Journal of the Indian Society of Remote Sensing

Estimation of Sea Surface Salinity Concentration from Landsat 8 OLI Data

--Manuscript Draft--

Manuscript Number: ISRS-D-19-00100

Full Title: Estimation of Sea Surface Salinity Concentration from Landsat 8 OLI Data

Article Type: Research Article

Abstract: Remote sensing technique to estimate the sea surface salinity has been widely
implemented. The interface between them was an algorithm that is developed using a
regression equation such as Son’s algorithm in previous research. However, the use of
this algorithm for waters in Indonesia, especially in Madura Strait, still requires a
number of adjustment since it is related to the characteristics of different areas in which
the algorithm was developed. The development of an applicable local algorithm was
performed by finding the best coefficient value in estimating sea surface salinity by
considering the value of its lowest NMAE (Normalized Mean Absolute Error). By using
salinity and in-situ Rrs(λ) (Reflectance of remote sensing) data, it has been found that
the coefficient for slope was -0.0092 and intercept was 1.4903 The developed
algorithm produce higher accuracy than the existing algorithm with NMAE of 0.51%.
This NMAE value is smaller than previous research.

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
Blinded Manuscript

1. Introduction
1
2 Salinity in seawater has an important role in the evolution of seawater microbial organism such as
3
4 macroalgae or seaweeds as well as other organisms living in the same place (Lobban & Harrison, 1997). This
5
6 phenomenon can occur since salinity affects the osmoregulation process in aquatic plants (Aslan, 1991; James W.
7
8 Nybakken, 2001). In addition, salinity can also be an indicator whether or not sea water can be converted into salt
9
10 (Nifizah, Jaelani, & Winarso, 2016). Thus, the information of salinity concentration becomes important for salt
11
12 farmers as well as in the area of planning development of seaweed farming.
13
14 Currently, are there two methods for measuring the concentration of seawater: the in-situ method, a
15
16 method of estimating the salinity by directly measure it on the sea as well as by collecting seawater sample then
17
18 measure it in the laboratory; and indirect method, the estimation of seawater salinity by using remote sensing data
19
20 (Budhiman, 2012; Scale, 2009; Wang & Xu, 2008). However, by considering the spatial and temporal
21
22 heterogeneity of water bodies, extracting water information by remote sensing techniques can be more effective
23
24 approach than an in-situ field measurement (Liu, Islam, & Gao, 2003). Sea surface salinity (SSS) concentration
25
26 could be retrieved by analyzing satellite images without direct contact with the object (George, 2005; Lillesand,
27
28 Kiefer, & Chipman, 2004).
29
30 SSS retrieval from satellite image is strongly dependent on the accuracy of surface reflectance produced
31
32 through atmospheric correction algorithm and parameter retrieval algorithm (L. M. L. M. Jaelani, Matsushita,
33
34 Yang, & Fukushima, 2013; L. M. Jaelani et al., 2016; L. M. Jaelani, Matsushita, Yang, & Fukushima, 2015).
35
36 Currently, there are several existing SSS retrieval algorithm that have been developed and frequently used by
37
38 researchers (Nifizah et al., 2016) such as algorithm of Young Baek Son (Son et al., 2012), Wouthuyzen Sam
39
40 (Wouthuyzen et al., 2008), Ahn YH (Ahn., Shanmugam, Moon, & Ryu, 2008), Yan Bai (Bai et al., 2013), Hiroaki
41
42
Sasaki (Sasaki, Siswanto, Nishiuchi, Tanaka, & Hasegawa, 2008) and Binding (Binding & Bowers, 2003).
43
44
Because of SSS algorithms are site-specific and time-dependent, these existing algorithms which were mainly
45
46
developed and tested on water area outside of Indonesia’s waters may not be suitable with the characteristic and
