Sei sulla pagina 1di 23

Biol. Rev. (2016), pp. 000–000.

1
doi: 10.1111/brv.12251

Relationships among multiple aspects


of agriculture’s environmental impact
and productivity: a meta-analysis to guide
sustainable agriculture
Richard N. German∗ , Catherine E. Thompson and Tim G. Benton
School of Biology, Faculty of Biological Sciences, University of Leeds, Leeds LS2 9JT, U.K.

ABSTRACT

Given the pressures on land to produce ever more food, doing it ‘sustainably’ is growing in importance. However,
‘sustainable agriculture’ is complex to define, not least because agriculture impacts in many different ways and it is
not clear how different aspects of sustainability may be in synergy or trade off against each other. We conducted a
meta-analysis to assess the relationships between multiple measures of sustainability using novel analytical methods,
based around defining the efficiency frontier in the relationship between variables, as well as using correlation analysis.
We define 20 grouped variables of agriculture’s impact (e.g. on soil, greenhouse gas, water, biodiversity) and find
evidence of both strong positive and negative correlations between them. Analysis based on the efficiency frontier
suggests that trade-offs can be ‘softened’ by exploiting the natural between-study variation that arises from a combination
of farming best practice and context. Nonetheless, the literature provides strong evidence of the relationship between
yields and the negative externalities created by farming across a range of measures.

Key words: sustainable agriculture, sustainable intensification, trade-offs, correlation, ecosystem services, food security,
efficiency frontier.

CONTENTS
I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
II. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
(1) Defining the scope of agricultural sustainability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
(2) Literature search and inclusion criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
(3) Results compilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
(4) Aggregation of response variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
(5) Measures of pairwise relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
(6) Meta-analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
(a) Meta-analysis: computing weighted averages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
(b) Meta-analysis: exploring between-study variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
III. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
(1) Strong associations between aspects of sustainability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
(2) Between-study variation in the associations between aspects of sustainability . . . . . . . . . . . . . . . . . . . . . . . . . 11
(3) Comparing Spearman’s rank correlation and choice potential scores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
IV. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
(1) Relationships among aspects of sustainability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
(a) Yield . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
(b) Biodiversity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
(c) Farm-scale monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

* Address for correspondence (Tel: +44(0)7843 739340; E-mail: richard_german11@hotmail.com).

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


2 R. N. German and others

(d) Farm management versus metric management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15


(e) Zero-tillage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
(2) Between-study variability in the relationships between aspects of sustainability . . . . . . . . . . . . . . . . . . . . . . 16
(3) Choice potential and correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
(4) Robustness of the meta-analysis of the literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
V. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
VI. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
VII. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
VIII. Supporting Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

I. INTRODUCTION on biodiversity, greenhouse gas emissions, water use, or even


a more generic definition of sustainability equating with
By the middle of this century, there will be about efficiency.
a third more people on the planet (UNDESA, 2013). But what ‘key indicators’ of sustainability should be
Additionally, on average, the world is getting richer, leading considered? There are a large number of potential axes which
to higher-calorie and more resource-intensive diets (Kearney, could measure the environmental impacts of management
2010). Projecting existing trends forwards suggests that, by and its economic sustainability. Many farms must import
2050, demand for food may increase 60–110% (Bruinsma, resources which may be non-renewable (e.g. rock phosphate,
2009; Tilman et al., 2011). Higher agricultural productivity fossil fuels), of limited or fluctuating availability (e.g. irrigation
may help to meet this demand, but in order to provide from groundwater and rivers), or subject to high and volatile
long-term food security agriculture must also be sustainable. prices (e.g. fuel, synthetic nitrogen fertilisers). The efficiency
That is, the way in which food is produced now must not of use of these resources affects farm profits, but the impact of
impair our ability to produce food, or any other function vital their production or diversion from other uses also contributes
for society, in the future (Brundtland, 1987). Maintaining the to a farm’s environmental footprint. Farm management
economic and social viability of farm businesses is a necessary affects ecosystem services (or disservices) produced by the
but not sufficient condition for sustainable agriculture. Each land, such as the regulation of water quality and flow
farm interacts with the rest of the world through use of rates, the balance of greenhouse gas emissions and their
external resources, and via flows of goods and services (or sequestration, control of acidifying or eutrophying gaseous
disservices). The environmental sustainability of agriculture emissions, the value of farm habitats for wildlife, and the
therefore needs to account for impacts (externalities) on aesthetic and recreational value of the landscape (Dale &
the environment, some at larger spatial scales than the farm Polasky, 2007; UK National Ecosystem Assessment, 2011).
itself. These externalities are considerable: agriculture utilises Management also influences a range of ecosystem services
nearly 40% of the earth’s surface (Foley et al., 2011); uses 70% supporting production at the farm scale, such as insect
of extracted water and, through agricultural run-off, pollutes pollination, soil fertility, and natural pest control (Zhang
rivers such that two-thirds of the world’s river basins are at et al., 2007). These externalities impact at many spatial and
or exceed capacity to buffer further nutrient input (Liu et al., temporal scales, making environmental accounting at the
2012); has degraded up to a quarter of the world’s agricultural scale of a farm business difficult. Management must also meet
soils (Grunwald, Thompson & Boettinger, 2011), leading ethical and welfare standards demanded by wider society,
to abandonment of land whilst new land is introduced to especially regarding treatment of workers and livestock.
agriculture via conversion of natural and semi-natural habitat Finally, if demand for food is constant or growing, it is
(Tilman et al., 2011); produces more greenhouse gases than also important to maintain yield per se (not merely profit),
any other sector apart from energy generation (Vermeulen, to avoid promoting indirect land-use change, and thus the
Campbell & Ingram, 2012); and is perhaps the biggest threat export of environmental costs to other locations.
to global biodiversity (Laurance, Sayer & Cassman, 2014; Given this wide range of important aspects of
Machovina & Feeley, 2014). sustainability, a key question (UK National Ecosystem
Unfortunately, accounting for externalities makes Assessment, 2011; Dicks et al., 2013) is ‘how are they
attempting to define ‘sustainable’ or ‘unsustainable’ farm related?’ If the impact of a new farm-management practice
management in absolute terms extremely difficult, because on all aspects of sustainability is positive then it implies
the same management intervention can have different that only a few need to be assessed to understand a
impacts depending on the local environmental context. farm’s overall sustainability. If, however, improving one
Instead, most approaches to sustainability or environmental measure leads to a decline or uncertain outcome in another
impact assessment aim to gauge change over time or between (e.g. increasing yield reduces biodiversity) it implies that
differing management practices, by monitoring desirable or (i) both need to be measured and (ii) informed decisions
undesirable changes in key social, economic, and biophysical are needed to decide on the balance between the two.
indicators (Saling & Kicherer, 2002; Gomez-Limon & Knowledge of relationships among aspects of sustainability
Sanchez-Fernandez, 2010), for example in terms of impacts may thus be of use in two ways: to help land managers to

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 3

understand constraints and potential for ‘win–wins’ among To complement this comprehensive analysis, we also
different aspects of sustainability; and to balance efficiency focus on one specific common management practice as
against quality when developing or applying sustainability a case study – zero-tillage agriculture. Zero-tillage (also
assessment methodologies. termed ‘no-tillage’ or ‘no-till’) agriculture involves planting,
While previous studies have explored trade-offs among tending and harvesting crops without ploughing or otherwise
multiple ecosystem services (often closely related to aspects disturbing the whole soil surface. Seeds are usually drilled
of sustainability) from correlations between two or more directly through residue on the soil surface, and mulches and
indicators mapped over large areas (e.g. Holland et al., cover crops in combination with herbicides are often used to
2011; Phalan et al., 2011; Maes et al., 2012; Maskell et al., control weeds. The resulting soil composition and structure
2013), they have not compared empirical responses of can differ markedly from that found in conventionally tilled
indicators to alternative management options in the same systems. Zero-tillage has gained popularity over the last half
place. This latter approach may better represent actual century, driven initially by the need to reduce high rates of
constraints in an agricultural setting, as the mechanisms soil erosion and conserve soil moisture, but with potential
underlying between-management correlations at the farm beneficial effects for development of soil biodiversity, carbon
scale may be different from those driving national-scale storage, and reduction in machinery costs (Lal, Reicosky &
spatial correlations. Other studies focus on demonstrated Hanson, 2007). However, the benefits of zero-tillage may
or likely direct interactions between ecosystem services (e.g. be contingent on the geological, climatic, socio-political, and
Pilgrim et al., 2010), or use expert estimation (e.g. Mouron agronomic context of a site. For example, a meta-analysis by
et al., 2012) or system models (e.g. Posthumus et al., 2010) of Manley et al. (2005) found that the soil carbon-storage benefits
the impacts of management scenarios on multiple aspects of conversion to zero-tillage were lower and costlier when
of sustainability. However, these tend to focus on a single growing wheat in the Great Plains area of the USA than for
type of agricultural system so inferences may not be valid for other crops in other areas. Here we use published literature
other systems. to summarise how zero-tillage management affects multiple
Here we conduct a meta-analysis of empirical studies aspects of sustainability, in comparison with conventional
from diverse agricultural systems, to synthesise what we tillage, and how the effects vary in different contexts.
know about how aspects of sustainability co-vary in response To summarise, this review aims to address the following
to management, and interpret what this may mean for questions: (i) which aspects of sustainability show negative
informing management decisions. Importantly, patterns of or positive associations and/or trade-offs with one another,
association between aspects of sustainability may differ when comparing alternative management options? (ii) How
with the management gradient in question (i.e. the type variable are the associations between aspects of sustainability
of management practice being varied in a study, such across different studies, and in which cases can variation
as fertilisation rate or tillage depth) or study context be predicted by contextual attributes of studies? (iii) Can
(e.g. local climate). The extent and predictability of this use of efficiency frontiers alongside correlation coefficients
context dependency may help to guide our interpretation show when management for multiple aspects of sustainability
of relationships, and highlight where caution is needed. We would prove most advantageous, compared to management
therefore aim to quantify the between-study variability of for a single objective?
pairwise relationships among aspects of sustainability, and
the degree to which this can be predicted by measurable
aspects of study context.
II. METHODS
A simple measure of association between variables is
the correlation coefficient. Correlations (and mean-trend
lines) provide information on the expected change in (1) Defining the scope of agricultural sustainability
other variables when management is optimised for one The term ‘sustainability’ is notoriously ambiguous, so in
variable. This approach has been used to infer constraints order to structure our analysis of relevant literature, we
for delivering multiple ecosystem services (Maskell et al., first developed a hierarchical classification of measures
2013). However, to quantify constraints when actively relevant to agricultural sustainability (Fig. 1). We first set
managing for multiple aspects of sustainability, a more out 13 broad themes (‘top-level’ themes hereafter): these
useful concept may be the ‘efficiency frontier’ (e.g. Nelson included ecosystem services flowing to and from farmland,
et al., 2009; Smith et al., 2012) – a line joining the cases resource-use efficiency, production and financial factors
where the value of one variable is at a maximum for a affecting business sustainability, and other issues such as
given value of another variable. The shape of this line welfare (Fig. 1, dark-grey boxes).
demonstrates the scope for satisfying multiple objectives Informed by a subset of literature, we divided each of the
(i.e. several aspects of sustainability) by choosing optimally top-level themes into ‘level 2’ categories. For example, our
among available options. We therefore explore the utility of top-level theme ‘Soil fertility’ was divided into ‘Soil nutrients’,
this concept alongside the analysis of correlation between ‘Soil biodiversity’, ‘Soil physical structure’ and ‘Soil carbon’
aspects of sustainability, to quantify the potential rewards of (Fig. 1, light-grey boxes). The level 2 categories served a dual
management for multiple, rather than single, objectives. purpose: (i) they formed the set of key words on which our

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


4 R. N. German and others

Worker Animal
Costs welfare welfare
Profit Yield quantity
Water flow
control Yield
Soil quality
moisture
Energy use
Water N efficiency
load Costs Welfare
Profit Yield
Water P Plant
load richness
Water Energy use
flow efficiency
Water pesticide Plant
load abundance
Water
Welfare
quality Costs Yield
Soil Plants Invertebrate
compactness richness
Profit Energy use
Soil efficiency
Soil physical
erosion Hydrology Invertebrate
structure
Invertebrates abundance
Soil N Biodiversity
content
Aspects of
Soil sustainability
Soil Vertebrate
fertility
Soil P nutrients Water-use Vertebrates richness
content efficiency
Pest
Soil Vertebrate
Soil K regulation abundance
carbon Nutrient-use
content Water-use
GHG efficiency
Pollination efficiency
emissions
Soil
Water-use
carbon Soil efficiency
biodiversity
Soil Nutrient-use
biodiversity efficiency
Nitrogen-use
Weeds efficiency
Total Animal GHG
weeds pests emissions
Pollination Phosphorus-use
Annual efficiency
weeds
CH4
Perennial emissions
Pest Natural
weeds CO2
abundance enemy N2O
emissions
abundance Pollination emissions

Fig. 1. Aggregation scheme of response variables. Schematic of the hierarchy used to group aspects of sustainability. White boxes
on the outside are the 37 factors representing biophysical variables recorded in studies with minimal aggregation (‘level 1’), light-grey
boxes show categories at the first level of aggregation (‘level 2’), and dark-grey boxes show the most aggregated ‘top-level’ themes.
This hierarchy of aggregation was an implicit part of several processes: defining the scope of the literature search, working from
top-level themes as a priori considerations; using level 2 categories to guide accept/reject decisions of journal articles, and aggregating
of level 1 variables into fewer categories during analysis, in order to reduce the complexity of the results. GHG, greenhouse gas.

literature search was built (see Section 4) and (ii) they formed matches and the zero-tillage search found 913 (see online
category labels, against which specific observed variables Appendix S1 for full details).
could be matched when deciding whether to include a study To be included in our analysis, a study had to
in our analysis. The specific variables observed in studies satisfy a number of criteria: (i) it must have studied an
were termed ‘level 1’ variables (Fig. 1, white boxes). agricultural system. (ii) It must have measured at least
two conceptually different response variables of relevance
to agricultural sustainability. We defined ‘conceptually
(2) Literature search and inclusion criteria different’ operationally as being related to different level
Our search used the set of level 2 categories as key 2 categories from the hierarchical relationships we devised
words to target studies where at least two variables of (Fig. 1). (iii) It must present results in a form yielding
relevance to ecosystem service provision or sustainability quantitative information (e.g. tables, graphs, or numbers
had been empirically measured in agricultural systems, in the text), or rank orders of treatments. (iv) It must have
across alternative agricultural management practices. We compared at least two places, treatments, or times that differ
searched the ISI Web of Knowledge, combining two strategies. meaningfully in agricultural management. Case studies of
We created a general search term accepting any agricultural sustainability against fixed standards such as life cycle analysis
crop, system or intervention, and a separate search for studies of a single system or location, for example, were rejected. (v)
testing the effect of zero versus conventional (and/or other) The response variables must be comparable across different
tillage practices. The general search found 2570 unique management treatments. For example, if in one treatment

