# Spatial Downscaling of MODIS Chlorophyll-a with Genetic Programming in South Korea

^{*}

## Abstract

**:**

^{nd}-, 3

^{rd}-, and 4

^{th}-degree) to illustrate their performances for downscaling MODIS Chl-a. The obtained results indicate that GP with R

^{2}= 0.927 and RMSE = 0.1642 on the winter day and R

^{2}= 0.763 and RMSE = 0.5274 on the summer day provides higher accuracy on both winter and summer days than all the applied MPR models because the GP model can automatically produce appropriate mathematical equations without any restrictions. In addition, the GP model is the least sensitive model to the changes in the input parameters. The improved downscaling data provide better information to monitor the status of oceanic and coastal marine ecosystems that are also critical for fisheries and fishing farming.

## 1. Introduction

^{nd}), third-order (3

^{rd}), and fourth-order (4

^{th}) polynomials. The developed models were utilized for Chl-a downscaling over the western coast of South Korea.

## 2. Study Area

^{2}) (Figure 1). The Yellow Sea is a shallow marine ecosystem with the average and maximum water depths of 44 m and 103 m, respectively [38]. There is clear seasonality in sea surface temperature (SST) over the Yellow Sea, where January is the coldest month with an average SST of 4–7 °C and July is the warmest, with an average SST of 26–27 °C [39]. There are a total of 339 fish species in the Korean West Sea [40]. Over the past few years, some fish species, such as small yellow croaker, hairtail, large yellow croaker, and flatfish have exhibited continuous declines due to overharvesting, degraded marine ecosystem quality, and several unknown factors [41].

## 3. Materials and Methods

_{4k}) and S-2A at 10 m (defined as X

_{10}), were acquired, and S-2A data were resampled to 4 km MODIS resolution (denoted as X

_{4k}); (2) the most important S-2A band combinations (X

_{4k}) were chosen by utilizing the support vector machine recursive feature elimination (SVM-RFE) method; (3) MODIS Chl-a downscaling from 4 km to 10 m was performed by regressing X

_{4k}to Y

_{4k}, calculating the residual at 4 km (${\mathsf{\epsilon}}_{4\mathrm{k}}$), and adding the interpolated residual (${\mathsf{\epsilon}}_{10}$) to the estimated fine-resolution Chl-a (${\widehat{\mathrm{Y}}}_{10}$); (4) the obtained downscaled maps were compared with visual comparison, validated with in situ data, and all the applied methods were assessed using sensitivity analysis. A complete explanation of the aforementioned steps is described in Section 3.1, Section 3.2, Section 3.3, Section 3.4.

#### 3.1. Data Description

#### 3.2. Preprocessing

_{10}). Since the study region had a remarkable number of islands (Figure 1), the reflectance caused by islands could decrease the accuracy of the applied models. Therefore, the third step of preprocessing aimed to separate water from land pixels to improve the downscaling results. The most commonly used indices, namely, the normalized difference vegetation index (NDVI) and normalized difference water index (NDWI), have already been employed for land-water separation purposes [54,55,56,57,58]. In the current study, land-water separation was performed by utilizing the NDWI index. The formula of this index is as follows:

_{10}) to 4 km resolution of MODIS Chl-a (X

_{4k}) using the nearest neighbor method [59]. Then, MODIS Chl-a and S-2A images at 4 km (Y

_{4k}and X

_{4k}) were used as dependent and independent variables, respectively, to develop GP and MPR models. Additionally, the developed models were applied to S-2A images at 10 m (X

_{10}) to estimate downscaled Chl-a at 10 m spatial resolution.

_{4k}). This method is a widely used technique for feature selection that has been utilized in a wide variety of remote sensing research studies [60,61,62]. To perform feature selection, multiple mathematical operations, including multiplication, addition, subtraction, rationing, averaging, and square transformation, were used to calculate 597 various combinations of the resampled S-2A bands at 4 km, and the computed combinations served as the input for the feature selection method. Then, SVM-RFE was trained on the calculated dataset using the MODIS Chl-a concentration (Y

_{4k}) and S-2A band combinations (X

_{4k}) as the predictand and predictor variables, respectively. As a result, relevant combinations for the downscaling procedure were chosen.

