The IJICS (International Journal of Informatics and Computer Science) Vol 5 No 1, March 2021, Page 57-63 ISSN 2548-8384 (online), ISSN 2548-8449 (print) Available Online at https://ejurnal.stmik-budidarma.ac.id/index.php/ijics/index DOI 10.30865/ijics.v5i1.2954 Estimation Chlorophyll-a Using Landsat-8 Imagery in Shallow Water: Effect of Atmospheric and Algorithm

Abdi Sukmono*, Lilik Kristianingsih Department of Geodetic Engineering, Faculty of Engineering, Diponegoro University, , Email: 1,*[email protected], [email protected] Coressponding Author: 1,*[email protected] Submitted: 06/03/2021; Accepted: 28/03/2021; Published: 29/03/2021 Abstract−Chlorophyll-a estimation using remote sensing technology is being challenged to improve its accuracy. Various algorithms and correction methods need to be studied, including the influence of the atmosphere. It can influence the passage of the electromagnetic wave from the sun to the object and from the object to the sensor that makes the difference on the image reflectance. There are two kinds of reflectance; which are ToA (Top of Atmosphere) reflectance and BoA (Bottom of Atmosphere) reflectance. ToA reflectance is the reflectance captured by the sensor yet BoA reflectance is the reflectance of the object corrected by the atmosphere. ToA reflectance is produced by radiometric calibration while BoA reflectance is made of atmospheric correction process. This research studies aims to compare those to reflectance to investigate the impact of atmospheric correction usage on chlorophyll-a case study. The waters of Wedung district is chosen as the research field because it is the largest area in Demak regency. This study used DOS (Dark Object Substractions), FLAASH (Fast Line of sight Atmospheric Analysis of Spectral Hypercubes), and 6SV (Second Simullations of a Satellite Signal in the Solar Spectrum) correction method. To compare between ToA and BoA reflectance in the calculation of chlorophyll-a, the writer used the algorithms of Wouthuyzen, Wibowo, Pentury, Much Jisin and also Lestari Laksmi. The result shown is that the usage of BoA image reflectance (atmospherically corrected) had a better model result than ToA image reflectance. This is indicated from the consecutive determination coefficient values of ToA reflectance which are 0,2292; 0,4562; 0,2292; 0,2252. Meanwhile the consecutive coefficient values of BoA reflectance by DOS correction method are 0,5251; 0,5575; 0,5251; 0,6939; by FLAASH correction method are 0,6168, 0,5041, 0,6168, 0,614; by 6SV method are 0,6436; 0,4033; 0,6436; 0,6365. The result of hypothesis and validation test is that atmospheric correction significantly influences on the calculation of chlorophyll-a in Wedung district except using Wibowo algorithm. Keywords: Chlorophyll-a Algorithm; Effect of Athmospheric; Landsat 8; Shallow Water

1. INTRODUCTION Chlorophyll-a is a very important parameter in the study of the level of productivity of ocean waters. According to [1] , Chlorophyll-a is a pigment of phytoplankton that can be used as a parameter marine productivity. Chlorophyll-a concentration above 0.2 mg / L indicate the presence of phytoplankton life. This can be used as a marker where fishing is combined with other oceanographic parameters. Because phytoplankton are the food source for fish surface. In the development of data chlorophyll-A is also often used as a parameter to indicate the quality of the water environment. Mapping the distribution of chlorophyll-a conventionally by taking samples of the data field is currently less effective, because it requires considerable time and energy are adequate. Though data on the distribution of chlorophyll- A is needed quickly to make ocean model. Therefore, today the determination of the value of chlorophyll-a lot of remote sensing using satellite technology. Some satellites have been designed and used as a means of mapping the distribution of chlorophyll-a. Determination of chlorophyll-a using remote sensing technology still needs to increase accuracy. One of the techniques used is atmospheric correction. It is a process to eliminate errors caused by atmospheric influences on the image. Atmospheric correction able to eliminate the influence from the atmosphere and return the reflectance value according to the actual object's reflectance value on the earth's surface [2]. The correction is done by considering various atmospheric parameters in the correction process, including seasonal factors, and the climatic conditions at the location of the image recording (eg tropical, sub-tropical, and others). It features the ability to fix atmospheric disturbances such as mist, smoke, and other. The influence of the atmosphere occur during the recording process the image in which the electromagnetic waves from the sun to the earth's surface and from the object to the sensor impaired as it passes through the atmosphere, the disorder can be either scattering or absorption. This will have an impact on image data obtained, in which the data recorded by the satellite sensors with data on the object will be different.