47
48
condition in the area of the current research. Therefore, the objective of this study is intended to validate the
49
50
51 performance of the existing algorithm and develop the new one that suitable for geographical conditions of the
52
53 waters in Indonesia, especially in Madura Strait (Muhsi, Sukojo, Taufik, & Aji, 2017, 2018).
54
55 2. Study Area
56
57 The study area was located at Madura Strait of East Java Province, Indonesia. It is located between 07°
58
59 08' 30"S - 07° 44' 27"S Latitude and 112° 39' 23"E - 114° 05' 24"E Longitude (see Figure 1); between two islands,
60
61
62 1
63
64
65
 i.e. Java and Madura. Madura Strait, in the north side of it, is bordered by four regencies: Bangkalan, Sampang
1
2 and Sumenep, and Pamekasan respectively. Over the past few years, the waters in the Strait of Madura were used
3
4 as raw materials for producing salt and as seaweed farmland. According to data from the Ministry of Marine
5
6 Affairs and Fisheries of the Republic of Indonesia, for a national salt production run by PT. Garam company,
7
8 Madura has contributed 17% of total production of 19201.9 tons-per-August 2017. Thus, it is crucial to know and
9
10 map the sea surface salinity in Madura strait as a consideration for using it in various interests, especially salt and
11
12 seaweed farming.
13
14 3. Materials and Methods
15
16 3.1. Data collection
17
18 Field data was taken in Madura strait waters, consisted of SSS data (collected by a refractometer) and in-
19
20 situ Rrs() (indirect recorded by a Spectroradiometer). The field campaign and acquisition of Landsat 8 data were
21
22 performed on June 2, 2016. From that field survey, there were 20 SSS data as well as apparent optical properties
23
24 of water (AOP) such as upward radiance (Lu), downward radiance atmosphere (Ls) and downward irradiance
25
26 atmosphere (Ed). Since the Rrs() value cannot be directly measured in the field (Mobley, 1999b), it was calculated
27
28
by using the equation of Gordon and Mobley (Gordon et al., 1988; Mobley, 1999b) in (Budhiman, 2012).
29
30
3.2. Insitu data processing
31
32
The AOP data was measured by using spectroradiometer Tri-OS RAMSES with a wavelength range
33
34
35 320nm–950nm and intervals of approximately 3.3nm. Tri-OS RAMSES used is hyperspectral spectroradiometer
36
37 which was fitted into Landsat 8 OLI bands: Band 1 (433nm-453nm), Band 2 (450nm-515nm), Band 3 (525nm-
38
39 600nm), Band 4 (630nm-680nm) dan Band 5 (845nm-885nm). The formula for calculating Rrs() as follows :
40
𝐿𝑤 (𝜆)
41 𝑅𝑟𝑠 (𝜆) = (1)
𝐸𝑑 (𝜆)
42
43
44 𝐿𝑤 (𝜆) = 𝐿𝑢 (𝜆) − 𝜌𝑠 𝐿𝑠 (𝜆) (2)
45
46 where Lw (water leaving radiance / Wm-2sr-1) is the radiance value obtained from the water column and radiance
47
48 reflected directly by the thin layer of sea surface (Nababan, Wirapramana, & Arhatin, 2013), Ls (downward
49
50 radiance atmosphere / Wm-2sr-1), ρs is a part of the reflectance on the surface of waters from reflected sunlight. ρs
51
52 can be calculated using Fresnel formula (Hecth, Coffey, & Dolan, 2002) (Budhiman, Suhyb, Vekerdy, & Verhoef,
53
54 2012). ρs can also use the constant value of research results (Mobley, 1999a) or the constant value of 0.02
55
56 (Nababan et al., 2013).
57
58 The first step in the calculation was started by calculating diffuse attenuation coefficient (Kd) from
59
60 downward irradiance atmosphere (Ed). Kd is the coefficient to see the propagation of light absorption by water
61
62 2
63
64
65
 column from the surface to the ocean depth (Zhongping Lee & Carder, 2005). Kd is also very important for marine
1
2 analysis such as water class classification or light intensity in water depth (Zhongping Lee & Carder, 2005;
3
4 ZhongPing Lee, Carder, & Arnone, 2002; Zhongping Lee, Du, & Arnone, 2005; Wei & Lee, 2013). Based on
5
6 research conducted by Lee at al., there are three methods for calculating the value of Kd. First, estimating the Kd