#### 3.3. Downscaling

_{4k}) and 10 m (X

_{10}), $\widehat{\mathrm{Y}}$ is the estimated MODIS Chl-a at 4 km (${\widehat{\mathrm{Y}}}_{4\mathrm{k}}$) and 10 m (${\widehat{\mathrm{Y}}}_{10}$), and f represents a nonlinear regression function developed by GP and MPR techniques.

_{4k}and X

_{4k}was established using all the applied methods as 2

^{nd}-degree MPR, 3

^{rd}-degree MPR, 4

^{th}-degree MPR, and GP. From the 636 pixels of the independent (X

_{4k}) and dependent (Y

_{4k}) variables, 445 pixels were used for training, and the remaining 191 pixels were reserved for the validation of the models. Standardization was a prerequisite step before training all the applied methods to ensure that all variables remained on the same scale. Therefore, the standardization of all independent variables (X

_{4k}) in the training and validation set was performed by subtracting the mean and dividing by the standard deviation of the training set. This preprocessing sped up the convergence and allowed efficient training of the network. Then, the estimated Chl-a at 4 km (${\widehat{\mathrm{Y}}}_{4\mathrm{k}}$) was subtracted from the original MODIS Chl-a at 4 km (Y

_{4k}) as follows:

#### 3.4. Sensitivity Analysis

_{i,n}is the elementary effects (EEs) assessed for the i-th input variable using the n-th sample point. EEs are employed to specify noninfluential inputs for a computationally costly mathematical model or for a model with a great number of input parameters, where the costs of evaluating other SA methods such as variance-based methods are not reasonably priced.

#### 3.5. Mathematical Background

#### 3.5.1. Multiple Polynomial Regression

^{th}order polynomial model with one variable Equation (6) is the general form of the polynomial model that indicates the nonlinear relationship between one predictand and one predictor variable.

^{nd}), third-order (3

^{rd}), and fourth-order (4

^{th}), were trained and compared.

#### 3.5.2. Genetic Programming

## 4. Results

#### 4.1. Preprocessing

^{2}= 0.996 for winter (panel (a)) and R

^{2}= 0.939 for summer (panel (b)). Therefore, it can be concluded that the grid-fill method exhibits good performance for the reconstruction of MODIS Chl-a values. Note that the reason for the better performance of the method on the winter day than on the summer day might be related to the presence of more missing values in the selected scene of the summer day that occurs due to high cloudiness on the summer day.

#### 4.2. Downscaling Results

^{nd}, 3

^{rd}, and 4

^{th}) and GP were trained with the four determined band combinations as predictors and MODIS Chl-a as a predictand at low-resolution (4 km). Then, a residual correction was performed utilizing the described methodology in the downscaling section (Section 3.3), and Equations (3)–(4) to produce high-resolution Chl-a maps (Y

_{10}). To assess the accuracy of the downscaling technique, the final downscaled maps (i.e., Y

_{10}) were validated with the original MODIS Chl-a maps at a pixel size of 4 km. For this purpose, sample values were extracted from the downscaled maps (Y

_{10}) within a 3 × 3 window around the center of each MODIS Chl-a pixel, and the mean of each sample was calculated and compared with the original MODIS Chl-a (Figure 5 and Table 3).

^{nd}-degree MPR. According to the performance indices, the accuracy of the models was ranked as GP > 2

^{nd}-degree MPR > 3

^{rd}-degree MPR > 4

^{th}-degree MPR for the winter day. For the summer day, the rank was GP > 3

^{rd}-degree MPR > 2

^{nd}-degree MPR > 4

^{th}-degree MPR. Overall, the GP exhibited the best performance for the winter and summer days.

^{2}= 0.927 and RMSE = 0.1642 on the winter day, compared to the MPR models (2

^{nd}-, 3

^{rd}-, and 4

^{th}-degree). The same result can be seen on the summer day (the best performance in the GP model), as shown in the bottom panels of Figure 5. Furthermore, it can be seen that all the applied models estimate Chl-a better at lower concentrations than at higher concentrations, particularly in the range of 1.5–3.5 mg m

^{−3}, as presented in Figure 5.

#### 4.3. Model Evaluation

#### 4.3.1. Visual Comparison

^{nd}-degree MPR. From Figure 6, the major trend of Chl-a is fairly modeled with GP and 2

^{nd}-degree MPR but specific and substantially high values are not captured in both models such as the values in the near coastal area. However, the residual model can additionally capture this high variability and produce a reliable estimate in the final stage, as seen in the last column of Figure 6. According to the results illustrated in Table 3, and the panels in the last column of Figure 6, 4

^{th}-degree MPR cannot estimate the Chl-a values at 10 m resolution as accurately as the other models, especially in coastal areas. This phenomenon indicates that the GP and 2

^{nd}-degree MPR can capture the most variation in high Chl-a concentrations at 4 km resolution.