Figure 1. Atmospheric influence on the recordings of satellite imagery [3]

Copyright © 2021, Abdi Sukmono. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Page 57

The IJICS (International Journal of Informatics and Computer Science) Vol 5 No 1, March 2021, Page 57-63 Abdi Sukmono, Estimation Chlorophyll-a Using Landsat-8 Imagery in Shallow Water: Effect of Atmospheric and Algorithm

Figure 1 shows that the atmosphere can influence the electromagnetic waves from the sun to the object and from the object to the sensor which causes an error in the image data, in which the image data obtained with the desired data is not the same such errors can be minimized by performing atmospheric correction. In satellite image processing, there are several models of atmospheric correction. In general atmospheric correction models are often used include: 1. DOS Atmospheric correction. According [4], the principle of this method is to improve the radiometric values (pixel value in the image due to atmospheric disturbances). If there is no atmosphere, dark colored object or usually in the form of water and the shadow of the cloud should have a pixel value of 0, if the object is not worth 0 then the value is biased. 2. FLAASH Atmospheric correction. It corrected image by suppressing or eliminating the effect of water vapor, oxygen, carbon dioxide, methane, ozone and molecular and aerosol scattering by radiative transfer codes MODTRAN-4. This correction is applied to each pixel [5]. 3. 6SV Atmospheric correction. Atmospheric correction methods 6S (Second Simulation of a Satellite Signal in the Solar Spectrum) was performed using a web-based 6S atmospheric code in http://6s.ltdri.org/ [6]. The recorded energy data recorded by the satellite is called reflectance. Reflectance, there are two kinds, namely reflectance TOA (Top of Atmosphere) and reflectance BoA (Bottom of Atmosphere). TOA reflectance is the reflectance captured by satellite sensors while the reflectance BoA is reflectance on the object that has been corrected atmosphere. TOA reflectance obtained by radiometric calibration and reflectance BoA obtained from atmospheric correction process. This study will compare the reflectance ToA (without atmospheric correction) and reflectance BoA. This is used to determine how its influence in the calculation of chlorophyll-a. Reflectance is used as input data on the algorithm used. In the process of atmospheric correction, every place needs a different approach algorithm model because each place has a different atmospheric character depending on the climate and character of the area. Therefore, the study of the use of atmospheric correction to a particular area needs to be studied further related to its effectiveness and models appropriate correction. There are some basic algorithms that can be used for calculation of chlorophyll-a using Landsat TM satellite imagery, such as: 1. Wouthuyzen Algorithm [7] Chl = 10,359 (TM2/TM1) – 2,355 (1) 2. Wibowo Algorithm [8] Log Chl = 2,41 (TM3/TM2) + 0,187 (2) 3. Pentury Algorithm [9] Chl = 2,3868 (TM2/TM1) -0,4671 (3) 4. Much Jisin dan Lestari Laksmi Algorithm [10] Chl = 17,912 (퐵1−퐵2) − 0,3343 (4) 퐵1+퐵2 Where TM1, TM2, TM3 is number of band on landsat TM satelit imagery

2. RESEARCH METHODOLOGY 2.1 Materials The data used in this study are: 1. Chlorophyll-a in situ On April 29, 2016. 2. Data Distribution coordinate points each sample point on 29 April, 2016. 3. Landsat 8 Path 120 Row 65 Acquisition of 29 April 2016. 4. Quickbird imagery in June 2015. 5. Data salinity dated October 15, 2013. 6. The horizontal visibility, wind speed and wind speed azimuth April 29th, 2016. 7. Pigments Concentration April 29, 2016. 8. Solar zenithal angel, Solar Azimuthal angel, angel zenithal Sensor, Sensor Azimuthal angel. 9. Data acquisition Landsat 8 April 29, 2016. The tools or equipment of research are : 1. GPS Handheld, for determine position of sample 2. Sample packing of chlorophyll-A Insitu 3. Spectrofotometer 2.2 Methods This study used a model of atmospheric correction 6SV, FLAASH and Dark of substraction (DOS). It was applied to four algorithm. The results of the four algorithms were regressed with the data in situ chlorophyll. The best model algorithm was determined from the highest value of R2 (coeficient of determination) and lowest RMSE. The algorithm selected, then statistically tested using the t test. This test to check how much influence the atmospheric correction to the accuracy