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 5

yield of maize was recorded but in another yield of wheat, encountered, the ranks for that variable were entered into
these were deemed incomparable and the study rejected. the same column. If it represented a different quantity, we
(vi) Some attempt to replicate treatments and/or control for created a new column. Definitions of the indicators used can
confounding variables should have been made. be found in Table 1.
Following selection, 88 studies comprising 100 separate Finally, we characterised the environmental and
results were retained from the general search, and 55 studies agricultural context of each study, so that we could
including 64 separate results were kept from the zero-tillage explore how these factors influenced the relationships
search. The full list of selected studies can be found in between response variables. We recorded the following:
Table S1. crop/livestock yield type; crop rotation type; climatic zone
[IPCC broad categories (Bickel et al., 2006)]; soil texture
(3) Results compilation (USDA soil texture triangle); continent; duration of the study
(in years); whether it was a manipulative or correlative study;
From the selected studies we extracted information on the whether the treatments were categorical or continuous; and
variation in response variables, the variation in the predictor the management variables defining the treatment groups.
variables (such as management or climate), and information These attributes were deliberately general, so that the same
on the agricultural or environmental context. For each set could be recorded from as many studies as possible.
response variable measured in a study, we recorded the
mean value in every treatment group, and then ranked
the treatments by this mean value. We assumed that (4) Aggregation of response variables
as the treatment mean of a response variable increases, As level 1 variables were highly study specific, to perform
this corresponds to a monotonic increase or decrease in analyses we aggregated ranks within level 2 categories to
‘desirability’, from the perspective of a farmer or from allow higher-level comparisons and to increase the number
society as a whole. This approach is recommended for of studies in which any given pair of variables was measured.
constructing composite metrics, such as soil-quality indices To aggregate variables, we calculated an unweighted mean
(Seybold et al., 1997). The rankings assigned to a variable of the ranks of the level 1 variables belonging to each level
across treatment groups applied to this ‘desirability’ scale, 2 category (Fig. 1). We used unweighted means of ranks
rather than raw values. For example, higher crop yields were because we had no reliable method with which to weight
ranked higher (more desirable), whereas higher soil erosion different raw responses.
rates were ranked lower (less desirable). Where a unimodal
change in desirability might have been more appropriate
(5) Measures of pairwise relationships
(e.g. for soil pH), we did not have enough information about
the optimum value or symmetry of the change in desirability We described the relationship between each pair of response
either side of the optimum to confidently rank treatments, variables across management treatments or gradients within
so these variables were not included. The assigned rankings a study using two measures: (i) The Spearman’s rank
from each study were then used in further analysis. Details correlation. This communicates the likely change in a
of the direction ranked by desirability for each biophysical response, X , if management is targeted purely for another
or socioeconomic indicator are provided in Table 1. response Y , and is ‘blind’ to response X (Thomas et al., 2013).
The statistical significance of differences between groups (ii) The shape of the efficiency frontier. We quantified this
was not taken into account, but when values were as the area under the frontier, and we term this the ‘Choice
indistinguishable by eye on a graph, or at the precision potential’ (Fig. 2). It illustrates how much scope there is for
given in the text, a tied rank was assigned. compromises or synergies by ‘choosing’ management for X
In cases where explanatory variables were continuous, we and Y simultaneously, i.e. to what degree you can have a
used the fitted trends published in the paper to rank the ‘win–win’ situation.
values at the ends of the line (or corners of the plane), and In order to calculate choice potential, a pair of response
treated these as dummy treatment groups. variables (X and Y ) are ranked across treatment groups
Although information is lost by using ranks, doing so within a study according to desirability (Fig. 2A). When
does not assume that the scale at which a variable is plotted against one another, a convex hull drawn around the
measured is linearly related to some true physical magnitude, upper right margin of the cloud of points describes the best
as using measured units would. This conservative approach is available value of response X for a given value of response
especially useful when studies use different units to measure Y (or vice versa), and is called the ‘efficiency frontier’ (Smith
the same variable. It provides a basis to record results et al., 2012). The shape of this line indicates how severe the
consistently, and to average results across studies. trade-off between X and Y is for the set of management
We did not assume a priori the set of precise biophysical practices studied, with a diagonal line sloping down to the
and socioeconomic variables that we would record from right being the most severe trade-off when using ranks, and a
papers, but allowed the literature itself to demonstrate the box reaching the top right corner indicating no trade-off. For
most appropriate indicators (level 1 variables in Fig. 1) for meta-analysis, the shape of the efficiency frontier must be
the broader aspects of sustainability we had identified. If summarised as a single number, and the area under the line is
a variable measured in a study matched one previously one way to do this. Choice potential is calculated as the area

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


6 R. N. German and others

Table 1. Definitions and ranking directions of the variables used in the meta-analysis. Variable name shows each level 1 variable (see
Fig. 1), and the definition and ranking direction columns define how these were coded from the set of precise study-level indicators
which they encompass. The exact biophysical indicator measured often differed between studies, but if they were deemed to reflect
the same underlying attribute they were grouped together into the same level 1 variable. A ranking direction of ‘+’ or ‘−’ indicates
that higher values of the indicator were ranked as more or less desirable, respectively. N is the number of separate results in which
each level 1 variable was measured. GHG, greenhouse gas

Ranking
Level 1 variable Definition direction N
Yield Yield quantity Per unit area production of focal goods during + 130
study period
Yield quality Quality definition as defined in each source + 6
article. For example, grain protein content for
cereals
Profit Profit Total profit from production for each treatment + 39
during study period
Costs Costs Total outgoings for the production process during − 25
study period
Energy-use efficiency Energy-use Production per unit of energy used + 12
efficiency
Water-use efficiency Water-use Production per unit of evapotranspiration (ET). + 27
efficiency In non-irrigated systems estimated ET was
measured as precipitation plus loss of soil
water. In irrigated systems, it was either
irrigation alone, or precipitation plus irrigation
quantity plus loss of soil water
Nutrient-use efficiency N-use efficiency Production per unit of either applied nitrogen + 10
(agronomic efficiency) or nitrogen uptake
(physiological efficiency) depending on the
study
P-use efficiency Production per unit of either applied phosphorus + 2
(agronomic efficiency) or phosphorus uptake
(physiological efficiency) depending on the
study
Plants Plant richness Species richness of non-crop plants studied + 15
(arable systems), or all plants (pastoral systems)
Plant Abundance of non-crop plants studied + 5
abundance
Invertebrates Invertebrate Species richness of non-pest invertebrates studied + 9
richness
Invertebrate Abundance of non-pest invertebrates studied + 9
abundance
Vertebrates Vertebrate Species richness of vertebrate group(s) + 7
richness studied – predominantly birds
Vertebrate Abundance of vertebrate group(s) + 3
abundance studied – predominantly birds
Pollination Pollination Pollination of crop in question, according to + 2
measure used in study (visitation, pollen
transfer, seed set etc.)
Animal pests Pest abundance Severity of animal pest abundance or damage, as − 10
assessed in the study
Natural enemy Abundance or effectiveness of natural predators + 6
abundance of crop pests, depending on study
Weeds Perennial weeds Abundance of perennial weeds, as assessed by the − 1
study
Annual weeds Abundance of annual weeds, as assessed by the − 1
study
Total weeds Abundance of all weeds, as assessed by the study − 5
GHG emissions CH4 emissions Methane emissions from production process − 1
CO2 emissions CO2 emissions from production process − 6
N2 O emissions Nitrous oxide emissions from production process − 4
Soil biodiversity Soil biodiversity Abundance or diversity of groups of soil fauna or + 15
fungi, depending on the study

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 7

Table 1. Continued

Ranking
Level 1 variable Definition direction N
Soil carbon Soil carbon Either mass per area, or mass per mass of soil, of + 67
carbon or organic matter, depending on the
study
Soil nutrients Soil K content Available potassium content either in mass per + 22
area, or mass per mass of soil depending on
study
Soil P content Available phosphorus content either in mass per + 39
area, or mass per mass of soil depending on
study
Soil N content Available nitrogen content either in mass per area, + 48
or mass per mass of soil depending on study
Soil physical structure Soil erosion Mass of soil lost over study period − 17
Soil Soil bulk density, penetration resistance, or − 26
compactness inverse porosity depending on the study
Water quality Water pesticide Concentration, or frequency of detection in − 3
load drainage water, depending on the study
Water P load Concentration of the studied − 7
phosphorus-containing compound in drainage
water
Water N load Concentration of the studied nitrogen-containing − 17
compound in drainage water (varied among
studies)
Water flow Soil moisture Either available water capacity, or total water + 31
content (mass per mass, or volume per volume)
depending on the study
Water flow Either infiltration rate of water into soil, or run-off Infiltration +, 19
control coefficient, depending on the study run-off −
Welfare Worker welfare Welfare scale varied by study; composite index in + 4
most cases
Animal welfare Welfare scale varied by study; composite index in + 4
most cases

under the efficiency frontier, relative to the minimum and (6) Meta-analysis
maximum possible given the number of treatments in a study
After obtaining Spearman’s rank correlations and choice
(Fig. 2B). Choice potential can take a value between 0 (perfect
potential scores for pairs of aspects of sustainability in each
trade-off) and 1 (no trade-off). The measure is asymmetrically study, we performed a meta-analysis. One of the strengths
related to the correlation coefficient (Fig. 2C); high values of meta-analyses is the ability to average over the context
of choice potential are always produced by a strong positive specificity of individual studies, to produce a more-reliable
correlation, but at more negative correlations a wide range estimate of the true size of an effect, or relationship between
of choice potentials are possible. Due to the low number variables, than can be provided by any single study.
of treatment groups in some studies, choice potential could However, an average correlation or choice potential
not always be defined, so in these extreme situations, we between two aspects of sustainability may conceal potentially
applied two rules: (i) where all ranks from one response were informative variation within a heterogeneous set of studies
tied, a choice potential of 1 was assigned and (ii) where coming from varied farming systems and geographic
all ranks of both responses were tied (i.e. no variation in locations. Therefore, in addition to estimating an overall
either variable), choice potential was not calculated for that average pairwise relationship among any two aspects of
study. sustainability, we explored its variability and whether this is
For studies comparing zero-tillage (ZT) with conventional- systematic to extracted contextual variables.
tillage treatments (CT), we also calculated the mean rankings
of response variables in the two treatments, then calculated
the sign of the difference between the means (ZT–CT). (a) Meta-analysis: computing weighted averages
Thus, each response could have a value of either 1, 0 or −1, For each study we first transformed Spearman’s rank
corresponding to a desirable, neutral or undesirable effect of correlation (rs ) values to Fisher’s z scale (using Fisher’s
zero-tillage. This is a simple measure, but is less affected by z = arctanh rs ), and choice potential scores (CP) to a logit
study design than are more quantitative measures. scale [using logitCP = log(CP/[1 − CP])] (Fig. 3, step 3).

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


8 R. N. German and others

(A) Treatment Response A Response B


ID (rank) (rank)
S 1 2
T 2 1

U 3 7
V 4 8
W 5 5
X 6 3

Y 7 6
Z 8 4

1.0
(B) (C)
8

V
7

0.8
U
Treatment rank: response B
6

Choice potential (CP)


Y
W

0.6
5
4

0.4

Z
rs = 0.29 X
CP = 0.76
3

0.2

S
2

T
0.0
1

1 2 3 4 5 6 7 8 –1.0 –0.5 0.0 0.5 1.0


Treatment rank: response A Spearman's rank correlation (rs)

Fig. 2. Graphical explanation of the choice potential score. Correlation cannot describe the relationship fully between two variables.
With weak positive or even a negative correlation, data may nonetheless exist towards the top-right corner of a scatterplot, scoring
highly on both variables; we assess this using the choice potential. (A) Eight different treatments (S–Z) ranked for desirability
according to two response variables (A and B) in a hypothetical study. (B) These ranks are plotted, a convex hull is drawn from the
axis ends around the points (dashed line), and the area under the hull (AUH: dark-grey region) calculated. The diagonal running
from top left to bottom right shows the smallest possible AUH when using ranks. The light-grey area is the maximum possible AUH
if the same treatment ranked best for both variables. The choice potential is the proportional area of the dark-grey area relative to the
possible area above the diagonal. (C) The relationship between Spearman’s rank correlation and choice potential, from permuting
the relationship between two arbitrary response variables 500 times. The red point shows the hypothetical study from A.

Transformation is necessary when dealing with bounded in the relationship of interest across studies, by incorporating
measures in order to make variance equal across the whole between-study variation into the estimate of the overall
range of values. Correlations of exactly 1 or −1 were assigned mean and uncertainty around it (Borenstein et al., 2009).
values of 0.99 and −0.99, respectively, and choice potential When studies have heterogeneous designs and contexts this
scores of exactly 0 or 1 were assigned values of 0.01 and 0.99, is more appropriate than the fixed-effects method (which
respectively, to give finite values in the transformed units. assumes there is a single true relationship).
To compute the average (and confidence interval) of We assigned study weights inversely proportional to the
Spearman’s rank correlation and choice potential between variance around the study-specific Fisher’s z or logitCP
different aspects of sustainability from multiple studies, we (Borenstein et al., 2009), estimated using a permutation
developed an approach summarised in Fig. 3. For each approach. In theory, the expected standard error of Fisher’s
pair of aspects of sustainability, we calculated the weighted z depends only on the number of observations (N ; treatment

average of Spearman’s rank correlations and choice potential groups in a study in our case), as 1/ (N − 3). However,
scores between them from multiple studies. We used a for studies with N < 4 this expectation is undefined, and
random-effects approach, which accounts for real variation we had no theoretical basis on which to define an expected

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 9

Single study Multiple studies Multiple studies


Single pair of responses Single pair of responses Multiple pairs of responses

1. Rank treatments in a study 5. Calculate the weighted mean z or 6. Back-transform cross-study mean
according to different response logitCP, and uncertainty, over multiple and confidence interval into rs or CP
variables (as in Fig. 2A). studies. units, for chosen pair of responses.

Study 1 Responses
2. Calculate correlation (rs) or choice
Study 2 A B C . . .
potential (CP) across treatments in that
study,for a chosen pair of response Study 3 A
-0.2
(-0.7 :+0.4)
variables. Study 4

Responses
Study 5 B
Study 6
3. Transform value into Fisher’s z (for rs) C
or logit (for CP) scale.
Study 7
Study 8 .
.
Study 9
.
4. Estimate within-study variance of Study 10
Fisher’s z or logitCP, from a null Mean 7. Repeat steps 1 to 6 for other pairs
distribution simulated for the of responses, forming a pairwise
-10 -5 0 5 10
appropriate number of treatment matrix of mean rs and CP with
groups (N). Transformed units confidence estimate.

Fig. 3. Work flow for calculation of average Spearman’s rank correlation (rs ) and choice potential (CP) between different aspects of
sustainability, from multiple studies. Schematic diagram illustrating the process of calculating weighted mean rs or CP values across
sets of studies. The work flow within each blue box results in a different level of output: left-hand box produces a study-specific
estimate of (transformed) rs or CP between the desirability rankings of a pair of response variables across treatments within that study;
the central box produces a weighted average of the (transformed) rs or CP values from multiple studies where this pair of responses
was measured (even when the treatments differed among studies); and the process in the right-hand box results in a pairwise matrix
of associations (e.g. Fig. 4), which can be used in further multivariate analyses (see online Fig. S2). Sets of studies used to calculate
averages may comprise all studies where a pair of response variables were measured (as in Fig. 4), or particular subsets having similar
study attributes (Fig. 6, Table 2).

variance for logitCP. To solve this problem, we created ‘null’ For each pair of aspects of sustainability, we then used
distributions of Fisher’s z and logitCP for each value of N , by regression trees to find combinations of study attributes which
repeatedly calculating them from two uncorrelated variables could be used to segment studies into subsets with similar
permuted 20000 times. The distribution of Fisher’s z and correlation or choice potential values. Due to the presence of
logitCP resulting from these permutations was symmetrical binary predictors, we used the ‘ctree’ function from the ‘party’
with a mean of 0, so we could use the variance of these null package in R to fit tree models, which unlike other methods is
distributions to approximate the variance of all Fisher’s z and not biased against predictors with few factor levels (Hothorn,
logitCP values calculated from the same N . The standard Hornik & Zeileis, 2006). We used a conservative building
error of Fisher’s z calculated in this way was a good match to process to find the strongest sources of between-study
the theoretical standard error for N ≥ 4, and for an N of 2 or variation, only adding splits passing a significance threshold
3 followed the same qualitative pattern (see online Fig. S1D), of 0.95 (argument ‘mincriterion’ = 0.95). The results of this
suggesting that the use of this approximation was reasonable. exploratory process are presented in Table 2 (for choice
potential) and Table S2 (for Spearman’s rank correlation).
All statistical analysis and calculations were performed
(b) Meta-analysis: exploring between-study variation using R 3.0.0 (R Development Core Team, 2013).
In order to identify informative patterns of between-study
variation, we looked for subsets of studies which differed
systematically in their contextual attributes and had similar
III. RESULTS
Spearman’s rank correlation and choice potential scores. To
do this, we coded each management practice that varied in
a study as a binary variable: Y if this factor was tested, N (1) Strong associations between aspects of
sustainability
if it was not. This allowed us to characterise studies by the
combination of management practices which were tested, Many aspects of sustainability showed consistent positive or
as many studies crossed several treatment factors. We also negative Spearman’s rank correlations (rs ) with one another
classified the environmental context of the study as described across management treatments, even when aggregated
in Section 5. Together, the management and contextual across the diverse range of studies (Fig. 4). On average,
variables of the study provided the attributes by which we variation in yield (production per unit area) was strongly
segmented the studies into subsets. positively correlated with variation in profit accrued over