#### 4.3.2. In Situ Validation

^{2}and RMSE are presented in Figure 9 for the winter day (left panel) and the summer day (right panel). The computed p-values (Figure 9) shows that R

^{2}of all models is statistically significant (p-values lower than 0.05). In general, the R

^{2}of the GP model was higher than that of the other models for the winter and summer days, which were 0.59 and 0.47, respectively, as shown in Figure 9. Additionally, the RMSE of the GP model was smaller than that of the other models for the winter and summer days, at 0.766 and 0.483, respectively.

#### 4.3.3. Sensitivity Analysis

^{nd}-degree MPR model is the most sensitive model (${\mu}_{}^{*}$ values ranged from 0.09 to 1.24 for the winter and 0.94 to 2.42 for the summer day), while the GP model is the least sensitive model (${\mu}_{}^{*}$ = 1.45 and 1.4 for B2/B3 band combination in winter and summer days, respectively) to the changes in the predictor variables. In addition, the sensitivity of the MPR models increases on the summer day compared to that of the winter day, while the GP model shows a slight decrease to the changes of the B2/B3 band combination on the summer day compared to that of the winter day.

^{nd}-degree MPR show comparable accuracy (Figure 5 and Table 3) on both winter and summer days, the GP model with the least sensitivity to the changes in the input parameters is more effective than the 2

^{nd}-degree MPR for downscaling Chl-a. Additionally, SA results show that the accuracy of all models is strongly related to the accuracy of the remote sensing data; this further confirms the need for atmospheric correction as an essential task in the downscaling procedure [71].

## 5. Discussion and Conclusions

^{nd}, 3

^{rd}, and 4

^{th}) and GP model to downscale MODIS Chl-a; (iv) assessment of the results with original MODIS Chl-a maps at a pixel size of 4 km and in situ measurements.

^{nd}-degree MPR, but specific and substantially high values are not captured with both models such as the values in near coastal areas. This drawback of the models can be solved with the residual correction that makes it an essential procedure to improve the accuracy of the spatial downscaling model.