Page 58 The IJICS (International Journal of Informatics and Computer Science) Vol 5 No 1, March 2021, Page 57-63 Abdi Suknomo, Estimation Chlorophyll-a Using Landsat-8 Imagery in Shallow Water: Effect of Atmospheric and Algorithm of estimation. Area of this research located at shallow marine in districs of Wedung, Demak, Central . The Research area showed in Figure 2.

Figure 2. Research area Research process and procedure of data processing in this study is described in the flow chart shown in Figure 2

Landsat 8 field surveys images

Quickbird Geometric images Corection laboratory test No

RMSE <1 Pixel

Yes

Radiometric Calibration

DN to DN to Radiance Reflectance ToA ToA

Reflectance BoA Reflectance BoA Reflectance BoA Using DOS using 6S model using FLAASH model

Cropping

Calculation Calculation Concentration Calculation Calculation Concentration Concentration of of Chlorophyll-a using Concentration of of Chlorophyll-a using Chlorophyll-a using Wouthuyzen’s chlorophyll-a using Wibowo’s Algorithms Much Jisin and Lestari Algorithms Pentury’s algorithms Algorithms

Correlation and Chlorophyl Regression Test l-a In Situ

No Selected Algorithms

Validation

Yes

Hypothesis test

Concentration maps of Chlorophyll-a

Figure 3. Flowchart of the research

Page 59 The IJICS (International Journal of Informatics and Computer Science) Vol 5 No 1, March 2021, Page 57-63 Abdi Sukmono, Estimation Chlorophyll-a Using Landsat-8 Imagery in Shallow Water: Effect of Atmospheric and Algorithm

Description of flowchart: 1. Landsat 8 in this study was the product level 1T. It was image that has been corrected geometric, but the geometric correction is still the global nature so dependent on the accuracy DEM used, therefore needs to be done to determine the geometric accuracy geometric image accuracy. Geometric accuracy test using Quickbird imagery in June 2015 which has been rectified with the accuracy of 0.6 meters. 2. Calibration radiometric as image preprocessing was done using rescalling method, In this stage Digital Number (DN) value in the image is processed to radians and reflectance TOA. 3. The third stage is the atmospheric correction, In this stage radians and TOA reflectance processed into BoA reflectance using DOS atmospheric correction, FLAASH and 6SV. 4. The next step is cropping the image, this step is done on TOA reflectance and BoA. Cropping was done using ROI (Region of Interest) from shapefile format. 5. Then the some estimation cholorphyll-s algorithm was run in separate processing between TOA reflectance and fourth reflectance BoA,The algorithms used were Wouthuyzen in the summer (1991), Wibowo et al. (1994), Pentury (1997), Much Jisin Arief and Laksmi Lestari (2006). 6. The image chlorophyll-a and in situ then tested regression and correlation to obtain a model of the relationship and the coefficient of determination 7. The coefficient of determination serve as the main parameters in selecting the best algorithm of each model of atmospheric correction. Furthermore, the algorithm models the best of each of atmospheric correction in the analysis to get one of the best. 8. Validation of models of atmospheric correction algorithms and chlorophyll-a have been obtained from this validation. It can be known the level model accuracy. 9. Conduct a test of hypothesis to determine whether or not the influence of atmospheric correction to the calculation of chlorophyll-a. Hypothesis testing was done by the formula t test for correlated samples of the two parties 10. Making the chlorophyll-a concentration map: It estimated by running the best algorithm model and the atmospheric correction method.