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


10 R. N. German and others

Table 2. Study attributes contributing to variation in choice potential between pairs of responses (level 1). Table describes results of
fitting regression-tree models to choice potential scores using study attributes as potential explanatory variables (‘Tree model’). Each
box represents a recursive partition of the set of scores, as a list of conditions and results. Each line in the list corresponds to a node
in the tree model, with numbers at the beginning of lines identifying a particular node, with ‘1)’ representing the root of the tree,
and other numbers referring to subsequent splits or terminal nodes. Statements in bold after a number show the splitting condition
applying to that node. Nodes marked with a star are terminal nodes. Terminal nodes show the weighted mean, confidence intervals
(in parentheses), and number of studies. Abbreviations: HORT, horticultural crops; C, cool; W, warm; TE, temperate; TR, tropical;
Y, yes the statement is true; N, no the statement is false
Explained
Variable 1 Variable 2 variance (%) Tree model
Yield (quantity) Water-use efficiency 81.7 1) Climate == C.TE.DRY, TR.DRY, W.TE.DRY, W.TE.MOIST
2)* 1.00 (0.97: 1), N = 18
1) Climate == C.TE.MOIST, TR.MOIST, TR.WET
3)* 0.79 (0.35: 0.96), N = 7
Yield (quantity) Soil carbon 38.0 1) Comparison = Conventional.vs.organic == N
2)* 0.94 (0.88: 0.97), N = 43
1) Comparison = Conventional.vs.organic == Y
3)* 0.53 (0.15: 0.88), N = 9
Yield (quantity) Soil N content 56.1 1) Comparison = Conventional.vs.organic == N
2) Continent == ASIA, EUROPE, N.AMERICA
3)* 0.97 (0.93: 0.99), N = 27
2) Continent == AFRICA, S.AMERICA
4)* 0.63 (0.25: 0.90), N = 4
1) Comparison = Conventional.vs.organic == Y
5)* 0.66 (0.13: 0.96), N = 7
Soil carbon Soil N content 70.0 1) Yield.type == CEREAL, HORT, LEGUMES, MAIZE, MIXED,
WOODY
2) Comparison = Residue.management == N
3)* 0.96 (0.91: 0.98), N = 30
2) Comparison = Residue.management == Y
4)* 0.90 (0.63: 0.98), N = 4
1) Yield.type == RICE
5)* 0.55 (0.03: 0.98), N = 5
Soil carbon Soil P content 24.8 1) Study type = Manipulative
2)* 0.87 (0.68: 0.95), N = 28
1) Study type = Correlative
3)* 0.25 (0.01: 0.88), N = 6
Soil N content Soil P content 63.9 1) Study type = Manipulative
2) Comparison = Fertiliser.regime == Y
3) Yield.type == CEREALS, RICE
4)* 1.00 (0.97: 1.00), N = 8
3) Yield.type == GRASS, MAIZE, OTHER CROP
5)* 0.98 (0.92:1.00), N = 3
2) Comparison = Fertiliser.regime == N
6)* 0.87 (0.59: 0.96), N = 18
1) Study type = Correlative
7)* 0.13 (0.00: 0.81), N = 5
Soil P content Soil K content 76.3 1) Comparison = Conventional.vs.organic == N
2)* 0.97 (0.94: 0.99), N = 18
1) Comparison = Conventional.vs.organic == Y
3)* 0.18 (0.00: 1.00), N = 4
Costs Profit 47.2 1) Duration <= 2yrs
2) Comparison = Irrigation.strategy == Y
3)* 0.84 (0.51: 0.96), N = 4
2) Comparison = Irrigation.strategy == N
4)* 0.14 (0.01: 0.67), N = 10
1) Duration > 2yrs
5)* 0.94 (0.69: 1.00), N = 11

the study period (rs = 0.67, C.I. = 0.32: 0.86), soil carbon of irrigation or irrigation plus ET; rs = 0.86, C.I. = 0.67:
storage (rs = 0.61, C.I. = 0.27: 0.82), soil nutrients (N, P 0.94), and weed control (in terms of reduced weed
and K content; rs = 0.47, C.I. = 0.21: 0.66), water-use abundance; rs = 0.95, C.I. = 0.68: 1). However, high yield
efficiency (in rain-fed systems measured as yield per unit was associated with poorer water quality (N, P and pesticide
of evapotranspiration (ET), in irrigated as yield per unit loading; rs = −0.74, C.I. = −0.95: −0.04), and higher total

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 11

production costs (rs = –0.52, C.I. = –0.78: –0.12). There related, with a stronger trade-off in Africa and South America
was also a well-supported negative correlation between (CP = 0.63 C.I. = 0.25: 0.90) than in Europe, Asia or North
profit and soil physical quality (compactness and erosion; America (CP = 0.97, C.I. = 0.93: 0.99) (Table 2).
rs = −0.78, C.I. = −0.97: −0.06). Detailed exploratory analysis of the context dependency
Although less strongly, yield tended to be negatively of zero-tillage agriculture identified two study attributes
associated with measures of species’ diversity and abundance. with very strong effects. First, the response of soil carbon,
Negative correlation was strongest for plants (abundance soil nutrients and soil physical structure differed markedly
and diversity combined; rs = −0.60, C.I. = −0.93: 0.27), between studies in different climatic zones. The first two
weaker for invertebrates (rs = −0.43, C.I. = −0.89: 0.48), showed negative responses in wet tropical regions, but largely
and weakest for vertebrates (rs = −0.17, C.I. = −0.92: 0.84). positive responses elsewhere (Fig. 7). Soil physical structure
By contrast, profit tended to be positively correlated with responded negatively or inconsistently everywhere except in
measures of biodiversity, perhaps reflecting that lower yields tropical dry climates, where a consistent positive effect was
in biodiversity-friendly treatments were often offset by lower seen. Additionally, the effect of zero-tillage on yield also
production costs (rs = 1.00 and 1.00 for costs versus plant and varied with climate, with positive effects in moist-temperate
invertebrate diversity, respectively). Profit was also positively zones, but inconsistent or negative effects in other climates
correlated with soil biodiversity (rs = 1.0, C.I. = 0.41: 1). It (Fig. 7).
is likely that the length of the study affects this relationship, The second attribute that influences the impact of
as long-term profit is seen to be higher in studies favouring zero-tillage on yield was the duration of the manipulation.
good soil quality. Most studies running for 5 years or more showed a positive
Water quality and flow characteristics (nutrient and effect of zero-tillage on yield, compared to a negative effect
pesticide loading, run-off coefficient or infiltration rate) in a greater proportion of shorter studies (see online Fig. S2).
from agricultural land tended to be positively associated When comparing studies measuring maize yield with those
with measures of soil quality and resource-use efficiency, involving other cereals, little difference in the impact of
indicating that managing to protect watersheds is likely to zero-tillage for multiple aspects of sustainability was observed
benefit long-term soil fertility, and vice versa. See Table 1 for (see online Fig. S3).
full definitions of the variables as used in the analysis.
(3) Comparing Spearman’s rank correlation and
(2) Between-study variation in the associations choice potential scores
between aspects of sustainability
Choice potential scores summarise the shape of the
The mean Spearman’s rank correlations (rs ) and choice efficiency frontier: the potential scope for optimising between
potential (CP) scores (Fig. 4) across all studies mask substantial variables. Choice potential scores for pairs of aspects
between-study variation (Fig. 5). For some pairs of variables, of sustainability were typically much larger than the
such as yield and soil N content, or soil carbon and soil K corresponding Spearman’s rank correlation, indicating that
content, the majority of Spearman’s rank correlation and trade-offs may be less severe, and ‘win–wins’ more likely than
choice potential values calculated from individual studies lay correlations might suggest (Fig. 4). Considering only pairs
within a fairly narrow range. For many other pairs however, of responses measured in 10 studies or more, the greatest
values spanned almost the entire possible range. This high average discrepancies between Spearman’s rank correlation
between-study variation contributed to the wide confidence coefficients (rs ) and choice potential scores (CP) occurred
intervals observed for cross-study average values (Fig. 4). between yield and water quality (rs = −0.74, CP = 0.57),
In some cases, the reported study context could explain a yield and plant diversity (rs = −0.60, CP = 0.48), yield and
significant amount of between-study variation in Spearman’s costs (rs = −0.52, CP = 0.50), soil physical structure and soil
rank correlation (see online Table S2) and choice potential nutrients (rs = 0.33, CP = 0.92), and soil physical structure
(Table 2). Climatic zone accounted for 81.7% of the variation and soil carbon (rs = 0.21, CP = 0.87).
in choice potential between crop yield and water use
efficiency, which was much higher in dry climates (CP = 1.00,
C.I. = 0.97: 1) than in wetter ones (CP = 0.79, C.I. = 0.35:
0.96) (Table 2), and choice potential between soil carbon IV. DISCUSSION
and soil N was much lower in studies involving rice than
in those involving other crops. In other cases, given the low We carried out a comprehensive meta-analysis to explore
sample size of studies, no significant explanatory factors were the relationship between different aspects of sustainability
identified. Regression trees also identified several pairs of in agriculture, comparing across alternative management
aspects of sustainability for which studies comparing whole practices for a range of farming systems around the world.
farming systems (typified by ‘conventional versus organic’ Many aspects were positively correlated with each other,
comparisons) produced a different relationship than studies suggesting that these aims are generally compatible. Equally,
focusing on other management gradients (Fig. 6; Table 2 other pairs of variables showed negative correlations, where
and see online Table S2). Study location, more than climate, managing to improve one is likely to be detrimental
was found to affect how crop yield and soil N content were to the other. In this study we also went beyond using

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


12 R. N. German and others

Variable 2

Nutrient-use efficiency
Soil physical structure

Energy use efficiency


Water-use efficiency

GHG emissions
Soil biodiversity

Invertebrates
Water quality
Soil nutrients

Animal pests
Vertebrates
Soil carbon
Water flow

Pollination
(A)

Welfare
Weeds
Plants

Costs

Profit
Yield
0 0.54 0.31 0.21 0.33 0.54 0.41 0.72 0.65 0.96 1 -1 -0.35 -0.78
Soil physical structure (41) (21) (5) (18) (11) (6) (30) (3)
-
(3)
- - - -
(2) (1) (1)
-
(6) (8)
0.91 0 0.85 0.89 0.33 0.5 0.23 0.74 0.44 0.05 0.24 -0.38 1 0.81 -0.53
Water flow (21) (47) (2) (15) (8) (1) (40) (9) (1) (2)
- - -
(3) (2) (1)
- -
(2) (4)
0.86 0.95 0 0.98 0.81 0.25 0.3 1 0.41 0.6 0.65 1 0.8 1 1
Soil biodiversity (5) (2) (15) (9) (7) (3) (13) (1) (1) (2) (3) (1)
- - - -
(1)
-
(2) (2)
0.87 0.97 0.98 0 0.64 0.3 0.61 0.73 0.66 -0.05 0.8 1 -1 -1 0.7 -0.87 -0.38 0.65
Soil carbon (18) (15) (9) (67) (44) (4) (52) (7) (7) (3) (1)
- -
(1) (1) (3) (6) (2) (8) (10)
0.92 0.89 0.98 0.95 0 -0.16 0.47 -0.03 0.75 0.56 0.32 0.05 -1 -0.02 0.28 0 0.22 -0.03
Soil nutrients (11) (8) (7) (44) (53) (6) (42) (7) (6) (2) (1)
- -
(2) (1) (2) (2) (1) (4) (7)
0.95 0.75 0.95 0.81 0.92 0 -0.74 0.87 1 0.99 1 1 -0.75 0.73 0.7 0.43 0.04
Water quality (6) (1) (3) (4) (6) (19) (14) (2) (1) (3) (1)
- -
(2)
-
(3) (3) (3) (6) (8)
0.88 0.85 0.84 0.92 0.94 0.57 0 0.86 0.54 -0.6 -0.43 -0.17 0.4 0.27 0.95 0.13 -0.71 0.47 -0.52 0.67
Yield (30) (40) (13) (52) (42) (14) (131) (25) (8) (10) (8) (6) (1) (9) (6) (7) (6) (3) (19) (27)
0.95 0.95 1 0.91 0.86 1 0.98 0 0.59 1 0.06 0.42
Water-use efficiency (3) (9) (1) (7) (7) (2) (25) (27) (3)
- - - -
(1)
- - - -
(6) (11)
1 0.91 0.91 0.95 1 0.87 0.95 0 1 1 1 -0.87 0
Nutrient-use efficiency -
(1) (1) (7) (6) (1) (8) (3) (10)
- - - -
(1)
- -
(1) (1) (2) (3)
Variable 1

0.98 0.73 1 0.89 0.97 1 0.48 0 0.4 0.86 0.69 0 1 0.5 1 0.81
Plants (3) (2) (2) (3) (2) (3) (10)
- -
(16) (7) (5)
-
(3)
-
(1) (1) (1) (2) (4)
0.79 1 0.67 1 0.81 0.9 0 0.42 0.97 0.61 1 1 0.49
Invertebrates - -
(3) (1) (1) (1) (8)
- -
(7) (14) (5) (2) (5)
- -
(2)
-
(1) (3)
1 0.79 1 0.92 0 0.82 0.87
Vertebrates - -
(1)
- - -
(6)
- -
(5) (5) (7)
-
(3)
- - - - -
(1)
1 1 0 0.4
Pollination - - - - - -
(1)
- - -
(2)
-
(2) (1)
- - - - - -

0.89 1 0.73 1 0.88 1 1 1 0.89 1 0.78 0 -1 -1


Animal pests -
(3)
-
(1) (2) (2) (9) (1) (1) (3) (5) (3) (1) (13)
- - - -
(1) (1)
1 0.81 0 0.2 0.98 0
Weeds (2) (2)
-
(1) (1)
-
(6)
- - - - - - -
(6)
- - - - -

1 1 0 0.7 0.29 0.59 0.63 0 0.12 -0.7 1 -1


GHG emissions (1) (1)
-
(3) (2) (3) (7)
- -
(1)
- - - - -
(8) (4) (2) (1) (1)
0 0.89 0.9 0.81 0.95 0.61 1 1 1 0.89 0 0.15 0.22 0.79
Energy use efficiency (1)
-
(1) (6) (2) (3) (6)
-
(1) (1) (2)
- - - -
(4) (12) (6) (5) (7)
0.5 0.94 0.97 1 1 0.44 0.88 0 -0.96 0.35
Welfare - - -
(2)
-
(3) (3)
-
(1) (1)
- - - - -
(2) (6) (6) (2) (4)
0.67 1 1 0.38 0.86 0.92 0.5 0.91 0.5 1 1 0 1 0.59 0.62 0 0.29
Costs (6) (2) (2) (8) (4) (6) (19) (6) (2) (2) (1)
- -
(1)
-
(1) (5) (2) (25) (25)
0.3 0.69 1 0.84 0.96 0.92 0.94 0.88 0.82 0.86 0.95 1 0 0 0.93 0.92 0.78 0
Profit (8) (4) (2) (10) (7) (8) (27) (11) (3) (4) (3) (1)
-
(1)
-
(1) (7) (4) (25) (39)

Fig. 4. (A) Cross-study averages of the between-treatment Spearman’s rank correlation and choice potential of desirability rankings,
for pairs of sustainability indicators. Each cell shows the weighted average, across all studies, of the Spearman’s rank correlation
(upper-right region) and choice potential score (bottom-left region) calculated in every study for that pair of variables using their
desirability rankings across treatments. The number of studies (N ) from which the average is calculated is shown in brackets. A
positive correlation indicates that, on average, more desirable values of variable 1 are found in treatments with more desirable values
of variable 2. More desirable rankings of a variable may however correspond to lower values of the indicator, in the case of ‘Costs’
for example (see Table 2 for full details of ranking directions). Cell colour is scaled for correlation and choice potential separately
(see legend), to aid visual interpretation. White cells with dashes indicate that these variables were not measured together. The
reduction in red/orange in the bottom-left versus the upper-right indicates that in some cases negative correlations nonetheless have
some ‘win–win’ data (as estimated by the choice potential scores) and there is thus potential for management to soften the trade-off.
(B) Confidence intervals for the weighted average correlation coefficients and choice potentials in A. The 2.5% (before colon) and
97.5% (after colon) confidence limits are shown. Cell colour scaled to size of confidence range (dark blue, small; light blue, large).
GHG, greenhouse gas.

measures of correlation, deriving a novel measure – choice While some pairs of variables showed a strong and
potential – which characterises trade-offs’ multiple objectives consistent relationship across studies, others were highly
in a way that correlations do not. Use of this measure variable. Crucially we showed that, where possible, splitting
highlighted that there was sometimes significant scope to these highly variable studies into subgroups with a similar
optimise management and find ‘win–win’ solutions even context helped to explain a high proportion of this variation.
when correlations were neutral or negative. Thus we were able to summarise global trends whilst still

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 13

Variable 2

Nutrient-use efficiency
Soil physical structure

Energy use efficiency


Water-use efficiency

GHG emissions
Soil biodiversity

Water quality

Invertebrates
Soil nutrients

Animal pests
Vertebrates
(B)