## Author Contributions

## Funding

## Acknowledgments

## Conflicts of Interest

## References

- Bierman, P.; Lewis, M.; Ostendorf, B.; Tanner, J. A review of methods for analysing spatial and temporal patterns in coastal water quality. Ecol. Indic.
**2011**, 11, 103–114. [Google Scholar] [CrossRef] - Blondeau-Patissier, D.; Gower, J.F.; Dekker, A.; Phinn, S.; Brando, V. A review of ocean color remote sensing methods and statistical techniques for the detection, mapping and analysis of phytoplankton blooms in coastal and open oceans. Prog. Oceanogr.
**2014**, 123, 123–144. [Google Scholar] [CrossRef][Green Version] - Cullen, J.J. The deep chlorophyll maximum: Comparing vertical profiles of chlorophyll a. Can. J. Fish. Aquat. Sci.
**1982**, 39, 791–803. [Google Scholar] [CrossRef] - Guan, X.; Li, J.; Booty, W.G. Monitoring lake simcoe water clarity using Landsat-5 TM images. Water Resour. Manag.
**2011**, 25, 2015–2033. [Google Scholar] [CrossRef] - Carpenter, S.R.; Caraco, N.F.; Correll, D.L.; Howarth, R.W.; Sharpley, A.N.; Smith, V.H. Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecol. Appl.
**1998**, 8, 559–568. [Google Scholar] [CrossRef] - Smith, V.H. Cultural eutrophication of inland, estuarine, and coastal waters. In Successes, Limitations, and Frontiers in Ecosystem Science; Springer: New York, NY, USA, 1998; pp. 7–49. [Google Scholar]
- Kar, D. Epizootic Ulcerative Fish Disease Syndrome; Academic Press: Cambridge, MA, USA, 2015. [Google Scholar]
- Bacher, C.; Grant, J.; Hawkins, A.J.; Fang, J.; Zhu, M.; Besnard, M. Modelling the effect of food depletion on scallop growth in Sungo Bay (China). Aquat. Living Resour.
**2003**, 16, 10–24. [Google Scholar] [CrossRef] - Huot, Y.; Babín, M.; Bruyant, F.; Grob, C.; Twardowski, M.S.; Claustre, H. Does chlorophyll a provide the best index of phytoplankton biomass for primary productivity studies? Biogeosci. Discuss.
**2007**, 4, 707–745. [Google Scholar] [CrossRef][Green Version] - Goetz, S.J.; Gardiner, E.P.; Viers, J.H. Monitoring freshwater, estuarine and near-shore benthic ecosystems with multi-sensor remote sensing: An introduction to the special issue. Remote Sens. Environ.
**2008**, 112, 3993–3995. [Google Scholar] [CrossRef] - Kutser, T. Passive optical remote sensing of cyanobacteria and other intense phytoplankton blooms in coastal and inland waters. Int. J. Remote Sens.
**2009**, 30, 4401–4425. [Google Scholar] [CrossRef] - Schaeffer, B.A.; Schaeffer, K.G.; Keith, D.; Lunetta, R.S.; Conmy, R.; Gould, R.W. Barriers to adopting satellite remote sensing for water quality management. Int. J. Remote Sens.
**2013**, 34, 7534–7544. [Google Scholar] [CrossRef] - Merico, A.; Brown, C.W.; Groom, S.B.; Miller, P.; Tyrrell, T. Analysis of satellite imagery forEmiliania huxleyiblooms in the Bering Sea before 1997. Geophys. Res. Lett.
**2003**, 30, 30. [Google Scholar] [CrossRef] - Gower, J. Productivity and plankton blooms observed with Seawifs and in-situ sensors. In Proceedings of the IGARSS 2001, Scanning the Present and Resolving the Future, Proceedings, IEEE 2001 International Geoscience and Remote Sensing Symposium (Cat. No.01CH37217), Sydney, NSW, Australia, 9–13 July 2001; pp. 2181–2183. [Google Scholar]
- Lavender, S.; Groom, S.B. The detection and mapping of algal blooms from space. Int. J. Remote Sens.
**2001**, 22, 197–201. [Google Scholar] [CrossRef] - Siegel, D.; Behrenfeld, M.J.; Maritorena, S.; McClain, C.; Antoine, D.; Bailey, S.W.; Bontempi, P.; Boss, E.; Dierssen, H.; Doney, S.C.; et al. Regional to global assessments of phytoplankton dynamics from the SeaWiFS mission. Remote Sens. Environ.
**2013**, 135, 77–91. [Google Scholar] [CrossRef][Green Version] - Hu, C.; Muller-Karger, F.E.; Taylor, C.J.; Carder, K.L.; Kelble, C.; Johns, E.; Heil, C.A. Red tide detection and tracing using MODIS fluorescence data: A regional example in SW Florida coastal waters. Remote Sens. Environ.
**2005**, 97, 311–321. [Google Scholar] [CrossRef] - Gower, J.; King, S.; Yan, W.; Borstad, G.; Brown, L. Use of the 709 nm band of MERIS to detect intense plankton blooms and other conditions in coastal waters. In Proceedings of the MERIS User Workshop, Frascati, Italy, 10–13 November 2003. [Google Scholar]
- Hu, C. A novel ocean color index to detect floating algae in the global oceans. Remote Sens. Environ.