3. RESULT AND DISCUSSION 3.1 Geometric accuracy test of Landsat 8 Imagery used The goal of the geometric accuracy test was determine the geometric accuracy of Landsat 8 acquisition of 29 April 2016. The level of accuracy of the image can be seen from the value of the horizontal accuracy. Geometric errors allowed is less than 1 pixel or 30 m on Landsat 8 imagery used. RMSEr = 6,376 CE90 = 1,5175 x RMSEr Horizontal accuracy = 1,5175 x 6,376 = 9,675 m In this study, RMSEr is the square difference between images coordinates Quickbird to coordinate Landsat 8. In accordance with Regulation of the Head (Perka) BIG No. 15 Year 2014 on Guidelines for Technical Accuracy Base Map the horizontal accuracy results (9.675 m) are included in the category of grade 3 Horizontal scale of 1: 25,000. 3.2 Relationship model of chlorophyll-a and reflectance TOA Table 1. Results relationship of chlorophyll-a and TOA reflectance Algorithm R2 RMSE Wouthuyzen 0,2292 0,732 Wibowo 0,4562 0,814 Pentury 0,2292 0,682 Jisin dan Lestari 0,2252 0,732 The use of TOA reflectance (without atmospheric correction) on the processing of chlorophyll-a in showed in Table 1, generally produce the low coefficient of determination, with an average of less than 0,5. It means that the resulting relationship model between chlorophyll-a and chlorophyll-a image in situ is very weak. 3.3 Relationship models of Chlorophyll-A on DOS Atmospheric Correction DOS Table 2. Results relationship of chlorophyll-a and BoA reflectance (dos method) Algorithm R2 RMSE Wouthuyzen 0,5251 1,137 Wibowo 0,5575 0,697 Pentury 0,5251 1,105 Jisin dan Lestari 0,6939 0,458

Page 60 The IJICS (International Journal of Informatics and Computer Science) Vol 5 No 1, March 2021, Page 57-63 Abdi Suknomo, Estimation Chlorophyll-a Using Landsat-8 Imagery in Shallow Water: Effect of Atmospheric and Algorithm

Based on Table 2, there was increasing coefficient determination of relation models between chlorophyll-A using atmospheric correction and Chlorophyll insitu when compared with chlorophyll-a calculation using ToA reflectance. Chlorophyll-a calculation using TOA reflectance on average worth of 0.2 while the chlorophyll-a method atmospheric correction DOS average of 0.5. The average value of RMSE with BoA reflectance value is 0.8, higher than the TOA reflectance is 0.74. 3.4 Relationship models of Chlorophyll-A on FLAASH Atmospheric Correction Table 3. Results relationship of chlorophyll-a and BoA reflectance (FLAASH method) Algorithm R2 RMSE Wouthuyzen 0,6168 0,513 Wibowo 0,5041 0,731 Pentury 0,6168 0,510 Jisin dan Lestari 0,614 0,517 Entire as shown in Table 3, it known that Pentury algorithm is an algorithm that fits well on FLAASH atmospheric correction with coefficient of determination of 0.6168 and RMSE of 0.510. 3.5 Relationship models of Chlorophyll-A on 6SV Atmospheric Correction Table 4. Results relationship of chlorophyll-a and Boa reflectance (6sv method) Algorithm R2 RMSE Wouthuyzen 0,6168 0,513 Wibowo 0,5041 0,731 Pentury 0,6168 0,510 Jisin dan Lestari 0,614 0,517 Based on Table 4, Pentury algorithm with two order polynomial relationship model is the most appropriate algorithm used in the atmospheric correction 6SV with coefficient of determination of 0.6436, the correlation coefficient of 0.802 and 0.497 RMSE values. 3.6 Analysis of The Best Relationship Model of Chlorophyll-A The best model obtained by analyzing the coefficient of determination on each algorithm, the coefficient of determination of the highest and lowest RMSE, assumed the best model. Coefficient of Determination 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Wouthuyzen Wibowo Pentury Much Jisin ToA 0.2292 0.4562 0.2292 0.2252 BoA (DOS) 0.5251 0.5575 0.5251 0.6939 BoA (FLAASH) 0.6168 0.5041 0.6168 0.614 BoA (6SV) 0.6436 0.4033 0.6436 0.6365 Figure 4. Coefficient of determination of the relationship model Based on Figure 4, the processing of chlorophyll-a using TOA reflectance (without atmospheric corection) and reflectance BOA gave different results. A bar chart shows the algorithm of chlorophyll-a by TOA reflectance indicates the coefficient of determination is lower than chlorophyll-A algorithm using reflectance BoA. Algorithms of chlorophyll a and atmospheric correction models most suitable for the District waters Wedung is Algorithm Much Jisin and Lestari with DOS atmospheric correction method with coefficient of determination of 0.6939. Then the algorithm Wouthuyzen and Pentury using atmospheric correction 6SV with coefficient of determination of 0.6436, as well as algorithms Wibowo using DOS atmospheric correction value determination coefficient of 0.6436