7
8 at a wavelength of 490 by using empirical algorithm; other Kd for the other wavelengths can be calculated
9
10 afterward (Austin & Petzold, 1981. 1986; Zhongping Lee & Carder, 2005). Second, using the value of chlorophyll-
11
12 a. After the value of the first chlorophyll a was found, then it can be used to estimate the values of other Kd and
13
14 chlorophyll-a (Zhongping Lee & Carder, 2005; Morel, 1988; Morel & Maritorena, 2001). Third, using a method
15
16 of numerical simulation of flow movement of light in the sea (Zhongping Lee & Carder, 2005). The method was
17
18 semi analytically calculation of Kd from Rrs() calculated previously using the formula (1). But, the value of
19
20 the absorption and backscattering of light has been previously calculated as an input coefficient to the semi-
21
22 analytical model for calculating Kd of each wavelength. That process uses the quasi-analytical method wherein
23
24 the Rrs is becoming its data input (Zhongping Lee & Carder, 2005; Wei & Lee, 2013).
25
26 Meantime, Son YB calculates the value of Kd using the formula of maximum normalized difference
27
28 carbon index (MNDCI) in developing the SSS algorithm (Son et al., 2012). Son YB determined the Kd value by
29
30 using multiple wave approach by assuming that when the particle concentration level increases, the peak radiance
31
32 shifted to a longer wavelength band; in this case, he used the maximum normalized difference carbon index
33
34
(MNDCI) formula. However, the multi-spectral MNDCI approach (Formula 3) was used for a consideration that
35
36
the values obtained would be more accurate than the wavelength ratio or single beam, particularly for waters
37
38
containing a mixture of organic and inorganic components. Through the value of MNDCI, the Kd will be
39
40
41 determined (formula 4). Meantime, this study used Son’s formula to calculate the value of Kd since it will also
42
43 be implemented to predict the sea surface salinity, especially in Madura Strait.
44 [𝑛𝐿𝑤(555)−𝑚𝑎𝑥(𝑛𝐿𝑤(412),𝑛𝐿𝑤(443),𝑛𝐿𝑤(490))]
45 MNDCI = (3)
[𝑛𝐿𝑤(555)+𝑚𝑎𝑥(𝑛𝐿𝑤(412),𝑛𝐿𝑤(443),𝑛𝐿𝑤(490))]
46
47 Where MNDCI is maximum normalized difference carbon index, nLw is spectral normalized water-
48
49 leaving radiance. While the Kd is determined using the following model :
50
51 3 +0.96𝑥𝑀𝑁𝐷𝐶𝐼 2 +1.14𝑥𝑀𝑁𝐷𝐶𝐼−0.25]
52 𝐾𝑑 = 10[0.70𝑥𝑀𝑁𝐷𝐶𝐼 (4)
53
54 Where Kd is attenuation coefficient, MNDCI is obtained from Formula 3. After the Kd for each
55
56 wavelength has been obtained, the analysis will be done using Son’s salinity algorithm (formula 5).
57
58 𝑆𝑆𝑆 = 10[−0.141𝑥𝑙𝑜𝑔10 (𝐾𝑑)+1.45] (5)
59
60
61
62 3
63
64
65
 Where SSS is sea surface salinity that would be predicted and Kd is the attenuation coefficient obtained
1
2 from Formula 4. SSS of Son algorithm results will be correlated with in-situ SSS data. Furthermore, the accuracy
3
4 of the algorithm model will be tested using Normalized Mean Absolute Error (NMAE) method.
5
6 1 𝑥 𝑒𝑠𝑡𝑖−𝑥 𝑚𝑒𝑎𝑠
7 𝑁𝑀𝐴𝐸(100%) = ∑ | | × 100 (6)
𝑁 𝑥 𝑚𝑒𝑎𝑠
8
9 Where NMAE is a value to see the average absolute error between the value of the SSS obtained from
10
11 measurement and in-situ which then normalized by percentage; x esti is in-situ SSS data and x meas is SSS data
12
13 obtained from the model. At last, this algorithm model then implemented using Landsat 8 data and validated using
14
15 in-situ data.
16
17
3.3. Satellite data processing
18
19
The data used in the implementation of the prediction method was obtained from Landsat 8 OLI
20
21
(path/row of 118 and row 65) that available for public on https://earthexplorer.usgs.gov/ or http://glovis.usgs.gov/
22
23
24 Landsat 8 OLI images were taken in the Landsat Collection 1 Level-2 (On-Demand) directory, that is
25
26 Landsat 8 OLI/TIRS C1 Level-2 data. In this case, to obtain the Rrs() value of sea surface on each pixel of Landsat