**2009**, 113, 2118–2129. [Google Scholar] [CrossRef] - Claustre, H.; Babin, M.; Merien, D.; Ras, J.; Prieur, L.; Dallot, S.; Prášil, O.; Dousova, H.; Moutin, T. Toward a taxon-specific parameterization of bio-optical models of primary production: A case study in the North Atlantic. J. Geophys. Res. Space Phys.
**2005**, 110, 110. [Google Scholar] [CrossRef] - Sathyendranath, S.; Watts, L.; Devred, E.; Platt, T.; Caverhill, C.; Maass, H. Discrimination of diatoms from other phytoplankton using ocean-colour data. Mar. Ecol. Prog. Ser.
**2004**, 272, 59–68. [Google Scholar] [CrossRef] - Shang, S.; Dong, Q.; Lee, Z.; Li, Y.; Xie, Y.; Behrenfeld, M. MODIS observed phytoplankton dynamics in the Taiwan Strait: An absorption-based analysis. Biogeosciences
**2011**, 8, 841–850. [Google Scholar] [CrossRef][Green Version] - Lee, Z.; Carder, K.L.; Arnone, R.A. Deriving inherent optical properties from water color: A multiband quasi-analytical algorithm for optically deep waters. Appl. Opt.
**2002**, 41, 5755–5772. [Google Scholar] [CrossRef] - Addesso, P.; Longo, M.; Maltese, A.; Restaino, R.; Vivone, G. Batch methods for resolution enhancement of TIR image sequences. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.
**2015**, 8, 3372–3385. [Google Scholar] [CrossRef] - Bechtel, B.; Zakšek, K.; Hoshyaripour, G. Downscaling land surface temperature in an urban area: A case study for Hamburg, Germany. Remote Sens.
**2012**, 4, 3184–3200. [Google Scholar] [CrossRef][Green Version] - Liu, D.; Pu, R. Downscaling thermal infrared radiance for subpixel land surface temperature retrieval. Sensors
**2008**, 8, 2695–2706. [Google Scholar] [CrossRef] [PubMed][Green Version] - Chen, C.; Zhao, S.; Duan, Z.; Qin, Z. An improved spatial downscaling procedure for TRMM 3B43 precipitation product using geographically weighted regression. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.
**2015**, 8, 4592–4604. [Google Scholar] [CrossRef] - Fang, J.; Du, J.; Xu, W.; Shi, P.; Li, M.; Ming, X. Spatial downscaling of TRMM precipitation data based on the orographical effect and meteorological conditions in a mountainous area. Adv. Water Resour.
**2013**, 61, 42–50. [Google Scholar] [CrossRef] - Immerzeel, W.; Rutten, M.; Droogers, P. Spatial downscaling of TRMM precipitation using vegetative response on the Iberian Peninsula. Remote Sens. Environ.
**2009**, 113, 362–370. [Google Scholar] [CrossRef] - Zhang, T.; Li, B.; Yuan, Y.; Gao, X.; Sun, Q.; Xu, L.; Jiang, Y. Spatial downscaling of TRMM precipitation data considering the impacts of macro-geographical factors and local elevation in the Three-River Headwaters Region. Remote Sens. Environ.
**2018**, 215, 109–127. [Google Scholar] [CrossRef] - Kaheil, Y.H.; Gill, M.K.; Bastidas, L.A.; Rosero, E.; McKee, M. Downscaling and assimilation of surface soil moisture using ground truth measurements. IEEE Trans. Geosci. Remote Sens.
**2008**, 46, 1375–1384. [Google Scholar] [CrossRef] - Shi, J.; Jiang, L.; Zhang, L.; Chen, K.; Wigneron, J.-P.; Chanzy, A.; Jackson, T. Physically based estimation of bare-surface soil moisture with the passive radiometers. IEEE Trans. Geosci. Remote Sens.
**2006**, 44, 3145–3153. [Google Scholar] [CrossRef] - Fu, Y.; Xu, S.; Zhang, C.; Sun, Y. Spatial downscaling of MODIS Chlorophyll-a using Landsat 8 images for complex coastal water monitoring. Estuar. Coast. Shelf Sci.
**2018**, 209, 149–159. [Google Scholar] [CrossRef] - Gao, F.; Kustas, W.; Anderson, M.C. A data mining approach for sharpening thermal satellite imagery over land. Remote Sens.