Page 61 The IJICS (International Journal of Informatics and Computer Science) Vol 5 No 1, March 2021, Page 57-63 Abdi Sukmono, Estimation Chlorophyll-a Using Landsat-8 Imagery in Shallow Water: Effect of Atmospheric and Algorithm

Algorithms Wouthuyzen and Pentury on TOA reflectance and BoA has the same coefficient of determination, therefore RMSE values need to be shown to find the best. RMSE value for each algorithm of chlorophyll-a is shown in the bar chart in Figure 4.

RMSE

1.2

1

0.8

0.6

0.4

0.2

0 Wouthuyzen Wibowo Pentury Much Jisin ToA 0.732 0.814 0.682 0.732 BoA (DOS) 1.137 0.697 1.105 0.458 BoA (FLAASH) 0.513 0.731 0.51 0.517 BoA (6SV) 0.498 0.803 0.497 0.502

Figure 5. RMSE of the relationship model The best atmospheric correction model for the District waters Wedung study area is FLAASH with an average coefficient of determination of 0.588 and RMSE of 0,568. FLAASH atmospheric correction models and 6SV have difference coefficients determination and RMSE were very small. Judging from each algorithm chlorophyll-a, the atmospheric correction 6SV superior method of atmospheric correction FLAASH, this is due to the low value of the coefficient of determination and the high value of RMSE on algorithms Wibowo so as to make the average coefficient of determination on atmospheric correction 6SV lower than average atmospheric correction FLAASH. 3.7 Analysis of Atmospheric Corection Effect On Estimation Chlorofphyll-A The influence of atmospheric correction in the calculation of chlorophyll-a is known to test the hypothesis bidirectional samples. X1 is the difference between the value of chlorophyll-a in the TOA reflectance (reflectance that has not corrected the atmosphere), whereas the X2 is the difference between the value of chlorophyll-a in the BoA reflectance (reflectance which has corrected the atmosphere), second variable were differenced with a value of chlorophyll-a insitu. Testing was conducted at a significance level (α) of 0,05 . IF -t table ≤ t calculation ≤ + t table then Ho is accepted and Ha rejected, meaning that there is no difference between before and after atmospheric correction. IF t calculation > + t table then Ho is rejected and Ha accepted, meaning that there is difference between before and after atmospheric correction. T test of Much Jisin algorithms and Lestari at atmospheric correction DOS a. Ho: There is no significant difference between the delta of chlorophyll-a in algorithms Much Jisin before and after correction atmospheric DOS. b. Ha: There are significant differences between the delta of chlorophyll-a in algorithms Much Jisin before and after correction atmospheric DOS. c. t> t table (2.484> 2.037), then Ho is rejected and Ha accepted. d. There are significant differences between the delta of chlorophyll-a in algorithms Much Jisin before and after correction atmospheric DOS. 3.8 Mapping of Chlorophyll-a Distribution Mapping chlorophyll-A was limited to the best algorithm is an algorithm Much Jisin and Wibowo (DOS atmospheric correction method). It was classified in five classes, 0-0.5; 0.51 to 1; 1.1-1.5; 1.51 to 2; > 2 mg / m3. Chlorophyll-a value for each region varies, so as to calculation interval chlorophyll-a concentration classes based on the value of chlorophyll- a supreme divided by the number of classes. Figure 5 shows that the chlorophyll-a concentration is dominated by 0-0.5 mg / m3. High chlorophyll-a concentration to be near the mainland and has narrowed toward the sea. As seen in the sample point number 6 and 7, the two points are on the map with a class of chlorophyll-a concentration of more than 2 mg / m3, this condition is comparable to the value of chlorophyll-a in situ.