27
28 8 OLI images, the pixel value should be divided by 10000 to obtain the sea surface reflectance value. Then, the
29
30 result will be divided by a constant value of phi (π) to obtain the Rrs()
31
32 4. Results and discussion
33
34 4.1. TriOS Ramses data and in-situ salinity correlation with the result of estimation
35
36
37
There was three data parameter were retrieved from in-situ AOP data collected using TriOS Ramses
38
39 spectroradiometer, they were upward radiance (Lu), downward radiance atmosphere (Ls) and downward irradiance
40
41 atmosphere (Ed), as shown in Figure 2.
42
43 Furthermore, the following step is calculating Rrs() value using Formula 1 and 2, then MNDCI and
44
45 Kd were calculated based onthe Rrs() (Formula 3 dan 4). However, this study used only band 2 and band 3
46
47 since they accord with the ratio formula in MNDCI as shown in Table 1. Here, the Kd acts as an input value in
48
49 Son algorithm to estimate the sea surface salinity. Based on the Kd in Table 1, two things can be obtained; first,
50
51 salinity values which is in the form of particle salinity unit (psu) as seen in Table 2, second, a graph of the
52
53 correlation between in-situ SSS values and the SSS values obtained from the estimation using the Son algorithm,
54
55 wherein the NMAE value is still high, i.e. 3,4% (see Figure 3).
56
57 4.2. Developed algorithm
58
59
60
61
62 4
63
64
65
 The algorithm model was then developed using the regression model (Figure 4) as the development of
1
2 the Son algorithm (Son et al., 2012). The algorithm will be used to estimate the sea surface salinity of Madura
3
4 Strait.
5
6 𝑆𝑆𝑆 = 10[−0.0092×(𝑙𝑜𝑔10 (𝐾𝑑))+1.4903] (7)
7
8 Where -0.0092 and 1.4903 were the coefficient and the constant respectively, obtained from the
9
10 development of the algorithm.
11
12 The result of the developed algorithm model was applied to different in-situ data with the data used
13
14 previously when constructing the model. The estimation results of the SSS using the new algorithm model was
15
16 shown in Table 3 and Figure 5.
17
18 4.3. Validation
19
20 In this current study, validation was intended to test the validity of the new algorithm model (Formula
21
22 7) in estimating the sea surface salinity.
23
24 The Normalized Mean Absolute Error (NMAE) value was used to measure the validity of the algorithm.
25
26 It is used to measure how close the SSS predicted value with the in-situ SSS, or to measure the level of absolute
27
28 error of SSS prediction model which is in the form of a percentage. Further, the results of the calculation using
29
30 the developed algorithm model was a very small NMAE value which is below 1%, i.e. 0.47%. This indicated that
31
32 the developed algorithm model can be applied to estimate the SSS value of Madura Strait.
33
34
4.4. Model Implementation
35
36
The developed algorithm then applied using Rrs() of Landsat 8 OLI. The value then converted into Kd
37
38
39 value using Son’s formula (Son et al., 2012). The sea surface salinity in Madura strait water then calculated based
40
41 on that Kd using algorithm model which was previously developed. The comparison of the two salinity values
42
43 obtained from Landsat 8 OLI and in-situ observation can be seen in Figure 6. And Table 4. The distribution map
44
45 of SSS in Madura strait recorded on June 2, 2016 was shown in Figure 7.
46
47 5. Conclusions
48
49 The development of an algorithm model to estimate the Sea Surface Salinity (SSS) in Madura strait
50
51 produced a high accuracy. The indication can be clearly seen on the average difference of the salinity estimation
52
53 results with in-situ salinity value which was very small, 0.0047; in other words, the NMAE value was just 0.47%.
54
55 Likewise, when it was applied to the Landsat 8 OLI images, the average difference between the estimated SSS