Soil carbon
Water flow

Pollination

Welfare
Weeds
Plants

Costs

Profit
Yield
-0.09: -0.87: -0.41: -0.35: -0.68: -0.15: -0.78: -0.86: -0.63: -0.99: -1: -0.97: -0.97:
Soil physical structure 0.86 0.96 0.7 0.78 0.97 0.78 1
-
1
- - - -
1 1 0.99
-
0.86 -0.06
0.77: -0.44: 0.68: -0.47: -0.99: -0.22: 0.04: -0.58: -0.87: -0.56: -1: -0.99: -0.48: -0.96:
Water flow 0.97 1 0.96 0.83 1 0.6 0.95 0.92 0.89
- - -
0.81 0.98 1
- -
1 0.68
0.12: 0.34: 0.93: 0.13: -0.94: -0.62: -0.99: -0.6: -0.81: -0.86: -0.99: -0.69: 0.41: 0.41:
Soil biodiversity 1 1 1 0.97 0.98 0.87 1 0.92 0.99 1 1
- - - -
1
-
1 1
0.61: 0.91: 0.92: 0.37: -0.92: 0.27: -0.44: 0.08: -0.86: -0.69: -0.99: -1: -1: -0.56: -1: -0.95: -0.55:
Soil carbon 0.96 1 1 0.81 0.98 0.82 0.98 0.91 0.84 1
- -
1 0.99 0.33 0.98 0.98 0.75 0.97
0.73: 0.63: 0.82: 0.91: -0.95: 0.21: -0.72: 0.27: -0.77: -0.93: -1: -0.92: -0.91: -0.89: -0.58:
Soil nutrients 0.98 0.98 1 0.97 0.91 0.66 0.69 0.93 0.98 0.98
- - -1: 1
0.99 0.91 0.97
-1: 1
0.95 0.54
0.56: 0.26: 0.11: 0.32: -0.95: -0.98: -0.99: 0.26: -0.99: -0.77: -0.99: -0.48: -0.54: -0.84: -0.86:
Water quality 1
0: 1
1 1 1 -0.04 1 1 1 1
- -
1
-
0.55 0.98 0.98 0.97 0.88
0.69: 0.69: 0.48: 0.84: 0.88: 0.16: 0.67: -0.34: -0.93: -0.89: -0.92: -0.91: -0.32: 0.68: -0.49: -0.96: -0.82: -0.78: 0.32:
Yield 0.96 0.93 0.97 0.96 0.97 0.9 0.94 0.91 0.27 0.48 0.84 0.98 0.71 1 0.65 0.23 0.98 -0.12 0.86
0.16: 0.77: 0.01: 0.52: 0.22: 0.15: 0.95: -0.38: -0.99: -0.73: -0.36:
Water-use efficiency 1 0.99 1 0.99 1 1 0.99 0.94
- - - -
1
- - - -
0.78 0.85
0.81: 0.31: 0.65: 0.78: 0.01: 0.56: 0.55: -0.99: -0.99: -0.99: -1:
Nutrient-use efficiency -
1 1 0.98 0.99 1 0.97 1
- - - -
1
- -
1 1 0.98
-1: 1
Variable 1

0.53: 0.11: 0.66: 0.32: 0.46: 0.54: 0.09: -0.51: -0.15: -0.71: -0.45: -0.99: -0.77: -0.85:
Plants 1 0.98 1 1 1 1 0.89
- -
0.89 1
-
0.99
- -1: 1
1 1 1 1
0.06: 0.55: 0.02: 0.01: 0.32: 0.5: -0.76: 0.58: -0.44: 0.68: -0.99: -0.8:
Invertebrates - -
1 1 1 1 0.97
- -
0.99 0.96 1 0.96
- -
1
-
1 0.97
0.01: 0.16: 0.84: 0.32: -0.51: -0.95:
Vertebrates - -
1
- - -
0.99
- -
1 1
-
1
- - - - -
1
0.55: 0.82: -0.91:
Pollination - - - - - -
1
- - -
1
-
0.98
- - - - - -

0.47: 0.01: 0.15: 0.53: 0.01: 0.01: 0.73: 0.44: 0.73: 0.04: -1: -1:
Animal pests -
0.99
-
1
0: 1
1 0.98 1 1 1 0.99 1 1
- - - -
0.99 0.99
0.38: 0.03: 0: 0.87:
Weeds 1 1
-
0.99
0: 1 -
1
- - - - - - - - - - - -

0.01: 0.01: 0: 0.01: 0.07: -0.88: -0.99: -0.99: -1:


GHG emissions 1 1
-
0.65
0: 1
0.92 0.96
- - 0: 1 - - - - -
0.93 0.66 1 0.99
0: 0.09: 0.22: 0.08: 0.46: 0.06: 0.01: 0.17: 0.66: 0.17: -0.8: -0.95: -0.49:
Energy use efficiency 0.99
-
1 1 1 1 0.97
-
1 1 1
- - - -
1 0.89 0.98 1
0.41: 0.58: 0.01: 0.17: 0.02: 0.36: -1: -0.91:
Welfare - - - 0: 1 -
1 1
-
1 1
- - - - -
0.97 0.99 0.63 0.98
0.05: 0.73: 0.56: 0.04: 0.17: 0.32: 0.24: 0.64: 0.15: 0.01: 0: 0.01: 0.04: -0.34:
Costs 0.99 1 1 0.9 1 1 0.76 0.98
0: 1
1 1
- -
0.99
-
1 0.98
0: 1
0.74
0.03: 0.06: 0.56: 0.24: 0.82: 0.47: 0.84: 0.53: 0.01: 0.27: 0.16: 0.17: 0: 0: 0.39: 0.23: 0.52:
Profit 0.85 0.99 1 0.99 1 1 0.98 0.98 1 1 1 1
-
0.99
-
0.99 1 1 0.92

Correlation coefficient Range of correlation coefficient C.I.


-1.00 -0.50 0.00 0.50 1.00 0.00 0.50 1.00 1.50 2.00

0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00

Choice potential Range of choice potential C.I.

Fig. 4. Continued.

allowing consideration of more local influences. Below, we other aspects of agricultural sustainability. In fact, on average
discuss these results in more detail, broader issues related to yield tends to be positively correlated with direct indicators
the sustainability of the whole food system, and the limitations of ecosystem services supporting production, such as facets
of our methods. of soil quality and resource-use efficiency (with the notable
exception of energy-use efficiency). We showed that on
(1) Relationships among aspects of sustainability average, management with higher yield tends to be associated
with higher profit, but also consistently higher costs of
(a) Yield production, implying that the price of failure is higher in the
Agricultural yield (harvested product per unit area) and short term. This further implies that there is the potential for a
profit (per hectare) are primary drivers of decisions at trade-off between the expectation of profits and its variability
the farm scale, and therefore a critical consideration in as shown by Abson, Fraser & Benton (2013). Unfortunately,
sustainable farm management. As a reflection of this, yield our literature search strategy found few studies explicitly
was measured in 131 out of the 164 results considered, so quantifying variability in yield or profit, or even providing
it was possible to examine closely the associations between disaggregated data from which a variability measure could
yield and other aspects of sustainability. Our results suggest be calculated. This should be an important focus of future
that yield is not overwhelmingly negatively associated with studies.

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


14 R. N. German and others

Spearman's rank correlation (r s ) Choice potential (CP )


Pair of response variables
-1.0 -0.5 0.0 0.5 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Response A Response B N
Soil carbon Yield (quantity) 52
Soil carbon Soil N content 39
Soil N content Yield (quantity) 38
Soil N content Soil P content 34
Soil carbon Soil P content 34
Soil P content Yield (quantity) 33
Soil moisture Yield (quantity) 31
Yield (quantity) Prof it 26
Costs Prof it 25
Yield (quantity) Water-use ef f iciency 25
Soil P content Soil K content 22
Soil compactness Yield (quantity) 22
Soil N content Soil K content 20
Yield (quantity) Costs 19
Soil K content Yield (quantity) 19
Soil carbon Soil K content 19
Soil compactness Soil carbon 16
Soil biodiversity Yield (quantity) 13
Yield (quantity) Water f low control 12
Soil compactness Soil moisture 12
Water-use ef f iciency Prof it 11
Water N load Yield (quantity) 11
Yield (quantity) Plant richness 10
Soil carbon Prof it 10
Soil compactness Soil N content 10

-1.0 -0.5 0.0 0.5 1.0 0.0 0.2 0.4 0.6 0.8 1.0

Fig. 5. Between-study variation in Spearman’s rank correlation and choice potential scores for pairs of level 1 variables.
Box-and-whisker plots showing the distribution of Spearman’s rank correlation (left) and choice potential scores (right) between
studies, for the desirability rankings of pairs of level 1 response variables measured together in 10 or more studies. Response A and
Response B columns specify the pair of responses corresponding to each row. N shows the number of studies measuring this pair
together. Boxes show the interquartile range, thick black lines show the median, whiskers show the most extreme points within 1.5
box lengths from the box edge, and empty circles show outliers.

(b) Biodiversity The negative correlations between yield and plants and
invertebrates may also be a desired outcome of management,
Biodiversity is often overlooked in farm-based decisions;
if plant and invertebrate biodiversity is also correlated
partly because primary drivers are crop yield or farm profit,
with the disservices associated with pests (illustrated by the
but also because the relationship between these primary
negative correlation between a farmer’s costs, and hence
drivers and biodiversity is difficult to measure. This is
profit, and pest species). However, where studies focused
because biodiversity is such a broad term, encompassing
on species that directly contribute to agriculturally useful
all species’ richness and abundance. Negative correlations
ecosystem services, such as pollinators and earthworms,
were observed between yield and plant, invertebrate
there was generally a positive correlation and high choice
and vertebrate biodiversity. Interestingly, the strength of potential with yield.
the trade-off is related to the scale of movement of Given that different taxonomic groups may respond
the groups (plants < invertebrates < vertebrates), suggesting to management differentially, and this response may also
that (i) more sessile groups are more vulnerable to depend on the scale of assessment (Gabriel et al., 2010) it is
intensive agriculture, or (ii) the scale of assessment differs difficult to simplify biodiversity into a single metric.
among groups – a farmland bird may utilise the resources
at a landscape scale, whereas a plant may be more
affected by management at the field scale (Gabriel et al., (c) Farm-scale monitoring
2010). Studies incorporating biodiversity and agronomic Measuring variables such as species richness and abundance
variables at large scales are rare, which is reflected of wildlife is time-consuming and costly, so the presence
in the wide confidence intervals around the cross-study of strong and consistent relationships will help to guide
average Spearman’s rank correlation and choice potential farmers and food retailers to use more readily and
(Fig. 4B). cheaply measured proxies. The recent reforms of the EU’s

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 15

Type of management compared Type of management compared


Conventional Cultivation Fertiliser Irrigation Conventional Cultivation Fertiliser Irrigation
Crop rotation Crop rotation
vs . organic regime regime regime vs . organic regime regime regime

Variable r s with Yield CP with Yield


0.76 0.76 0.83 -1 0.9 0.97 1 0
Profit (-0.22: 0.98, N =11) (0.41: 0.91, N =9)
-
(0.62: 0.93, N =5) (-1: -0.68, N =2) (0.49: 0.99, N =11) (0.88: 1, N =9)
-
(0.96: 1, N =5) (0: 0.34, N =2)
0.47 0 0.97 1
Welfare (-0.82: 0.98, N =3)
- - -
(-0.96: 0.96, N =1) (0.58: 1, N =3)
- - -
(0.55: 1, N =1)
-0.22 0.76
Vertebrates (-0.95: 0.87, N =3)
- - - -
(0.05: 1, N =3)
- - - -

-0.32 0 0.62 0.65 0.36 0.74 1 1


Animal pests (-0.97: 0.9, N =2) (-0.88: 0.88, N =2) (-0.39: 0.95, N =1)
-
(-0.1: 0.93, N =2) (0.01: 0.97, N =2) (0.11: 0.98, N =2) (0.81: 1, N =1)
-
(0.9: 1, N =2)
-0.32 0.54 0.52 0.04 0.62 0.87 0.94 0.94 0.92 0.89
Soil nutrients (-0.85: 0.54, N =8) (0.19: 0.76, N =20) (0.24: 0.72, N =11) (-0.96: 0.96, N =2) (0.19: 0.85, N =4) (0.5: 0.98, N =8) (0.84: 0.98, N =20) (0.85: 0.97, N =11) (0.39: 1, N =2) (0.61: 0.98, N =4)
-0.42 -1 0.36 0 0.55 0 0.94 1
GHG emissions (-0.96: 0.79, N =4) (-1: 0.99, N =1) (-0.35: 0.81, N =1)
-
(-0.96: 0.96, N =1) (0.01: 1, N =4) (0: 0.99, N =1) (0.55: 1, N =1)
-
(0.55: 1, N =1)
-0.47 0.76 0.75 -0.02 0.53 0.96 0.95 0.48
Soil carbon (-0.89: 0.37, N =9) (0.38: 0.92, N =34) (0.25: 0.93, N =12)
-
(-0.83: 0.82, N =2) (0.15: 0.88, N =9) (0.9: 0.98, N =34) (0.83: 0.99, N =12)
-
(0.02: 0.98, N =2)
-0.47 0.75
Invertebrates (-0.97: 0.77, N =4)
- - - -
(0.18: 0.98, N =4)
- - - -

-0.54 0.67 0.95 0.45 0.91 1


Nutrient-use efficiency (-0.86: 0.09, N =3) (-0.23: 0.95, N =5) (0.38: 1, N =1)
- -
(0.1: 0.86, N =3) (0.62: 0.98, N =5) (0.73: 1, N =1)
- -

-0.67 0 0.37 0.9


Plants (-0.99: 0.76, N =5)
-
(-0.88: 0.88, N =2)
- -
(0.03: 0.93, N =5)
-
(0.07: 1, N =2)
- -

-0.71 0 0.61 1
Energy-use efficiency (-0.96: 0.23, N =6)
- - -
(-0.96: 0.96, N =1) (0.06: 0.97, N =6)
- - -
(0.55: 1, N =1)
-0.78 0.75 0.93 0.33 0.95 1
Soil biodiversity (-0.96: -0.18, N =7) (-0.4: 0.98, N =6) (0.64: 0.99, N =3)
- -
(0.06: 0.79, N =7) (0.68: 1, N =6) (0.9: 1, N =3)
- -

-0.78 -0.87 0 -0.32 0.4 0.5 1 0.71


Water quality (-0.98: 0.12, N =7)
-
(-1: 0.72, N =2) (-1: 1, N =1) (-0.97: 0.9, N =2) (0.04: 0.92, N =7)
-
(0.01: 0.99, N =2) (0.01: 1, N =1) (0: 1, N =2)
-0.81 -0.32 -0.51 -0.8 0.13 0.69 0.68 0.33
Costs (-0.99: 0.32, N =9) (-0.87: 0.58, N =6)
-
(-0.89: 0.3, N =3) (-1: 0.69, N =1) (0.01: 0.69, N =9) (0.2: 0.95, N =6)
-
(0.29: 0.92, N =3) (0: 0.98, N =1)
-0.92 0.68 0.59 1 0.08 0.95 0.88 1
Soil physical structure (-1: 0.4, N =3) (0.23: 0.89, N =25) (-0.56: 0.96, N =4)
-
(-0.99: 1, N =1) (0: 0.81, N =3) (0.87: 0.98, N =25) (0.58: 0.98, N =4)
-
(0.01: 1, N =1)
-1 0.91 0.52 0.75 0.93 0 0.99 0.97 0.97 0.99
Water-use efficiency (-1: 0.99, N =1) (0.79: 0.96, N =14) (0.08: 0.79, N =3) (0.1: 0.95, N =7) (0.4: 1, N =3) (0: 0.99, N =1) (0.96: 1, N =14) (0.88: 1, N =3) (0.88: 1, N =7) (0.84: 1, N =3)
0.41 -0.38 1 -0.4 0.91 0.69 1 0.52
Water flow -
(-0.01: 0.7, N =34) (-0.95: 0.77, N =4) (0.6: 1, N =1) (-0.86: 0.4, N =3)
-
(0.82: 0.96, N =34) (0.05: 0.99, N =4) (0.55: 1, N =1) (0.11: 0.91, N =3)
0.95 1 0.98 1
Weeds -
(0.68: 1, N =6)
- -
(0.84: 1, N =1)
-
(0.87: 1, N =6)
- -
(0.73: 1, N =1)

Fig. 6. Cross-study averages of the between-treatment Spearman’s rank correlation (rs ) and choice potential (CP) of the desirability
ranking of yield with other variables, broken down by management factor studied. Each cell contains the weighted average of
Spearman’s rank correlation and choice potential (as in Fig. 4), calculated across subsets of studies investigating similar management
factors. Confidence intervals for the coefficients are also given in each cell, with the number of studies (N ) comparing the variables
under each management factor. For example, across the 11 studies comparing organic and conventional treatments, the mean
within-study correlation between yield and profit was 0.76, (where across all studies regardless of management the correlation was
0.67, see Fig. 4). For the five studies comparing different irrigation treatments, the mean within-study correlation between yield and
profit was 0.83. Cells are colour-coded to aid visual interpretation. White cells indicate that no studies compared yield with this
variable. GHG, greenhouse gas.