**2012**, 4, 3287–3319. [Google Scholar] [CrossRef][Green Version] - Ghosh, A.; Joshi, P. Hyperspectral imagery for disaggregation of land surface temperature with selected regression algorithms over different land use land cover scenes. ISPRS J. Photogramm. Remote Sens.
**2014**, 96, 76–93. [Google Scholar] [CrossRef] - Hutengs, C.; Vohland, M. Downscaling land surface temperatures at regional scales with random forest regression. Remote Sens. Environ.
**2016**, 178, 127–141. [Google Scholar] [CrossRef] - Mehr, A.D.; Nourani, V.; Kahya, E.; Hrnjica, B.; Sattar, A.M.A.; Yaseen, Z.M. Genetic programming in water resources engineering: A state-of-the-art review. J. Hydrol.
**2018**, 566, 643–667. [Google Scholar] [CrossRef] - Koh, C.-H.; Khim, J.S. The Korean tidal flat of the Yellow Sea: Physical setting, ecosystem and management. Ocean Coast. Manag.
**2014**, 102, 398–414. [Google Scholar] [CrossRef] - Park, K.-A.; Lee, E.-Y.; Chang, E.; Hong, S. Spatial and temporal variability of sea surface temperature and warming trends in the Yellow Sea. J. Mar. Syst.
**2015**, 143, 24–38. [Google Scholar] [CrossRef] - Zhang, C.I.; Lim, J.-H.; Kwon, Y.; Kang, H.J.; Kim, H.; Seo, Y.I. The current status of west sea fisheries resources and utilization in the context of fishery management of Korea. Ocean Coast. Manag.
**2014**, 102, 493–505. [Google Scholar] [CrossRef] - Zhang, C.; Kim, S. Living marine resources of the Yellow Sea ecosystem in Korean waters: Status and perspectives. In Large Marine Ecosystems of the Pacific Rim; Wiley, Blackwell Science: Cambridge, MA, USA, 1999; pp. 163–178. [Google Scholar]
- Ye, H.; Li, J.; Li, T.; Shen, Q.; Zhu, J.; Wang, X.; Zhang, F.; Zhang, J.; Zhang, B. Spectral classification of the Yellow Sea and implications for coastal ocean color remote sensing. Remote Sens.
**2016**, 8, 321. [Google Scholar] [CrossRef][Green Version] - Cho, Y.-K.; Kim, M.-O.; Kim, B.-C. Sea fog around the Korean Peninsula. J. Appl. Meteorol.
**2000**, 39, 2473–2479. [Google Scholar] [CrossRef] - Letelier, R.M. An analysis of chlorophyll fluorescence algorithms for the moderate resolution imaging spectrometer (MODIS). Remote Sens. Environ.
**1996**, 58, 215–223. [Google Scholar] [CrossRef] - Sarthyendranath, S. Remote Sensing of Ocean Colour in Coastal, and Other Optically-Complex, Waters; International Ocean Colour Coordinating Group (IOCCG): Dartmouth, NS, Canada, 2000. [Google Scholar]
- Hu, C.; Lee, Z.; Franz, B. Chlorophyll aalgorithms for oligotrophic oceans: A novel approach based on three-band reflectance difference. J. Geophys. Res. Ocean.
**2012**, 117. [Google Scholar] [CrossRef][Green Version] - Pahlevan, N.; Schott, J.R. Leveraging EO-1 to evaluate capability of new generation of landsat sensors for coastal/inland water studies. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.
**2013**, 6, 360–374. [Google Scholar] [CrossRef] - Pahlevan, N.; Lee, Z.; Wei, J.; Schaaf, C.; Schott, J.R.; Berk, A. On-orbit radiometric characterization of OLI (Landsat-8) for applications in aquatic remote sensing. Remote Sens. Environ.
**2014**, 154, 272–284. [Google Scholar] [CrossRef] - Gower, J.; King, S.; Borstad, G.; Brown, L. Detection of intense plankton blooms using the 709 nm band of the MERIS imaging spectrometer. Int. J. Remote Sens.
**2005**, 26, 2005–2012. [Google Scholar] [CrossRef] - Gower, J.; King, S.; Borstad, G.; Brown, L. The importance of a band at 709 nm for interpreting water-leaving spectral radiance. Can. J. Remote Sens.
**2008**, 34, 287–295. [Google Scholar] - Moses, W.J.; Gitelson, A.; Berdnikov, S.; Povazhnyy, V. Satellite estimation of chlorophyll-a concentration using the red and NIR bands of MERIS—The Azov Sea case study. IEEE Geosci. Remote Sens. Lett.
**2009**, 6, 845–849. [Google Scholar] [CrossRef] - Pahlevan, N.; Chittimalli, S.K.; Balasubramanian, S.V.; Vellucci, V. Sentinel-2/Landsat-8 product consistency and implications for monitoring aquatic systems. Remote Sens. Environ.
**2019**, 220, 19–29. [Google Scholar] [CrossRef] - Dawson, A. (2018, April 23). ajdawson/gridfill: Version 1.0.1 (Version v1.0.1). Zenodo.
- Gao, H.; Birkett, C.; Lettenmaier, D.P. Global monitoring of large reservoir storage from satellite remote sensing. Water Resour. Res.