Page 62 The IJICS (International Journal of Informatics and Computer Science) Vol 5 No 1, March 2021, Page 57-63 Abdi Suknomo, Estimation Chlorophyll-a Using Landsat-8 Imagery in Shallow Water: Effect of Atmospheric and Algorithm

Figure 6. Result of Chlorophyll-a Estimation in Study Area

4. CONCLUSION This research can be concluded that the atmospheric correction effect on the calculation of chlorophyll-a in the District waters Wedung. The use of reflectance BoA (corrected atmospheric) has the model results were better than the TOA reflectance (without atmospheric correction). Based on regression analysis, hypothesis testing, and validation test showed that the algorithm is most appropriate for the calculation of chlorophyll-a in the waters of the District Wedung algorithm Much Jisin and Lestari, the suitability is highest on the algorithm of chlorophyll-a Much Jisin and Lestari on DOS atmospheric correction. The result shown is that the usage of BoA image reflectance (atmospherically corrected) had a better model result than ToA image reflectance. This is indicated from the consecutive determination coefficient values of ToA reflectance which are 0,2292; 0,4562; 0,2292; 0,2252. Meanwhile the consecutive coefficient values of BoA reflectance by DOS correction method are 0,5251; 0,5575; 0,5251; 0,6939; by FLAASH correction method are 0,6168, 0,5041, 0,6168, 0,614; by 6SV method are 0,6436; 0,4033; 0,6436; 0,6365. The result of hypothesis and validation test is that atmospheric correction significantly influences on the calculation of chlorophyll-a in Wedung district except using Wibowo algorithm. REFERENCES

[1] M. Gilbert, A. Domin, & A. Becker. Estimation of Primary Productivity by Chlorophyll a in vivo Fluorescence in Freshwater Phytoplankton. Photosynthetica 38, 111–126 (2000). https://doi.org/10.1023/A:1026708327185 [2] V. P. Siregar, N. W. Prabowo, S. B. Agus, T. Subarno. The effect of atmospheric correction on object based image classification using SPOT-7 imagery: a case study in the Harapan and Kelapa Islands. IOP Conf. Series: Earth and Environmental Science 176 (2018) 012028. IOP Publishing. doi :10.1088/1755-1315/176/1/012028 [3] L. M. Jaelani. 2016. Basic Theory of Atmospheric Corection. https://lmjaelani.com/2016/04/slide-teori-dasar-koreksi-atmosfer/ , 2016, retrived 20 Juli 2020. [4] Ardiansyah (2015), Image Processing of Remote Sensing Data Using ENVI 5.1 and ENVI LiDAR. Selatan : Lasbig Inderaja Islim. [5] P. Danoedoro (2012), Introduction of Digital Remote Sensing. : Andi Offset. [6] E. F. Vermote, D. Tanré, J. L, Deuzé, M. Herman, & J. Morcrette, . (1997), Second Simulation of the Satellite Signal in the Solar Spectrum, 6S: An Overview, IEEE Transactions on Geoscience and Remote Sensing, Vol. 35, No. 3, p. 675-686. [7] S. Wouthuyzen dan A. R. Siahainenia. Chlorophyll-a Concentration in Kayeli, Ambon and Elpaputih Bays, Maluku Province, Indonesia Estimated from Landsat-5 TM Data and its Relation with the Meteorological Parameters. Proc. 9th (1999) JSPS Joint Sem. Fish. Sci: 29-39pp. [8] A. Wibowo, B. Sumartono, & W. Setyantini. “The Application of Satellite Data For Improvement Site Selection and Monitoring Shrimp Pond Culture” , In: Proceeding National Conference on ERS-1, Landsat, SPOT, Jakarta, Indonesia. 1993. [9] R. Pentury R, “The algorithm for predicting chlorophyll-a concentrations in Ambon Bay waters using Landsat-ETM imagery”. M.Sc Thesis. Bogor Agriculture University, Indonesia, 1997. [10] M. J. Arief & L. Laksmi. Analysis of the suitability of pond waters in Demak Regency in terms of chlorophyll-a value, water surface temperature, and suspended solids using Landsat 7 ETM + satellite imagery. Jurnal Penginderaan Jauh dan Pengolahan Data Citra Digita Vol 3 No 1 Juni 2006.

Page 63