common agricultural policy (Hart & Baldock, 2011; Pe’er example being between conventionally managed fields or
et al., 2014), and some ‘sustainability’ certification schemes farms and an alternative system, such as organic manage-
[e.g. ‘WildCare’ (ABSustain, 2014)], use the proportion ment. A potential explanation for this is that a prescribed
of uncropped land and presence of vegetation features set of management practices reduces a farmers potential
as proxies for other aspects of biodiversity conservation to mix-and-match a range of practices to the particular
performance. However, the high among-study variability social, economic and environmental context, potentially
observed for many of the relationships in this analysis making trade-offs more severe. For example, several studies
suggests that using ‘one-size-fits-all’ proxies in a variety in our meta-analysis found that a conventional arable
of contexts may be of questionable value: the higher the regime had higher crop yields but lowered earthworm
variability among studies the more likely that an average activity compared to an organic system (Mader et al., 2002;
proxy would not work in any particular context. Therefore, Pulleman et al., 2003; Gabriel et al., 2013), whereas use
it may be more sensible to identify subsets of circumstances of manure or reduced tillage in a conventional system
(e.g. management attributes or environmental contexts), and can increase both simultaneously (Li et al., 2007b; Torabi
create ‘typologies’ where similar relationships among aspects et al., 2008; Endale et al., 2010). This suggests that there
of sustainability are observed, and tailor monitoring and is, in some cases, considerable scope for conventional
assessment methodologies to these individually. farming to benefit from some agro-ecological management
practices.
(d) Farm management versus metric management
(e) Zero-tillage
Interestingly trade-offs were stronger in those studies
that contrasted whole-farming approaches rather than Our analysis of context dependence in the effect of
single-management practices (Fig. 6), the most common zero-tillage may serve as a template for how existing

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


16 R. N. German and others

Number of results
Cool temperate, dry Warm temperate, dry Tropical, dry
10 8 6 4 2 0 2 4 6 8 10 10 8 6 4 2 0 2 4 6 8 10 6 4 2 0 2 4 6

Water flow Positive Water flow Water flow


No effect
Profit Negative Profit Profit
Costs Costs Costs
GHG emissions GHG emissions GHG emissions
Soil carbon Soil carbon Soil carbon
Soil nutrients Soil nutrients Soil nutrients
Yield Yield Yield
Water-use efficiency Water-use efficiency Water-use efficiency
Nutrient-use efficiency Nutrient-use efficiency Nutrient-use efficiency
Soil physical structure Soil physical structure Soil physical structure
Weeds Weeds Weeds
Animal pests Animal pests Animal pests
Soil biodiversity Soil biodiversity Soil biodiversity

Response Cool temperate, moist Warm temperate, moist Tropical, wet


variable 5 4 3 2 1 0 1 2 3 4 5 12 8 4 0 4 8 12 4 3 2 1 0 1 2 3 4

Water flow Water flow Water flow


Profit Profit Profit
Costs Costs Costs
GHG emissions GHG emissions GHG emissions
Soil carbon Soil carbon Soil carbon
Soil nutrients Soil nutrients Soil nutrients
Yield Yield Yield
Water-use efficiency Water-use efficiency Water-use efficiency
Nutrient-use efficiency Nutrient-use efficiency Nutrient-use efficiency
Soil physical structure Soil physical structure Soil physical structure
Weeds Weeds Weeds
Animal pests Animal pests Animal pests
Soil biodiversity Soil biodiversity Soil biodiversity

Fig. 7. Effect of zero-tillage versus conventional tillage treatments on multiple response variables, split by climate type. Bar charts
showing the number of studies finding a desirable (green), undesirable (red) or neutral effect (dark grey) of zero-tillage treatments on
several aspects of sustainability, relative to conventional tillage treatments. Each panel shows results from studies in a single climate
zone, according to IPCC categorisation (Bickel et al., 2006). Where no bar is present, that variable was not measured in any studies
from that climate zone. Note that y-axis scaling differs between panels. GHG, greenhouse gas.

data could be collated to examine the effects of (2) Between-study variability in the relationships
specific management interventions on multiple aspects of between aspects of sustainability
sustainability. We found that zero-tillage increases soil One of the key objectives of this paper was to
carbon content and water flow relative to other cultivation assess the consistency of patterns of co-variation among
treatments in most parts of the world (Fig. 7), but not aspects of sustainability across different production systems,
in studies from wet tropical areas. Moreover, total yields management gradients, and environmental contexts. We
under zero-tillage were lower than with conventional tillage need to produce a way to summarise broad global trends
in 12 of 17 studies shorter than 5 years in duration, but between aspects of sustainability, whilst retaining enough
higher in 20 out of 34 studies lasting longer than 5 years sensitivity to explore the impact of local conditions in order
(see online Fig. S2), so the time necessary to achieve a to seek ways to mitigate trade-offs. Too much emphasis on
positive yield benefit also varies among studies. Proponents the former leads to a rigid definition of sustainability that
of zero-tillage agriculture should emphasise these caveats, may have negative consequences in some contexts, whereas
and ensure that advice on its potential costs and benefits is too much emphasis on the latter leads to the conclusion that
taken from studies with a close climatic match to the site sustainability ‘depends’ uniquely on context and thus leads
in question. Other authors have stressed the importance to confusion and lack of generality.
of other management factors used in combination with There is considerable variability around the average
zero-tillage, such as residue management, herbicide use pairwise relationships for many sustainability variables
and cover crop choice, in determining the benefits of (Fig. 5). However, some relationships, such as that between
zero-tillage (e.g. Lal, 2009; Franzluebbers, 2010). These soil carbon and soil potassium content are much tighter,
factors did not explain a significant amount of between-study and perhaps reveal patterns which are not too context
variation in our meta-analysis, but this may be due to dependent. Co-variation between aspects of sustainability
the difficulty of consistently recording these from the may arise either through independent responses to shared
articles considered. The level of detail and comparability drivers, direct interactions between aspects of sustainability,
of the contextual information recorded varied widely among or a mixture of these (Bennett, Peterson & Gordon, 2009;
studies. Pilgrim et al., 2010). For example, soil compaction and water

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 17

infiltration rate (part of the ‘water flow’ level 2 response) multifunctional landscapes. This is the notion of ‘sustainable
are mechanistically linked, as more porous soils allow intensification’ as espoused by Garnett, Appleby & Balmford
greater water flow, but this may occur either through direct (2013). The advantage of planning for multiple objectives
effects of management on soil texture (including rotations over using a single variable as a proxy has been demonstrated
that vary crop root architecture), or indirectly through previously by other authors (Nelson et al., 2008; Thomas
effects on earthworms which act to increase soil porosity et al., 2013) when comparing strategies for reconciling species
(Edwards et al., 1990). Clearly, developing system-specific conservation and carbon storage, and is likely to apply to
mechanistic understanding of the processes driving observed other pairs of aspects of sustainability when correlations are
correlations must be one long-term goal, and as systems not strong. To be clear, we do not claim that choice potential
models proliferate to cover a wider array of physical, is somehow ‘superior’ to correlation; rather that when used
physiological, ecological and social processes, these may together they provide better information for decision-making
eventually be joined up by tools such as ARIES (Villa than either does alone.
et al., 2009) or InVEST (The Natural Capital Project,
2011) to produce realistic predictions of the outcome of (4) Robustness of the meta-analysis of the literature
management for many different aspects of sustainability, for
whole landscapes and regions. However, until this is a reality, There are several issues that should be borne in mind when
using consistently observed relationships among different interpreting the meta-analysis of the literature presented
aspects of sustainability is a useful guide to likely outcomes for above.
practitioners. First, the high uncertainties around estimates of correlation
and choice potential underline that the power of this
analysis is limited by a relatively small sample of
(3) Choice potential and correlation
studies that have assessed multiple services across different
Correlations or mean-trend lines provide information on situations. Furthermore, key word searching, whilst exactly
the likely outcome for other aspects of sustainability when reproducible, does not reveal all of the literature. We found
choosing management optimally for a single aspect of that many articles were not captured by our general search
sustainability, ‘blind’ to the others, or using one indicator as due to the lack of relevant key words, but nonetheless
a proxy for others. Measures of the shape of the efficiency satisfied all of our criteria. The low number or complete
frontier, such as the choice potential, better characterise absence of studies in our sample comparing some aspects of
constraints to ‘jointness in production’ (sensu Wossink & sustainability (white cells, Fig. 4; Tables 1 and 2) is however
Swinton, 2007). The results of the literature review show that likely to reflect a real dearth of evidence, which in itself
pairwise choice potential scores were often more favourable is a useful finding. The majority of relevant experimental
than the corresponding correlation coefficient, suggesting studies in our sample did not aim to measure sustainability
that good management compromises are possible even as broadly as possible, but instead focused on restricted parts
in the presence of strong negative correlations (Fig. 4). of agricultural systems in order to answer specific scientific
A particularly striking example of this is the relationship questions. On the other hand, studies recording many aspects
between crop yield and water quality; the Spearman’s rank of sustainability were more often unreplicated case studies,
correlation is significantly negative (−0.74) but the choice and therefore unsuitable for inclusion in our analysis. There
potential is nonetheless quite high (0.57). This indicates that, was a distinct paucity of studies which both measured
at least for some studies, high yield can be associated with a wide range of aspects of sustainability, and compared
good water quality. responses among alternative, replicated management
Although less certain, it is significant that yield and treatments.
biodiversity can jointly benefit from the right choice Secondly, two simplifications used in the literature analysis
of management, despite Spearman’s rank correlations are likely to have biased the Spearman’s rank correlations
across treatments being negative. This has implications and choice potential scores calculated for individual studies
for the land-sparing/sharing debate (Phalan et al., 2011), towards more extreme values, and require the results to
where density–yield curves for species have typically been be interpreted with some circumspection. Firstly, the data
quantified by lines fitted through the centre of a cloud of gathered from articles were treatment means, not individual
points. In light of our results, however, when repeatable observations; treatment means tend to average uncorrelated
management practices cause predictable variation around a variation in two variables. In the extreme case with only
mean trend line, an efficiency frontier may provide a truer two treatment groups, correlations calculated from raw
representation of the constraints to yield and population observations of close to zero will become either −1 or 1. The
density. Thus, there is no universal answer to the debate: second bias towards extremes comes from using the ranks of
both sharing and sparing may be appropriate depending on treatment means, rather than quantitative data. This tends
the context (Hodgson et al., 2010; Shackelford et al., 2015). to increase absolute values of correlation, especially with
These results suggest the potential for ‘smarter’ strategies low numbers of treatment groups, by restricting the possible
for sustainable land management: by investing in strategic number of positions at which points can lie. This effect can
planning for multiple ecosystem services at field, farm and be observed in Fig. 5, where even when most values lie near
landscape scales it is possible to have both high-yielding and one end of the possible range, values at the other extreme

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


18 R. N. German and others

are still present. Obtaining observation-level data from many V. CONCLUSIONS


studies would solve these issues, but would have required far
more time than we had available. (1) It is possible to reconcile many aspects of farm
Third, our aggregation of variables at several levels before sustainability through careful choice of management,
analysis could lead to difficulties in interpretation, in addition and land managers can, through these actions, have a
to increased among-study variability in the relationships considerable ‘bottom-up’ impact on the sustainability of
among aspects of sustainability. At the lowest level, two our food system.
studies which we categorised as measuring the same level (2) Yield and profit do trade off against important aspects
1 variable (see Fig. 1) often did not measure precisely the of sustainability, such as quality of water leaving the farm.
same biophysical indicators, in exactly the same way (see Conflicting aims must be balanced by land managers,
Table 1). We then further aggregated variables into level 2 ultimately based on the relative weight that they, and those
categories in order to derive a manageable set of comparisons. downstream in the supply chain, put on different aspects of
Among-study differences are therefore likely to occur at least sustainability.
in part due to differing behaviour of the exact indicators (3) The potential for trade-offs and win–win situations
measured. By using rankings, we attempted to minimise among aspects of sustainability varies from place to place.
the impact of this and make the quantities measured in For a buyer at a food retailer or wholesaler, and by extension
different studies more comparable. Given that the focus of for a consumer, this context dependency of management
our meta-analysis was broad, we believe that it was worth may allow them to exert perhaps a larger influence on how
trading off some precision, to gain a greater frequency of sustainably food is grown at a global scale than can the actions
comparisons and clarity of analysis. of individual land managers. By insisting on adherence to
Fourth, publication bias is an issue for all literature rigorous standards tailored to a particular product and
analyses. Bias is perhaps more likely to influence the location, and flexibly sourcing a given commodity from a
publication of results showing the direction of effect for location where the system is the least sensitive to adverse
particular response variables, such as yield, than those impacts (e.g. buying cotton from areas with abundant
showing particular relationships among response variables. rainfall) or where sympathetic management does not reduce
Our zero-tillage case study may thus be more susceptible to yield and profit, those higher up the food chain have the
this kind of bias than the rest of our analyses, as it focused on power to create the market forces that will drive increased
the directionality of outcomes. sustainability from the top down.
Finally, perhaps the most important source of uncertainty (4) Our research highlights that there is a great need
affecting the interpretation of our analysis is the issue of for context-specific evidence of the effects of different
incomplete, or unrepresentative coverage of agricultural management interventions on a whole suite of aspects
systems. The reasons for this are twofold. First, the available of sustainability, both for identifying management which
published literature may over-represent some kinds of will avoid trade-offs (bottom-up), and to develop useful
locations, crops, or types of management intervention while assessment methods (top-down). Given the patchy coverage
under-representing others, or omitting them entirely. For available from the academic research that we sourced, we
example, 11 of the studies included in our analysis concerning suggest that if food producers were routinely to collect data
wheat crops were undertaken in China, compared to only five concerning the effects of different production practices on
in India, despite India producing about three-quarters of the multiple axes of sustainability, this would greatly increase
weight of wheat that China did in 2012 (FAOSTAT, 2015). the evidence base available. However, to make use of this
Second, our literature-search strategy may have omitted large volume of varied evidence, it must be condensed into
certain classes of evidence. For example, a substantial body a form which can provide a clear and balanced basis for
of management and conservation experience is not reported management decisions (Dicks, Walsh & Sutherland, 2014).
in the academic literature at all. We feel that attempting In ecology and conservation, initiatives such as Conser-
to estimate or control for the effects of these two sources vation Evidence (http://www.conservationevidence.com/)
of bias would have introduced an extra level of complexity and the Collaboration for Environmental Evidence
to our analyses for an uncertain effect on quality, so we (http://www.environmentalevidence.org/) have used hier-
did not attempt to do so. In future analyses of this kind, archical frameworks to summarise and organise evidence of
selecting a sample of studies stratified by agricultural system the effectiveness of conservation management interventions,
and geographic variables may help ensure representativeness and the sustainable food production knowledge exchange
of overall estimates. However, this would require a larger program run by the Natural Environment Research Coun-
set of suitable studies than was available as this study was cil (NERC) in the UK (http://nercsustainablefood.com/)
conducted. For these reasons, we urge the reader to bear has adopted this approach to summarise how management
in mind that the average relationships among aspects of interventions impact on individual aspects of agricultural
sustainability emerging from our meta-analysis apply strictly sustainability. We argue that such a hierarchical approach
to the set of studies we considered, and should not be could be extended easily to encompass management impacts
interpreted as representing the entirety of world agriculture, on multiple aspects of agricultural sustainability simultane-
or a subset thereof. ously, as exemplified here with our case study of zero-tillage