56
57 and the in-situ SSS reached NMAE of 0.51%.
58
59 Acknowledgment
60
61
62 5
63
64
65
 We are thankful for the Indonesian National Institute of Aeronautics and Space for the supporting in-
1
2 field campaign.
3
4 References
5
6 Ahn., Y. H., Shanmugam, P., Moon, J. E., & Ryu, J. H. (2008). Satellite remote sensing of a low-salinity water
7
8 plume in the East China Sea. Annales Geophysicae, 26(7), 2019–2035.
9
10 Aslan, L. M. (1991). Budidaya Rumput Laut. Yogyakarta: Penerbit Kanisius.
11
12 Austin, R. W., & Petzold, T. J. (1981). The Determination of the Diffuse Attenuation Coefficient of Sea Water
13
14 Using the Coastal Zone Color Scanner. Marine Science Springer, Boston, MA, 13, 239–256.
15
16 Austin, R. W., & Petzold, T. J. (1986). Spectral dependence of the diffuse attenuation coefficient of light in
17
18 ocean waters. Opt. Eng, 25, 473–479.
19
20 Bai, Y., Pan, D., Cai, W., He, X., Wang, D., & Tao, B. (2013). Remote sensing of salinity from satellite-derived
21
22 CDOM in the Changjiang River dominated East China Sea. JOURNAL OF GEOPHYSICAL RESEARCH:
23
24 OCEANS, 118, 227–243. https://doi.org/10.1029/2012JC008467
25
26 Binding, C., & Bowers, D. (2003). Measuring the salinity of the Clyde Sea from remotely sensed ocean color.
27
28 Estuarine, Coastal and Shelf Science, 57, 605–611.
29
30 Budhiman, S. (2012). Perbandingan Karakteristik Spektral (Spectral Signature) Parameter Kualitas Perairan
31
32 Pada Kanal Landsat ETM + dan Envisat Meris (Comparion of Water Constituents Spectral Signature on
33
34 Landsat ETM+ and Envisat Meris Band. Jurnal Pengideraan Jauh, 9(2), 76–89.
35
36 Budhiman, S., Suhyb, M., Vekerdy, Z., & Verhoef, W. (2012). Deriving optical properties of Mahakam Delta
37
38 coastal waters , Indonesia using in situ measurements and ocean color model inversion. ISPRS Journal of
39
40 Photogrammetry and Remote Sensing, 68, 157–169. https://doi.org/10.1016/j.isprsjprs.2012.01.008
41
42
George, J. (2005). Fundamentals of Remote Sensing (Second Edi). Hyderabad: Universities Press (India).
43
44
Gordon, H., Brown, J. W., Evans, R. H., Smith, R. C., Gordon, R., Brown, O. B., … Baker, K. (1988). A
45
46
semianalytic radiance model of ocean color, (July 2014). https://doi.org/10.1029/JD093iD09p10909
47
48
Hecth, E., Coffey, M., & Dolan, P. (2002). OPTICS Fourth Edition (4th ed.). San Francisco: Addison Wesley.
49
50
51 Jaelani, L. M. L. M., Matsushita, B., Yang, W., & Fukushima, T. (2013). Evaluation of four MERIS
52
53 atmospheric correction algorithms in Lake Kasumigaura, Japan. International Journal of Remote Sensing,
54
55 34(24), 8967–8985. https://doi.org/10.1080/01431161.2013.860660
56
57 Jaelani, L. M., Limehuwey, R., Kurniadin, N., Pamungkas, A., Koenhardono, E. S., & Sulisetyono, A. (2016).
58
59 Estimation of Total Suspended Sediment and Chlorophyll-A Concentration from Landsat 8-Oli: The
60
61
62 6
63
64
65
 Effect of Atmospher and Retrieval Algorithm. IPTEK The Journal for Technology and Science, 27(1), 16–
1
2 23. https://doi.org/10.12962/j20882033.v27i1.1217
3
4 Jaelani, L. M., Matsushita, B., Yang, W., & Fukushima, T. (2015). An improved atmospheric correction
5
6 algorithm for applying MERIS data to very turbid inland waters. International Journal of Applied Earth
7
8 Observation and Geoinformation, 39, 128–141. https://doi.org/10.1016/j.jag.2015.03.004
9
10 James W. Nybakken. (2001). Marine Biology : An Ecological Approach (5th ed.). New York: Benjamin-
11
12 Cummings Publishing Company.
13
14 Lee, Z., & Carder, K. L. (2005). Diffuse Attenuation Coefficient of Downwelling Irradiance : An Evaluation of
15
16 Remote Sensing Methods. Journal of Geophysical Research, 110(10).
17
18 Lee, Z., Carder, K. L., & Arnone, R. a. (2002). Deriving inherent optical properties from water color: a
19
20 multiband quasi-analytical algorithm for optically deep waters. Applied Optics, 41(27), 5755–5772.