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 19

agriculture, and to make use of the potential wealth of Bennett, E. M., Peterson, G. D. & Gordon, L. J. (2009). Understanding
relationships among multiple ecosystem services. Ecology Letters 12, 1394–1404.
evidence from food producers. *Berg, H., Berg, C. & Thanh Tam, N. (2012). Integrated rice-fish farming:
safeguarding biodiversity and ecosystem services for sustainable food production in
the Mekong Delta. Journal of Sustainable Agriculture 36, 859–872.
Bickel, K., Richards, G., Köhl, M. & Rodrigues, R. L. V. (2006). Consistent
representation of lands. In 2006 IPCC Guidelines for National Greenhouse Gas Inventories,
VI. ACKNOWLEDGEMENTS Chapter 3 (eds H. S. Eggleston, L. Buendia, K. Miwa, T. Ngara, K. Tanabe).
IGES, Hayama.
*Bilalis, D. J. & Karamanos, A. J. (2010). Organic maize growth and mycorrhizal
We would like to thank Prof. Bill Sutherland, Dr Lynn root colonization response to tillage and organic fertilization. Journal of Sustainable
Dicks and their postgraduate students for stimulating Agriculture 34, 836–849.
discussions and suggestions. Dr Richard Barnes and Dr Chris *Bisseleua, D. H. B., Missoup, A. D. & Vidal, S. (2009). Biodiversity
conservation, ecosystem functioning, and economic incentives under cocoa
Brown provided important sense-checks from an industry agroforestry intensification. Conservation Biology 23, 1176–1184.
perspective. Two anonymous referees supplied valuable *Bjorklund, J., Limburg, K. E. & Rydberg, T. (1999). Impact of production
comments. The UK Natural Environment Research Council intensity on the ability of the agricultural landscape to generate ecosystem services:
an example from Sweden. Ecological Economics 29, 269–291.
(NERC) provided funding for this work. *de Boer, I. J. M. & Cornelissen, A. M. G. (2002). A method using sustainability
indicators to compare conventional and animal-friendly egg production systems.
Poultry Science 81, 173–181.
*Bonny, B. P., Prasad, R. M. & Paulose, S. (2010). Agro-ecosystem performance
VII. REFERENCES index (API)–a quantitative approach to evaluate the sustainability of rice production
systems. Journal of Sustainable Agriculture 34, 758–777.
Borenstein, M., Hedges, L. V., Higgins, J. P. T. & Rothstein, H. R. (2009).
References marked with asterisk have been cited within the supporting information Introduction to Meta-Analysis. Wiley, Chichester.
*Abe, S. S., Hashimoto, S., Umezane, T., Yamaguchi, T., Yamamoto, S., *Borowy, A. (2012). Growth and yield of stake tomato under no-tillage cultivation
Yamada, S., Endo, T. & Nakata, N. (2010). Agronomic and environmental using hairy vetch as a living mulch. Acta Scientiarum Polonorum-Hortorum Cultus 11,
performance of rapeseed oilcake in the lowland rice farming of Japan. Communications 229–252.
in Soil Science and Plant Analysis 41, 1944–1958. Bruinsma, J. (2009). The resource outlook to 2050: by how much do land, water and
Abson, D., Fraser, E. & Benton, T. (2013). Landscape diversity and the resilience crop yields need to increase by 2050?. In Proceedings of the Expert Meeting on How to Feed
of agricultural returns: a portfolio analysis of land-use patterns and economic returns
the World in 2050, Food and Agriculture Organisation of the United Nations, Rome,
from lowland agriculture. Agriculture and Food Security 2, 1–15.
24–26 June 2009.
ABSustain (2014). Wildcare. Available at Http://www.wildcare.co.uk Accessed
Brundtland, G. (1987). Our Common Future: Report of the 1987 World Commission on
5.5.2014.
Environment and Development. Oxford University Press, Oxford.
*Adamtey, N., Cofie, O., Ofosu-Budu, K. G., Ofosu-Anim, J., Laryea, K. B. &
*Brunett-Pérez, L., González-Esquivel, C. E. & García-Hernández, L.
Forster, D. (2010). Effect of N-enriched co-compost on transpiration efficiency and
(2005). Evaluación de la sustentabilidad de dos agroecosistemas campesinos de
water-use efficiency of maize (Zea mays L.) under controlled irrigation. Agricultural
producción de maíz y leche, utilizando indicadores. Livestock Research for Rural
Water Management 97, 995–1005.
Development 17, 78.
*Agbede, T. M. & Ojeniyi, S. O. (2009). Tillage and poultry manure effects on
*Bullock, J. M. & Pywell, R. F. (2005). Rhinanthus: a tool for restoring diverse
soil fertility and sorghum yield in southwestern Nigeria. Soil & Tillage Research 104,
grassland? Folia Geobotanica 40, 273–288.
74–81.
*Buntin, G. D., All, J. N., McCracken, D. V. & Hargrove, W. L. (1994).
*Allen, V. G., Baker, M. T., Segarra, E. & Brown, C. P. (2007). Integrated
Cover crop and nitrogen fertility effects on southern corn-rootworm (coleoptera,
irrigated crop-livestock systems in dry climates. Agronomy Journal 99, 346–360.
chrysomelidae) damage in corn. Journal of Economic Entomology 87, 1683–1688.
*Allen, V. G., Brown, C. P., Kellison, R., Segarra, E., Wheeler, T., Dotray, P.
*Cambardella, C. A., Moorman, T. B., Andrews, S. S. & Karlen, D. L. (2004).
A., Conkwright, J. C., Green, C. J. & Acosta-Martinez, V. (2005). Integrating
cotton and beef production to reduce water withdrawal from the Ogallala aquifer Watershed-scale assessment of soil quality in the loess hills of southwest Iowa. Soil &
in the southern high plains. Agronomy Journal 97, 556–567. Tillage Research 78, 237–247.
*Araya, T., Cornelis, W. M., Nyssen, J., Govaerts, B., Getnet, F., Bauer, *Campanelli, G. & Canali, S. (2012). Crop production and environmental effects
H., Amare, K., Raes, D., Haile, M. & Deckers, J. (2012). Medium-term effects in conventional and organic vegetable farming systems: the case of a long-term
of conservation agriculture based cropping systems for sustainable soil and water experiment in mediterranean conditions (central Italy). Journal of Sustainable Agriculture
management and crop productivity in the Ethiopian highlands. Field Crops Research 36, 599–619.
132, 53–62. *Candido, V., D’Addabbo, T., Basile, M., Castronuovo, D. & Miccolis, V.
*Arshad, M. A., Gill, K. S. & Izaurralde, R. C. (1998). Wheat production, weed (2008). Greenhouse soil solarization: effect on weeds, nematodes and yield of tomato
population and soil properties subsequent to 20 years of sod as affected by crop and melon. Agronomy for Sustainable Development 28, 221–230.
rotation and tillage. Journal of Sustainable Agriculture 12, 131–154. *Castellini, C., Boggia, A., Cortina, C., Dal Bosco, A., Paolotti, L.,
*Attard, E., Recous, S., Chabbi, A., De Berranger, C., Guillaumaud, Novelli, E. & Mugnai, C. (2012). A multicriteria approach for measuring the
N., Labreuche, J., Philippot, L., Schmid, B. & Le Roux, X. (2011). Soil sustainability of different poultry production systems. Journal of Cleaner Production 37,
environmental conditions rather than denitrifier abundance and diversity drive 192–201.
potential denitrification after changes in land uses. Global Change Biology 17, *Castoldi, N. & Bechini, L. (2010). Integrated sustainability assessment of cropping
1975–1989. systems with agro-ecological and economic indicators in northern Italy. European
*Ayisi, K. K., Mposi, M. S. & van den Berg, J. (2001). Grain yield response and Journal of Agronomy 32, 59–72.
Chilo partellus infestation in diverse sorghum-cowpea intercrop arrangements. South *Cavigelli, M. A., Teasdale, J. R. & Conklin, A. E. (2008). Long-term agronomic
African Journal of Plant and Soil 18, 39–42. performance of organic and conventional field crops in the mid-Atlantic region.
*Bai, Y., He, J., Li, H., Wang, Q., Chen, H., Kuhn, N. J., Hikel, H., Chen, F. & Agronomy Journal 100, 785–794.
Gong, Y. (2009). Soil structure and crop performance after 10 years of controlled *Cho, Y. S., Lee, B. Z., Choe, Z. R. & Ockerby, S. E. (2001). An evaluation of a
traffic and traditional tillage cropping in the dryland Loess Plateau in China. Soil no-tillage, unfertilised, direct-sown, wheat-rice cropping system in Korea. Australian
Science 174, 113–119. Journal of Experimental Agriculture 41, 53–60.
*Balezentiene, L. & Kusta, A. (2012). Reducing greenhouse gas emissions in *Clough, Y., Barkmann, J., Juhrbandt, J., Kessler, M., Wanger, T. C.,
grassland ecosystems of the central Lithuania: multi-criteria evaluation on a basis of Anshary, A., Buchori, D., Cicuzza, D., Darras, K., Putra, D. D., Erasmi, S.,
the ARAS method. Scientific World Journal 2012, 1–12. Pitopang, R., Schmidt, C., Schulze, C. H., Seidel, D., Steffan-Dewenter,
*Basamba, T. A., Barrios, E., Amezquita, E., Rao, I. M. & Singh, B. R. (2006). I., Stenchly, K., Vidal, S., Weist, M., Wielgoss, A. C., Tscharntke, T.
Tillage effects on maize yield in a Colombian savanna oxisol: soil organic matter (2011). Combining high biodiversity with high yields in tropical agroforests.
and P fractions. Soil & Tillage Research 91, 131–142. Proceedings of the National Academy of Sciences of the United States of America 108, 8311–8316.
*Bayer, C., Mielniczuk, J., Amado, T. J. C., Martin-Neto, L. & Fernandes, S. *Cociu, A. I. (2011). Soil properties, winter wheat yield, its components and economic
V. (2000). Organic matter storage in a sandy clay loam Acrisol affected by tillage efficiency when different tillage systems are applied. Romanian Agricultural Research 28,
and cropping systems in southern Brazil. Soil & Tillage Research 54, 101–109. 121–130.

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


20 R. N. German and others

*Costa, F. S., Albuquerque, J. A., Bayer, C., Fontoura, S. M. V. & Wobeto, *Ghosh, P. K., Das, A., Saha, R., Kharkrang, E., Tripathi, A. K., Munda,
C. (2003). Physical properties of a south Brazilian Oxisol as affected by no-tillage G. C. & Ngachan, S. V. (2010). Conservation agriculture towards achieving food
and conventional tillage systems. Revista Brasileira de Ciencia do Solo 27, 527–535. security in North East India. Current Science 99, 915–921.
*Dalal, R. C., Wang, W., Allen, D. E., Reeves, S. & Menzies, N. W. (2011). *Glover, J. D., Culman, S. W., DuPont, S. T., Broussard, W., Young, L.,
Soil nitrogen and nitrogen-use efficiency under long-term no-till practice. Soil Science Mangan, M. E., Mai, J. G., Crews, T. E., DeHaan, L. R., Buckley, D. H.,
Society of America Journal 75, 2251–2261. Ferris, H., Turner, R. E., Reynolds, H. L. & Wyse, D. L. (2010). Harvested
Dale, V. H. & Polasky, S. (2007). Measures of the effects of agricultural practices on perennial grasslands provide ecological benchmarks for agricultural sustainability.
ecosystem services. Ecological Economics 64, 286–296. Agriculture, Ecosystems & Environment 137, 3–12.
*Das, D. K., Chaturvedi, O. P., Jha, R. K. & Kumar, R. (2011). Yield, soil health Gomez-Limon, J. A. & Sanchez-Fernandez, G. (2010). Empirical evaluation
and economics of aonla (Emblica officinalis Gaertn.)-based agri-horticultural systems of agricultural sustainability using composite indicators. Ecological Economics 69,
in eastern India. Current Science 101, 786–790. 1062–1075.
Dicks, L., Bardgett, R., Bell, J., Benton, T., Booth, A., Bouwman, J., *Govaerts, B., Fuentes, M., Mezzalama, M., Nicol, J. M., Deckers, J.,
Brown, C., Bruce, A., Burgess, P., Butler, S., Crute, I., Dixon, F., Etchevers, J. D., Figueroa-Sandoval, B. & Sayre, K. D. (2007). Infiltration,
Drummond, C., Freckleton, R., Gill, M., Graham, A., Hails, R., Hallett, soil moisture, root rot and nematode populations after 12 years of different tillage,
J., Hart, B., Hillier, J., Holland, J., Huxley, J., Ingram, J., King, V., residue and crop rotation managements. Soil & Tillage Research 94, 209–219.
MacMillan, T., McGonigle, D., McQuaid, C., Nevard, T., Norman, S., *Gowda, M. J. C. & Jayaramaiah, K. M. (1998). Comparative evaluation of rice
Norris, K., Pazderka, C., Poonaji, I., Quinn, C., Ramsden, S., Sinclair, D., production systems for their sustainability. Agriculture, Ecosystems & Environment 69,
Siriwardena, G., Vickery, J., Whitmore, A., Wolmer, W., Sutherland, W. 1–9.
(2013). What do we need to know to enhance the environmental sustainability of *Green, V. S., Cavigelli, M. A., Dao, T. H. & Flanagan, D. C. (2005). Soil
agricultural production? A prioritisation of knowledge needs for the UK food system. physical properties and aggregate-associated C, N, and P distributions in organic
Sustainability 5, 3095–3115. and conventional cropping systems. Soil Science 170, 822–831.
Dicks, L. V., Walsh, J. C. & Sutherland, W. J. (2014). Organising evidence for Grunwald, S., Thompson, J. A. & Boettinger, J. L. (2011). Digital soil mapping
environmental management decisions: a ‘4S’ hierarchy. Trends in Ecology & Evolution and modeling at continental scales: finding solutions for global issues. Soil Science
29, 607–613. Society of America Journal 75, 1201.
*Dogan, E., Copur, O., Kahraman, A., Kirnak, H. & Guldur, M. E. (2011). *Gwenzi, W., Gotosa, J., Chakanetsa, S. & Mutema, Z. (2009). Effects of
Supplemental irrigation effect on canola yield components under semiarid climatic tillage systems on soil organic carbon dynamics, structural stability and crop yields
conditions. Agricultural Water Management 98, 1403–1408. in irrigated wheat (Triticum aestivum L.)-cotton (Gossypium hirsutum L.) rotation in
Edwards, W., Shipitalo, M., Owens, L. & Norton, L. (1990). Effect of Lumbricus semi-arid Zimbabwe. Nutrient Cycling in Agroecosystems 83, 211–221.
terrestris L. burrows on hydrology of continuous no-till corn fields. Geoderma 46, *Haas, G., Wetterich, F. & Kopke, U. (2001). Comparing intensive, extensified
73–84. and organic grassland farming in southern Germany by process life cycle assessment.
*Eltun, R., Korsaeth, A. & Nordheim, O. (2002). A comparison of environmental, Agriculture, Ecosystems & Environment 83, 43–53.
soil fertility, yield, and economical effects in six cropping systems based on an 8-year *Haenke, S., Scheid, B., Schaefer, M., Tscharntke, T. & Thies, C. (2009).
Increasing syrphid fly diversity and density in sown flower strips within simple vs.
experiment in Norway. Agriculture, Ecosystems & Environment 90, 155–168.
complex landscapes. Journal of Applied Ecology 46, 1106–1114.
Endale, D. M., Schomberg, H. H., Jenkins, M. B., Franklin, D. H. & Fisher,
*Hai An Phan, H., Huon, S., des Tureaux, T. H., Orange, D., Jouquet, P.,
D. S. (2010). Management implications of conservation tillage and poultry litter use
Valentin, C., De Rouw, A. & Toan Tran, D. (2012). Impact of fodder cover on
for Southern Piedmont USA cropping systems. Nutrient Cycling in Agroecosystems 88,
runoff and soil erosion at plot scale in a cultivated catchment of North Vietnam.
299–313.
Geoderma 177, 8–17.
FAOSTAT (2015). Food and agricultural commodities production. Food and
Hart, K. & Baldock, D. (2011). Greening the CAP: Delivering Environmental Outcomes
Agriculture Organisation of the United Nations. Available at http://faostat.fao.org/
through Pillar One. Institute for European Environmental Policy, London.
site/339/default.aspx Accessed 16.8.2015.
*He, J., Kuhn, N. J., Zhang, X. M., Zhang, X. R. & Li, H. W. (2009). Effects
*Feiza, V., Feiziene, D., Auskalnis, A. & Kadziene, G. (2010). Sustainable tillage:
of 10 years of conservation tillage on soil properties and productivity in the
results from long-term field experiments on Cambisol. Zemdirbyste-Agriculture 97,
farming-pastoral ecotone of Inner Mongolia, China. Soil Use and Management 25,
3–14.
201–209.
*Fernandes, L. A. d. O. & Woodhouse, P. J. (2008). Family farm sustainability in
*He, J., Li, H., Rasaily, R. G., Wang, Q., Cai, G., Su, Y., Qiao, X. & Liu,
southern Brazil: an application of agri-environmental indicators. Ecological Economics
L. (2011). Soil properties and crop yields after 11 years of no tillage farming in
66, 243–257. wheat-maize cropping system in North China Plain. Soil & Tillage Research 113,
*Ferreiro-Dominguez, N., Rigueiro-Rodriguez, A. & Mosquera-Losada, M. 48–54.
R. (2011). Response to sewage sludge fertilisation in a Quercus rubra L. silvopastoral *He, J., McHugh, A. D., Li, H. W., Wang, Q. J., Li, W. Y., Rasaily, R. G. & Li,
system: soil, plant biodiversity and tree and pasture production. Agriculture, Ecosystems H. (2012). Permanent raised beds improved soil structure and yield of spring wheat
& Environment 141, 49–57. in arid north-western China. Soil Use and Management 28, 536–543.
Foley, J. A., Ramankutty, N., Brauman, K. A., Cassidy, E. S., Gerber, J. S., *Hernanz, J. L., Lopez, R., Navarrete, L. & Sanchez-Giron, V. (2002).
Johnston, M., Mueller, N. D., O’Connell, C., Ray, D. K., West, P. C., Long-term effects of tillage systems and rotations on soil structural stability and
Balzer, C., Bennett, E. M., Carpenter, S. R., Hill, J., Monfreda, C., organic carbon stratification in semiarid central Spain. Soil & Tillage Research 66,
Polasky, S., Rockstrom, J., Sheehan, J., Siebert, S., Tilman, D., Zaks, 129–141.
D. P. M. (2011). Solutions for a cultivated planet. Nature 478, 337–342. Hodgson, J. A., Kunin, W. E., Thomas, C. D., Benton, T. G. & Gabriel, D.
Franzluebbers, A. J. (2010). Achieving soil organic carbon sequestration with (2010). Comparing organic farming and land sparing: optimizing yield and butterfly
conservation agricultural systems in the Southeastern United States. Soil Science populations at a landscape scale. Ecology Letters 13, 1358–1367.
Society of America Journal 74, 347–357. Holland, R. A., Eigenbrod, F., Armsworth, P. R., Anderson, B. J., Thomas,
Gabriel, D., Sait, S. M., Hodgson, J. A., Schmutz, U., Kunin, W. E. & Benton, C. D., Heinemeyer, A., Gillings, S., Roy, D. B. & Gaston, K. J. (2011).
T. G. (2010). Scale matters: the impact of organic farming on biodiversity at different Spatial covariation between freshwater and terrestrial ecosystem services. Ecological
spatial scales. Ecology Letters 13, 858–869. Applications 21, 2034–2048.
Gabriel, D., Sait, S. M., Kunin, W. E. & Benton, T. G. (2013). Food production Hothorn, T., Hornik, K. & Zeileis, A. (2006). Unbiased recursive partitioning:
vs. biodiversity: comparing organic and conventional agriculture. Journal of Applied a conditional inference framework. Journal of Computational and Graphical Statistics 15,
Ecology 50, 355–364. 651–674.
*Galvez, L., Douds, D. D., Drinkwater, L. E. & Wagoner, P. (2001). Effect of *Huang, G., Chai, Q., Feng, F. & Yu, A. (2012). Effects of different tillage systems
tillage and farming system upon VAM fungus populations and mycorrhizas and on soil properties, root growth, grain yield, and water use efficiency of winter wheat
nutrient uptake of maize. Plant and Soil 228, 299–308. (triticum aestivum L.) in arid northwest China. Journal of Integrative Agriculture 11,
Garnett, T., Appleby, M. & Balmford, A. (2013). Sustainable intensification in 1286–1296.
agriculture: premises and policies. Science 341, 33–34. *Huang, C., Deng, L., Gao, X., Zhang, S., Luo, T. & Ren, Q. (2010). Effects of
*Geiger, F., Bengtsson, J., Berendse, F., Weisser, W. W., Emmerson, M., fungal residues return on soil enzymatic activities and fertility dynamics in a paddy
Morales, M. B., Ceryngier, P., Liira, J., Tscharntke, T., Winqvist, C., soil under a rice-wheat rotation in Chengdu Plain. Soil & Tillage Research 108, 16–23.
Eggers, S., Bommarco, R., Part, T., Bretagnolle, V., Plantegenest, *Huang, M. B., Shao, M. G., Zhang, L. & Li, Y. S. (2003). Water use efficiency
M., Clement, L. W., Dennis, C., Palmer, C., Onate, J. J., Guerrero, I., and sustainability of different long-term crop rotation systems in the Loess Plateau
Hawro, V., Aavik, T., Thies, C., Flohre, A., Haenke, S., Fischer, C., of China. Soil & Tillage Research 72, 95–104.
Goedhart, P. W., Inchausti, P. (2011). Persistent negative effects of pesticides on *Huang, G. B., Zhang, R. Z., Li, G. D., Li, L. L., Chan, K. Y., Heenan, D. P.,
biodiversity and biological control potential on European farmland (vol 11, pg 97, Chen, W., Unkovich, M. J., Robertson, M. J., Cullis, B. R. & Bellotti, W.
2010). Basic and Applied Ecology 12, 386–387. D. (2008). Productivity and sustainability of a spring wheat-field pea rotation in a