21
22 Lee, Z., Du, K., & Arnone, R. (2005). A model for the diffuse attenuation coefficient of downwelling irradiance.
23
24 Journal of Geophysical Research, 110(C2), 1–25. https://doi.org/10.1029/2004JC002275
25
26 Lillesand, T. M. K. M., Kiefer, R. W., & Chipman, J. (2004). Remote Sensing and Image Interpretation (5th
27
28 ed.). New York: John Wiley {&} Sons, Inc.
29
30 Liu, Y., Islam, M. A., & Gao, J. (2003). Quantification of shallow water quality parameters by means of remote
31
32 sensing. Progress in Physical Geography, 27(1), 24–43. https://doi.org/10.1191/0309133303pp357ra
33
34 Lobban, C. S., & Harrison, P. J. (1997). Seaweed Ecology and Physiology. Cambridge: Cambridge University
35
36 Press.
37
38 Mobley, C. D. (1999a). Estimation of Remote Sensing Reflectance from Above-Surface Measurements. Applied
39
40 Optics, 38(36), 7442–7455.
41
42
Mobley, C. D. (1999b). Estimation of the remote-sensing reflectance from above-surface measurements.
43
44
Applied Optics, 38(36), 7442–7455.
45
46
Morel, A. (1988). Optical modeling of the upper ocean in relation to its biogenous matter content (case 1
47
48
waters). J. Geophys. Res, 93, 10749– 10768.
49
50
51 Morel, A., & Maritorena, S. (2001). Bio-optical properties of oceanic waters: A reappraisal. J. Geophys. Res,
52
53 106, 7163–7180.
54
55 Muhsi, M., Sukojo, B. M., Taufik, M., & Aji, P. (2017). Model Pendugaan Kandungan Sulfat di Air Laut
56
57 Menggunakan Citra Satelit Landsat 8 OLI. In Konferensi Nasional Teknik Sipil dan Infrastruktur (pp. 13–
58
59 22). Jember: Jurusan Teknik Sipil Universitas Jember.
60
61
62 7
63
64
65
 Muhsi, M., Sukojo, B. M., Taufik, M., & Aji, P. (2018). Estimation Algorithm of Sulfate Concentration at The
1
2 Sea Surface Based on Landsat 8 OLI Data. Journal of Theoretical and Applied Information Technology,
3
4 96(17), 5741–5753.
5
6 Nababan, B., Wirapramana, A. A., & Arhatin, R. E. (2013). Spectral of Remote Sensing Reflectance of Surface
7
8 Waters. Jurnal Ilmu Dan Teknologi Kelautan Tropis, 5(1), 69–84.
9
10 Nifizah, Jaelani, L. M., & Winarso, G. (2016). Evaluasi Algoritma Wouthuyzen dan Son untuk Pendugaan Sea
11
12 Surface Salinity (SSS) (Studi Kasus: Perairan Utara Pamekasan). JURNAL TEKNIK ITS, 5(September).
13
14 Sasaki, H., Siswanto, E., Nishiuchi, K., Tanaka, K., & Hasegawa, T. (2008). Mapping the low salinity
15
16 Changjiang Diluted Water using satellite- retrieved colored dissolved organic matter ( CDOM ) in the East
17
18 China Sea during high river flow season. GEOPHYSICAL RESEARCH LETTERS, 35, 1–6.
19
20 https://doi.org/10.1029/2007GL032637
21
22 Scale, S. (2009). SATELLITE REMOTE SENSING : SALINITY MEASUREMENTS, 127–132.
23
24 Son, Y. B., Gardner, W. D., Richardson, M. J., Ishizaka, J., Ryu, J.-H., Kim, S.-H., & Lee, S. H. (2012). Tracing
25
26 offshore low-salinity plumes in the Northeastern Gulf of Mexico during the summer season by use of
27
28 multispectral remote-sensing data. Journal of Oceanography, 68(5), 743–760.
29
30 https://doi.org/10.1007/s10872-012-0131-y
31
32 Wang, F., & Xu, Y. J. (2008). Development and application of a remote sensing-based salinity prediction model
33
34 for a large estuarine lake in the US Gulf of Mexico coast, 184–194.
35
36 https://doi.org/10.1016/j.jhydrol.2008.07.036
37
38 Wei, J., & Lee, Z. (2013). Model of the attenuation coefficient of daily photosynthetically available radiation in
39
40 the upper ocean. Methods in Oceanography, 8, 56–74. https://doi.org/10.1016/j.mio.2013.12.001
41
42
Wouthuyzen, S., Tarigan, S., Indarto, H., Supriyadi, Sediadi, A., Sugarin, … Ishizaka, J. (2008). Pengukuran
43
44
Salinitas Permukaan Teluk JakartaMelalui Penginderaan Warna Laut MenggunakanData Multi-Temporal
45
46
Citra Satelit Landsat-7 ETM+. In PIT MAPIN XVII. Bandung.
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62 8
63
64
65
Figure