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 21

semi-arid environment under conventional and conservation tillage systems. Field Liu, C., Kroeze, C., Hoekstra, A. Y. & Gerbens-Leenes, W. (2012). Past and
Crops Research 107, 43–55. future trends in grey water footprints of anthropogenic nitrogen and phosphorus
*Hudewenz, A., Klein, A.-M., Scherber, C., Stanke, L., Tscharntke, T., inputs to major world rivers. Ecological Indicators 18, 42–49.
Vogel, A., Weigelt, A., Weisser, W. W. & Ebeling, A. (2012). Herbivore and *Liu, W. Z. & Zhang, X. C. (2007). Optimizing water and fertilizer input using an
pollinator responses to grassland management intensity along experimental changes elasticity index: a case study with maize in the loess plateau of china. Field Crops
in plant species richness. Biological Conservation 150, 42–52. Research 100, 302–310.
*Iijima, M., Izumi, Y., Yuliadi, E., Sunyoto, S., Afandi, A. & Utomo, M. (2003). *Liu, E. K., Zhao, B. Q., Mei, X. R., So, H. B., Li, J. & Li, X. Y. (2010). Effects
Erosion control on a steep sloped coffee field in Indonesia with alley cropping, of no-tillage management on soil biochemical characteristics in northern China.
intercropped vegetables, and no-tillage. Plant Production Science 6, 224–229. Journal of Agricultural Science 148, 217–223.
*Izumi, Y., Yuliadi, E., Sunyoto, S. & Iijima, M. (2004). Maize-soybean-cowpea *Lyon, D. J., Stroup, W. W. & Brown, R. E. (1998). Crop production and soil
sequential cropping as a sustainable crop production for acid-infertile clay soils in water storage in long-term winter wheat-fallow tillage experiments. Soil & Tillage
Indonesia. Plant Production Science 7, 356–362. Research 49, 19–27.
*Jat, M. L., Gathala, M. K., Ladha, J. K., Saharawat, Y. S., Jat, A. S., Kumar, Machovina, B. & Feeley, K. J. (2014). Meat consumption as a key impact on tropical
V., Sharma, S. K., Kumar, V. & Gupta, R. (2009). Evaluation of precision nature: a response to Laurance et al. Trends in Ecology & Evolution 29, 430–431.
land leveling and double zero-till systems in the rice-wheat rotation: water use, Mader, P., Fliessbach, A., Dubois, D., Gunst, L., Fried, P. & Niggli, U. (2002).
productivity, profitability and soil physical properties. Soil & Tillage Research 105, Soil fertility and biodiversity in organic farming. Science 296, 1694–1697.
112–121. Maes, J., Paracchini, M. L., Zulian, G., Dunbar, M. B. & Alkemade, R. (2012).
*Jaynes, D. B., Colvin, T. S., Karlen, D. L., Cambardella, C. A. & Meek, D. Synergies and trade-offs between ecosystem service supply, biodiversity, and habitat
W. (2001). Nitrate loss in subsurface drainage as affected by nitrogen fertilizer rate. conservation status in Europe. Biological Conservation 155, 1–12.
Journal of Environmental Quality 30, 1305–1314. *Malhi, S. S. & Lemke, R. (2007). Tillage, crop residue and N fertilizer effects on
*Jedlicka, J. A., Greenberg, R. & Letourneau, D. K. (2011). Avian conservation crop yield, nutrient uptake, soil quality and nitrous oxide gas emissions in a second
practices strengthen ecosystem services in California vineyards. PLoS One 6, e27437. 4-yr rotation cycle. Soil & Tillage Research 96, 269–283.
*Kang, G. S., Beri, V., Sidhu, B. S. & Rupela, O. P. (2005). A new index to assess *Malley, Z. J. U., Mzimbiri, M. K. & Mwakasendo, J. A. (2009). Integrating local
soil quality and sustainability of wheat-based cropping systems. Biology and Fertility of knowledge with science and technology in management of soil, water and nutrients:
Soils 41, 389–398. implications for management of natural capital for sustainable rural livelihoods.
*Karlen, D. L., Berry, E. C., Colvin, T. S. & Kanwar, R. S. (1991). 12-year International Journal of Sustainable Development and World Ecology 16, 151–163.
tillage and crop-rotation effects on yields and soil chemical-properties in northeast Manley, J., Van Kooten, G. C., Moeltner, K. & Johnson, D. W. (2005). Creating
iowa. Communications in Soil Science and Plant Analysis 22, 1985–2003. carbon offsets in agriculture through no-till cultivation: a meta-analysis of costs and
*Kauffman, J. B., Thorpe, A. S. & Brookshire, E. N. J. (2004). Livestock exclusion carbon benefits. Climatic Change 68, 41–65.
and belowground ecosystem responses in riparian meadows of Eastern Oregon. Maskell, L. C., Crowe, A., Dunbar, M. J., Emmett, B., Henrys, P., Keith, A.
Ecological Applications 14, 1671–1679. M., Norton, L. R., Scholefield, P., Clark, D. B., Simpson, I. C. & Smart,
Kearney, J. (2010). Food consumption trends and drivers. Philosophical Transactions of S. M. (2013). Exploring the ecological constraints to multiple ecosystem service
the Royal Society of London, Series B: Biological Sciences 365, 793–807. delivery and biodiversity. Journal of Applied Ecology 50, 561–571.
*Klein, A.-M., Brittain, C., Hendrix, S. D., Thorp, R., Williams, N. & Kremen, *Masto, R. E., Chhonkar, P. K., Singh, D. & Patra, A. K. (2008). Alternative
C. (2012). Wild pollination services to California almond rely on semi-natural habitat. soil quality indices for evaluating the effect of intensive cropping, fertilisation and
Journal of Applied Ecology 49, 723–732. manuring for 31 years in the semi-arid soils of India. Environmental Monitoring and
*Knudsen, M. T., Qiao, Y.-H., Luo, Y. & Halberg, N. (2010). Environmental Assessment 136, 419–435.
assessment of organic soybean (Glycine max.) imported from China to Denmark: a *Melero, S., Jesus Lopez-Bellido, R., Lopez-Bellido, L., Munoz-Romero, V.,
case study. Journal of Cleaner Production 18, 1431–1439. Moreno, F. & Manuel Murillo, J. (2011). Long-term effect of tillage, rotation
*Krauss, J., Gallenberger, I. & Steffan-Dewenter, I. (2011). Decreased and nitrogen fertiliser on soil quality in a Mediterranean Vertisol. Soil & Tillage
functional diversity and biological pest control in conventional compared to organic Research 114, 97–107.
crop fields. PLoS One 6, e19502. *Mendez, V. E., Shapiro, E. N. & Gilbert, G. S. (2009). Cooperative management
*Kumar, S., Kadono, A., Lal, R. & Dick, W. (2012). Long-term no-till impacts on and its effects on shade tree diversity, soil properties and ecosystem services of coffee
organic carbon and properties of two contrasting soils and corn yields in Ohio. Soil plantations in western El Salvador. Agroforestry Systems 76, 111–126.
Science Society of America Journal 76, 1798–1809. *Millar, G. D. & Badgery, W. B. (2009). Pasture cropping: a new approach to
*Lal, R. (1998). Soil quality changes under continuous cropping for seventeen seasons integrate crop and livestock farming systems. Animal Production Science 49, 777–787.
of an alfisol in western Nigeria. Land Degradation & Development 9, 259–274. *Mina, B. L., Saha, S., Kumar, N., Srivastva, A. K. & Gupta, H. S. (2008).
Lal, R. (2009). Soils and food sufficiency. A review. Agronomy for Sustainable Development Changes in soil nutrient content and enzymatic activity under conventional and
29, 113–133. zero-tillage practices in an Indian sandy clay loam soil. Nutrient Cycling in Agroecosystems
Lal, R., Reicosky, D. L. & Hanson, J. D. (2007). Evolution of the plow over 10,000 82, 273–281.
years and the rationale for no-till farming. Soil & Tillage Research 93, 1–12. *Minns, A., Finn, J., Hector, A., Caldeira, M., Joshi, J., Palmborg, C., Schmid,
*Laliberte, E. & Tylianakis, J. M. (2012). Cascading effects of long-term land-use B., Scherer-Lorenzen, M., Spehn, E., Troumbis, A. & Project, B (2001). The
changes on plant traits and ecosystem functioning. Ecology 93, 145–155. functioning of European grassland ecosystems: potential benefits of biodiversity to
*Laryea, K. B. & Unger, P. W. (1995). Grassland converted to agriculture. Outlook on Agriculture 30, 179–185.
cropland–soil-conditions and sorghum yield. Soil & Tillage Research 33, 29–45. *Moraru, P. I. & Rusu, T. (2012). Effect of tillage systems on soil moisture, soil
Laurance, W. F., Sayer, J. & Cassman, K. G. (2014). Agricultural expansion and temperature, soil respiration and production of wheat, maize and soybean crops.
its impacts on tropical nature. Trends in Ecology & Evolution 29, 107–116. Journal of Food Agriculture & Environment 10, 445–448.
*Leiva, F. R. & Morris, J. (2001). Mechanization and sustainability in arable farming *Morell, F. J., Lampurlanes, J., Alvaro-Fuentes, J. & Cantero-Martinez,
in England. Journal of Agricultural Engineering Research 79, 81–90. C. (2011). Yield and water use efficiency of barley in a semiarid Mediterranean
*Letourneau, D. K., Armbrecht, I., Salguero Rivera, B., Montoya Lerma, agroecosystem: long-term effects of tillage and N fertilization. Soil & Tillage Research
J., Jimenez Carmona, E., Constanza Daza, M., Escobar, S., Galindo, V., 117, 76–84.
Gutierrez, C., Duque Lopez, S., Lopez Mejia, J., Acosta Rangel, A. M., *Motsi, K. E., Chuma, E. & Mukamuri, B. B. (2004). Rainwater harvesting for
Herrera Rangel, J., Rivera, L., Arturo Saavedra, C., Marina Torres, sustainable agriculture in communal lands of Zimbabwe. Physics and Chemistry of the
A. & Reyes Trujillo, A. (2011). Does plant diversity benefit agroecosystems? A Earth 29, 1069–1073.
synthetic review. Ecological Applications 21, 9–21. *de Moura, E. G., Albuquerque, J. M. & Aguiar, A. d. C. F. (2008). Growth and
*Li, H., Gao, H., Wu, H., Li, W., Wang, X. & He, J. (2007a). Effects of 15 years productivity of corn as affected by mulching and tillage in alley cropping systems.
of conservation tillage on soil structure and productivity of wheat cultivation in Scientia Agricola 65, 204–208.
northern China. Australian Journal of Soil Research 45, 344–350. Mouron, P., Heijne, B., Naef, A., Strassemeyer, J., Hayer, F., Avilla, J.,
Li, X.-G., Wang, Z.-F., Ma, Q.-F. & Li, F.-M. (2007b). Crop cultivation and intensive Alaphilippe, A., Hoehn, H., Hernandez, J., Mack, G., Gaillard, G.,
grazing affect organic C pools and aggregate stability in and grassland soil. Soil & Sole, J., Sauphanor, B., Patocchi, A., Samietz, J., Bravin, E., Lavigne,
Tillage Research 95, 172–181. C., Bohanec, M., Golla, B., Scheer, C., Aubert, U., Bigler, F. (2012).
*Li, C., Yang, J., Zhang, C., Zhang, Z., Zheng, M., Ahmad, S. & Cao, C. (2010). Sustainability assessment of crop protection systems: sustainOS methodology and
Effects of short-term tillage and fertilization on grain yields and soil properties of its application for apple orchards. Agricultural Systems 113, 1–15.
rice production systems in central China. Journal of Food Agriculture & Environment 8, *Mueller-Lindenlauf, M., Deittert, C. & Koepke, U. (2010). Assessment of
577–584. environmental effects, animal welfare and milk quality among organic dairy farms.
*Liu, C. A., Jin, S. L., Zhou, L. M., Jia, Y., Li, F. M., Xiong, Y. C. & Li, X. Livestock Science 128, 140–148.
G. (2009). Effects of plastic film mulch and tillage on maize productivity and soil Nelson, E., Mendoza, G., Regetz, J., Polasky, S., Tallis, H., Cameron, D.
parameters. European Journal of Agronomy 31, 241–249. R., Chan, K. M. A., Daily, G. C., Goldstein, J., Kareiva, P. M., Lonsdorf,

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


22 R. N. German and others

E., Naidoo, R., Ricketts, T. H. & Shaw, M. R. (2009). Modeling multiple *Schulthess, F., Neuenschwander, P. & Gounou, S. (1997). Multi-trophic
ecosystem services, biodiversity conservation, commodity production, and tradeoffs interactions in cassava, Manihot esculenta, cropping systems in the subhumid
at landscape scales. Frontiers in Ecology and the Environment 7, 4–11. tropics of West Africa. Agriculture, Ecosystems & Environment 66, 211–222.
Nelson, E., Polasky, S., Lewis, D. J., Plantinga, A. J., Lonsdorf, E., White, Seybold, C. A., Mausbach, M. J., Karlen, D. L. & Rogers, H. H. (1997).
D., Bael, D. & Lawler, J. J. (2008). Efficiency of incentives to jointly increase Quantification of soil quality. In Soil Processes and the Carbon Cycle (eds R. Lal, J.
carbon sequestration and species conservation on a landscape. Proceedings of the M. Kimble, R. F. Follett and B. A. Stewart), pp. 387–404. CRC Press,
National Academy of Sciences of the United States of America 105, 9471–9476. Washington.
*Ofosu, E. A., van der Zaag, P., van de Giesen, N. C. & Odai, S. N. (2010). Shackelford, G. E., Steward, P. R., German, R. N., Sait, S. M. & Benton,
Productivity of irrigation technologies in the White Volta basin. Physics and Chemistry T. G. (2015). Conservation planning in agricultural landscapes: hotspots of conflict
of the Earth 35, 706–716. between agriculture and nature. Diversity and Distributions 21, 357–367.
*Ogban, P. I. & Babalola, O. (1999). Preliminary evaluation of tillage systems on *Shukla, M. K., Lal, R. & Ebinger, M. (2004). Soil quality indicators for the
poorly-drained inland valley bottom soils. Tropical Agriculture 76, 236–240. North Appalachian experimental watersheds in Coshocton Ohio. Soil Science 169,
*Pacini, C., Wossink, A., Giesen, G., Vazzana, C. & Huirne, R. (2003). 195–205.
Evaluation of sustainability of organic, integrated and conventional farming systems: *Singh, V. K., Dwivedi, B. S., Shukla, A. K. & Mishra, R. P. (2010). Permanent
a farm and field-scale analysis. Agriculture, Ecosystems & Environment 95, 273–288. raised bed planting of the pigeonpea-wheat system on a typic Ustochrept: effects on
*Pandey, C., Shukla, S. & Obreza, T. A. (2007). Development and evaluation soil fertility, yield, and water and nutrient use efficiencies. Field Crops Research 116,
of soil moisture-based seepage irrigation management for water use and quality. 127–139.
Journal of Irrigation and Drainage Engineering-Asce 133, 435–443. *Singh, V., Srivastava, A., Singh, R. K. & Savita, U. S. (2011). Effect of tillage
*Panneerselvam, P., Hermansen, J. E. & Halberg, N. (2011). Food security of practices and residue management on soil quality and crop yield under maize (Zea
small holding farmers: comparing organic and conventional systems in India. Journal mays)-based cropping system in Mollisol. Indian Journal of Agricultural Sciences 81,
of Sustainable Agriculture 35, 48–68. 1019–1025.
Pe’er, G., Dicks, L. V., Visconti, P., Arlettaz, R., Báldi, A., Benton, T. G., *Smith, G. D., Coughlan, K. J., Yule, D. F., Laryea, K. B., Srivastava, K. L.,
Collins, S., Dieterich, M., Gregory, R. D., Hartig, F., Henle, K., Hobson, Thomas, N. P. & Cogle, A. L. (1992). Soil-management options to reduce runoff
P. R., Kleijn, D., Neumann, R. K., Robijns, T., Schmidt, J., Shwartz, A., and erosion on a hardsetting alfisol in the semiarid tropics. Soil & Tillage Research 25,
Sutherland, W. J., Turbé, A., Wulf, F., Scott, A. V. (2014). EU agricultural 195–215.
reform fails on biodiversity. Science 344, 1090–1092. Smith, F. P., Gorddard, R., House, A. P. N., McIntyre, S. & Prober, S.
Phalan, B., Onial, M., Balmford, A. & Green, R. E. (2011). Reconciling food M. (2012). Biodiversity and agriculture: production frontiers as a framework for
production and biodiversity conservation: land sharing and land sparing compared. exploring trade-offs and evaluating policy. Environmental Science & Policy 23, 85–94.
Science 333, 1289–1291. *Smith, R. G., Gross, K. L. & Robertson, G. P. (2008). Effects of crop diversity
*Philpott, S. M., Bichier, P., Rice, R. & Greenberg, R. (2007). Field-testing on agroecosystem function: crop yield response. Ecosystems 11, 355–366.
ecological and economic benefits of coffee certification programs. Conservation Biology *Smukler, S. M., Sanchez-Moreno, S., Fonte, S. J., Ferris, H., Klonsky, K.,
21, 975–985. O’Geen, A. T., Scow, K. M., Steenwerth, K. L. & Jackson, L. E. (2010).
Biodiversity and multiple ecosystem functions in an organic farmscape. Agriculture,
Pilgrim, E. S., Macleod, C. J. A., Blackwell, M. S. A., Bol, R., Hogan, D.
Ecosystems & Environment 139, 80–97.
V., Chadwick, D. R., Cardenas, L., Misselbrook, T. H., Haygarth, P. M.,
*Snapp, S. S., Gentry, L. E. & Harwood, R. (2010). Management intensity–not
Brazier, R. E., Hobbs, P., Hodgson, C., Jarvis, S., Dungait, J., Murray, P.
biodiversity–the driver of ecosystem services in a long-term row crop experiment.
J. & Firbank, L. G. (2010). Interactions among agricultural production and other
Agriculture, Ecosystems & Environment 138, 242–248.
ecosystem services delivered from European temperate grassland systems. In Advances
*Sohail-Ijaz, S. & Ali, S. (2011). Replacing clean fallow with minimum tilled pulse
in Agronomy (Volume 109, ed. D. L. Sparks), pp. 117–154. Elsevier Academic Press,
legume in a subtropical dryland at Pothwar, Pakistan. Agrociencia 45, 43–53.
Burlington.
*de Souza, H. N., de Goede, R. G. M., Brussaard, L., Cardoso, I. M., Duarte,
*Porazinska, L., McSorley, R., Duncan, L. W., Gallaher, R. N., Wheaten,
E. M. G., Fernandes, R. B. A., Gomes, L. C. & Pulleman, M. M. (2012).
T. A. & Parsons, L. R. (1998). Relationships between soil chemical status, soil
Protective shade, tree diversity and soil properties in coffee agroforestry systems in
nematode community, and sustainability indices. Nematropica 28, 249–262.
the Atlantic Rainforest biome. Agriculture, Ecosystems & Environment 146, 179–196.
Posthumus, H., Rouquette, J. R., Morris, J., Cowing, D. J. G. & Hess, T. M.
*Tapia-Vargas, M., Tiscareno-Lopez, M., Stone, J. J., Oropeza-Mota, J. L. &
(2010). A framework for the assessment of ecosystem goods and services; a case study
Velazquez-Valle, M. (2001). Tillage system effects on runoff and sediment yield
on lowland floodplains in England. Ecological Economics 69, 1510–1523.
in hillslope agriculture. Field Crops Research 69, 173–182.
*Pretty, J., Smith, G., Goulding, K. W. T., Groves, S. J., Henderson, *Tarkalson, D. D., Hergert, G. W. & Cassman, K. G. (2006). Long-term
I., Hine, R. E., King, V., van Oostrum, J., Pendlington, D. J., Vis, J. effects of tillage on soil chemical properties and grain yields of a dryland winter
K. & Walter, C. (2008). Multi-year assessment of Unilever’s progress towards wheat-sorghum/corn-fallow rotation in the great plains. Agronomy Journal 98, 26–33.
agricultural sustainability II: outcomes for peas (UK), spinach (Germany, Italy), The Natural Capital Project (2011). Integrated valuation of environmental services
tomatoes (Australia, Brazil, Greece, USA), tea (Kenya, Tanzania, India) and oil and tradeoffs.
palm (Ghana). International Journal of Agricultural Sustainability 6, 63–88. *Thierfelder, C. & Wall, P. C. (2012). Effects of conservation agriculture on soil
Pulleman, M., Jongmans, A., Marinissen, J. & Bouma, J. (2003). Effects of organic quality and productivity in contrasting agro-ecological environments of Zimbabwe.
versus conventional arable farming on soil structure and organic matter dynamics Soil Use and Management 28, 209–220.
in a marine loam in the Netherlands. Soil Use and Management 19, 157–165. Thomas, C. D., Anderson, B. J., Moilanen, A., Eigenbrod, F., Heinemeyer,
R Development Core Team (2013). R: A Language and Environment for Statistical Computing. A., Quaife, T., Roy, D. B., Gillings, S., Armsworth, P. R. & Gaston, K. J.
2.14.1 Edition. R Foundation for Statistical Computing, Vienna. (2013). Reconciling biodiversity and carbon conservation. Ecology Letters 16, 39–47.
*Ramesh, P., Ghosh, P. K., Reddy, K. S., Ajay , Ramana, S. & Choudhary, R. *Thomas, G. A., Dalal, R. C., Weston, E. J., Holmes, C. J., King, A. J., Orange,
S. (2005). Assessment of biomass, productivity and sustainability of soybean based D. N. & Lehane, K. J. (2007). Zero tillage and nitrogen fertiliser application in wheat
cropping systems at three levels of nitrogen in deep vertisols of semi-arid tropical and barley on a Vertosol in a marginal cropping area of south-west Queensland.
India. Journal of Sustainable Agriculture 26, 43–59. Australian Journal of Experimental Agriculture 47, 965–975.
*Rasul, G. & Thapa, G. B. (2004). Sustainability of ecological and conventional Tilman, D., Balzer, C., Hill, J. & Befort, B. L. (2011). Global food demand
agricultural systems in Bangladesh: an assessment based on environmental, economic and the sustainable intensification of agriculture. Proceedings of the National Academy of
and social perspectives. Agricultural Systems 79, 327–351. Sciences of the United States of America 108, 20260–20264.
*Reddy, K. N., Zablotowicz, R. M., Locke, M. A. & Koger, C. H. (2003). Cover *Tiwari, K. R., Sitaula, B. K., Bajracharya, R. M. & Borresen, T. (2010).
crop, tillage, and herbicide effects on weeds, soil properties, microbial populations, Effects of soil and crop management practices on yields, income and nutrients losses
and soybean yield. Weed Science 51, 987–994. from upland farming systems in the Middle Mountains region of Nepal. Nutrient
*Rose, L., Coners, H. & Leuschner, C. (2012). Effects of fertilization and cutting Cycling in Agroecosystems 86, 241–253.
frequency on the water balance of a temperate grassland. Ecohydrology 5, 64–72. Torabi, H., Naghdibadi, H. A., Omidi, H., Amirshekari, H. & Miransari, M.
*Salinas-Garcia, J. R., Diaz-Franco, A., Garza-Cano, E. & Garza-Cano, (2008). Effects of soil tillage, canola (Brassica napus L.) cultivars and planting date on
I. (2005). Effects of tillage and biofertilization on the soil properties affecting canola yield, and oil and some biological and physical properties of soil. Archives of
sustainability of the dry bean production. Ciencia y Tecnologia Alimentaria 5, 30–34. Agronomy and Soil Science 54, 175–188.
Saling, P. & Kicherer, A. (2002). Eco-efficiency analysis by BASF: the method. UK National Ecosystem Assessment (2011). The UK National Ecosystem Assessment: Synthesis
International Journal of Life Cycle Assessment 7, 203–218. of the Key Findings. UNEP-WCMC, Cambridge.
*Sankar, G. R. M., Sharma, K. L., Dhanapal, G. N., Shankar, M. A., Mishra, UNDESA (2013). World Population Prospects: The 2012 Revision. Key Findings and Advance
P. K., Venkateswarlu, B. & Grace, J. K. (2011). Influence of soil and fertilizer Tables. United Nations Department of Economic and Social Affairs, New York.
nutrients on sustainability of rainfed finger millet yield and soil fertility in semi-arid *Vasconcelos Pugliesi, A. C., Marinho, M. d. A., Marques, J. F. & Freitas
alfisols. Communications in Soil Science and Plant Analysis 42, 1462–1483. Lucarelli, J. R. (2011). Economical valuation of the erosion effect in soil

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society


Multiple measures of agricultural sustainability 23

managemnet systems employing the replacement cost method. Bragantia 70, *Yan, X. & Gong, W. (2010). The role of chemical and organic fertilizers on yield,
113–121. yield variability and carbon sequestration- results of a 19-year experiment. Plant and
*Verbruggen, E., Kiers, E. T., Bakelaar, P. N. C., Roling, W. F. M. & van Soil 331, 471–480.
der Heijden, M. G. A. (2012). Provision of contrasting ecosystem services by soil *Zhang, S., Li, P., Yang, X., Wang, Z. & Chen, X. (2011a). Effects of tillage and
communities from different agricultural fields. Plant and Soil 350, 43–55. plastic mulch on soil water, growth and yield of spring-sown maize. Soil & Tillage
*Verga, E. G., Leynaud, G. C., Lescano, J. N. & Bellis, L. M. (2012). Is livestock Research 112, 92–97.
grazing compatible with amphibian diversity in the High Mountains of Crdoba, *Zhang, S., Zhang, X., Huffman, T., Liu, X. & Yang, J. (2011b). Soil loss, crop
Argentina? European Journal of Wildlife Research 58, 823–832. growth, and economic margins under different management systems on a sloping
Vermeulen, S. J., Campbell, B. M. & Ingram, J. S. I. (2012). Climate change and field in the black soil area of Northeast China. Journal of Sustainable Agriculture 35,
food systems. Annual Review of Environment and Resources 37, 195–222. 293–311.
Villa, F., Ceroni, M., Bagstad, K., Johnson, G. & Krivov, S. (2009). ARIES Zhang, W., Ricketts, T. H., Kremen, C., Carney, K. & Swinton, S. M.
(artificial intelligence for ecosystem services): a new tool for ecosystem services (2007). Ecosystem services and dis-services to agriculture. Ecological Economics 64,
assessment, planning, and valuation. In Proceedings of the 11th Annual BIOECON 253–260.
Conference on Economic Instruments to Enhance the Conservation and Sustainable Use of *Zuazo, V. H. D., Pleguezuelo, C. R. R., Martinez, J. R. F., Rodiguez, B. C.,
Biodiversity. Venice, Italy, September, 2009. Raya, A. M. & Galindo, P. P. (2008). Harvest intensity of aromatic shrubs vs. soil
*Wang, Q., Chen, H., Li, H., Li, W., Wang, X., McHugh, A. D., He, J. & Gao, H. erosion: an equilibrium for sustainable agriculture (SE Spain). Catena 73, 107–116.
(2009). Controlled traffic farming with no tillage for improved fallow water storage
and crop yield on the Chinese Loess Plateau. Soil & Tillage Research 104, 192–197.
*Wang, Y., Tu, C., Cheng, L., Li, C., Gentry, L. F., Hoyt, G. D., Zhang, X. &
Hu, S. (2011). Long-term impact of farming practices on soil organic carbon and VIII. SUPPORTING INFORMATION
nitrogen pools and microbial biomass and activity. Soil & Tillage Research 117, 8–16.
*Ward, P. R., Micin, S. F. & Fillery, I. R. P. (2012). Application of eddy
covariance to determine ecosystem-scale carbon balance and evapotranspiration in Additional supporting information may be found in the
an agroforestry system. Agricultural and Forest Meteorology 152, 178–188. online version of this article.
*Winqvist, C., Bengtsson, J., Aavik, T., Berendse, F., Clement, L. W., Figure S1. Factors contributing to uncertainty around mean
Eggers, S., Fischer, C., Flohre, A., Geiger, F., Liira, J., Pärt, T., Thies,
C., Tscharntke, T., Weisser, W. W. & Bommarco, R. (2011). Mixed effects of Spearman’s rank correlation (rs ) estimates.
organic farming and landscape complexity on farmland biodiversity and biological
control potential across Europe. Journal of Applied Ecology 48, 570–579 Blackwell
Figure S2. Effect of zero-tillage versus conventional tillage
Publishing Ltd. treatments on multiple response variables, split by study
Wossink, A. & Swinton, S. M. (2007). Jointness in production and farmers’ duration.
willingness to supply non-marketed ecosystem services. Ecological Economics 64,
297–304. Figure S3. Effect of zero-tillage versus conventional tillage
*Woyessa, Y. E. & Bennie, A. T. P. (2004). Factors affecting runoff and soil loss
under simulated rainfall on a sandy Bainsvlei Amalia soil. South African Journal of Plant
treatments on multiple response variables, split by crop type.
and Soil 21, 203–208. Table S1. Full list of studies used in the meta-analysis.
*Wyngaard, N., Echeverria, H. E., Sainz Rozas, H. R. & Divito, G. A. (2012).
Fertilization and tillage effects on soil properties and maize yield in a Southern Table S2. Study attributes contributing to variation in
Pampas Argiudoll. Soil & Tillage Research 119, 22–30.
*Yadvinder, S., Gupta, R. K., Gurpreet, S., Jagmohan, S., Sidhu, H. S. & Bijay,
Spearman’s rank correlation between pairs of responses
S. (2009). Nitrogen and residue management effects on agronomic productivity and (level 1).
nitrogen use efficiency in rice-wheat system in Indian Punjab. Nutrient Cycling in
Agroecosystems 84, 141–154. Appendix S1. Literature key word searches.

(Received 28 April 2015; revised 13 December 2015; accepted 15 December 2015 )

Biological Reviews (2016) 000–000 © 2016 Cambridge Philosophical Society

Potrebbero piacerti anche