Figure 1. Study area

upward radiance downward radiance

1600 120.00
Intensity mW/(m^2 nm)

Intensity mW/(m^2 nm)

1400 100.00
1200
80.00
1000
800 60.00

600 40.00
400
20.00
200
0 0.00
0.000 200.000 400.000 600.000 800.0001000.0001200.000 0.000 200.000 400.000 600.000 800.000 1000.000
Wavelength (nm) Wavelength (nm)

downward irradiance
30.00
Intensity mW/(m^2 nm)

25.00
20.00
15.00
10.00
5.00
0.00
0.000 200.000 400.000 600.000 800.000 1000.0001200.000
Wavelength (nm)

Figure 2. Graph of water upward radiance, atmosphere downward radiance and downward
irradiance

1
 33.00

32.50
NMAE : 3.4%
32.00
SSS ESTIMATION RESULT

31.50
31.00

30.50
30.00
29.50
29.00

28.50
30.7 30.8 30.9 31 31.1 31.2 31.3 31.4
SSS IN SITU

Figure 3. The Correlation of the estimated and the in-situ SSS using Son algorithm

1.496
1.495
1.494
LOG SSS INSITU

1.493
1.492
1.491
1.49

y = -0.0092x + 1.4903 1.489

1.488
1.487
-0.5 -0.4 -0.3 -0.2 -0.1 0

LOG KD INSITU

Figure 4. Graphical model of the developed algorithm

2
 31.3

NMAE : 0.47%
31.2
SSS ESTIMATION RESULST

31.2

31.1

31.1

31.0
30.7 30.8 30.9 31 31.1 31.2 31.3 31.4 31.5
SSS INSITU

Figure 5. In-situ and estimated SSS correlation: new algorithm

31.20
NMAE = 0.51%
31.15
n = 20
31.10
SSS LANDSAT 8

31.05

31.00

30.95

30.90

30.85
30.7 30.8 30.9 31 31.1 31.2 31.3 31.4 31.5

SSS INSITU

Figure 6. The correlation of in-situ SSS and estimated SSS from Landsat 8 OLI

3
Figure 7. The distribution map of SSS in Madura Strait

4
Table

Table 1. Rrs(), MNDCI and Kd values

Station Rrs MNDCI Kd

Band 2 Band 3
2 0.016769 0.017564 0.023152 0.598552
4 0.018121 0.021336 0.081479 0.708418
6 0.016342 0.019711 0.093437 0.734883
8 0.011965 0.01083 -0.04978 0.495612
10 0.008756 0.005702 -0.21122 0.349746
12 0.011457 0.008881 -0.12668 0.415487
14 0.010555 0.007372 -0.17758 0.373676
16 0.011419 0.007939 -0.1798 0.37201
18 0.015111 0.014813 -0.00994 0.547882
20 0.019214 0.024202 0.114879 0.786342

Table 2. The in-situ and the estimation SSS with Son algorithm
Station Sea Surface Salinity
In situ Estimation
2 31.15 30.30
4 30.78 29.59
6 30.89 29.43
8 30.91 31.12
10 31.18 32.68
12 31.18 31.90
14 31.25 32.38
16 31.3 32.40
18 31.23 30.68
20 31.28 29.16

1
Table 3. The in-situ and estimated SSS results with new algorithm
Station Sea Surface Salinity
In situ Estimation
1 31.18 31.1
3 30.95 31.0
5 30.82 31.0
7 30.99 31.1
9 31.01 31.2
11 31.00 31.2
13 31.27 31.1
15 31.38 31.2
17 31.27 31.1
19 31.26 31.0

Table 4. In-situ SSS and estimation SSS calculated with new algorithm to the Landsat 8 OLI images
Station Sea Surface Salinity
In situ Estimation
1 31.18 30.88
2 31.15 30.99
3 30.95 30.97
4 30.78 30.94
5 30.82 30.92
6 30.89 30.97
7 30.99 31.04
8 30.91 31.06
9 31.01 31.11
10 31.18 31.16
11 31.00 31.09
12 31.18 31.03
13 31.27 31.09
14 31.25 31.18
15 31.38 30.98
16 31.3 31.18
17 31.27 31.12
18 31.23 30.99
19 31.26 30.97
20 31.28 30.91