**2012**, 48, 48. [Google Scholar] [CrossRef][Green Version] - McFeeters, S.K. The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. Int. J. Remote Sens.
**1996**, 17, 1425–1432. [Google Scholar] [CrossRef] - Mohebzadeh, H.; Fallah, M. Quantitative analysis of water balance components in Lake Urmia, Iran using remote sensing technology. Remote Sens. Appl. Soc. Environ.
**2019**, 13, 389–400. [Google Scholar] [CrossRef] - Sima, S.; Ahmadalipour, A.; Tajrishy, M. Mapping surface temperature in a hyper-saline lake and investigating the effect of temperature distribution on the lake evaporation. Remote Sens. Environ.
**2013**, 136, 374–385. [Google Scholar] [CrossRef] - Mohebzadeh, H. Extracting A-L relationship for Urmia Lake, Iran using MODIS NDVI/NDWI indices. J. Hydrogeol. Hydrol. Eng.
**2018**, 7, 1. [Google Scholar] [CrossRef] - Lee, T.; Singh, V.P. Statistical Downscaling for Hydrological and Environmental Applications; CRC Press: Boca Raton, FL, USA, 2018; Volume 1, p. 165. [Google Scholar]
- Tuia, D.; Pacifici, F.; Kanevski, M.; Emery, W.J. Classification of very high spatial resolution imagery using mathematical morphology and support vector machines. IEEE Trans. Geosci. Remote Sens.
**2009**, 47, 3866–3879. [Google Scholar] [CrossRef] - Zheng, Z.; Zeng, Y.; Li, S.; Huang, W. A new burn severity index based on land surface temperature and enhanced vegetation index. Int. J. Appl. Earth Obs. Geoinf.
**2016**, 45, 84–94. [Google Scholar] [CrossRef] - Ebrahimy, H.; Azadbakht, M. Downscaling MODIS land surface temperature over a heterogeneous area: An investigation of machine learning techniques, feature selection, and impacts of mixed pixels. Comput. Geosci.
**2019**, 124, 93–102. [Google Scholar] [CrossRef] - Rossi, R.E.; Dungan, J.L.; Beck, L.R. Kriging in the shadows: Geostatistical interpolation for remote sensing. Remote Sens. Environ.
**1994**, 49, 32–40. [Google Scholar] [CrossRef] - Karl, J. Spatial predictions of cover attributes of rangeland ecosystems using regression kriging and remote sensing. Rangel. Ecol. Manag.
**2010**, 63, 335–349. [Google Scholar] [CrossRef] - Petropoulos, G.; Srivastava, P.K. Sensitivity Analysis in Earth Observation Modelling; Elsevier: Amsterdam, The Netherlands, 2016. [Google Scholar]
- Morris, M.D. Factorial sampling plans for preliminary computational experiments. Technometrics
**1991**, 33, 161–174. [Google Scholar] [CrossRef] - Rao, C.R.; Toutenburg, H.; Shalabh, H.C.; Schomaker, M. Linear models and generalizations. In Least Squares and Alternatives, 3rd ed.; Springer: Berlin/Heidelberg, Germany; New York, NY, USA, 2008. [Google Scholar]
- Koza, J.R. Genetic Programming: On the Programming of Computers by Means of Natural Selection; MIT Press: Cambridge, MA, USA; London, UK, 1992; Volume 1. [Google Scholar]
- Stephens, T. Gplearn Model, Genetic Programming; Copyright, 2015. [Google Scholar]
- Geem, Z.W.; Kim, J.H.; Loganathan, G. A new heuristic optimization algorithm: Harmony search. Simulation
**2001**, 76, 60–68. [Google Scholar] [CrossRef] - Nazeer, M.; Nichol, J. Development and application of a remote sensing-based Chlorophyll-a concentration prediction model for complex coastal waters of Hong Kong. J. Hydrol.
**2016**, 532, 80–89. [Google Scholar] [CrossRef]

**Figure 2.**Downscaling workflow. Note that the goal of the present research is to downscale Moderate Resolution Imaging Spectroradiometer (MODIS) Chl-a from coarse resolution (4 km) to high resolution (10 m).

**Figure 4.**1:1 scatter plots of the original MODIS Chl-a vs. reconstructed Chl-a by the grid-fill method on (

**a**) the winter day (2016.2.2) and (

**b**) the summer day (2016.8.4).

**Figure 5.**1:1 scatter plots of the original MODIS Chl-a at 4 km pixel size vs. downscaled MODIS Chl-a at 10 m pixel size; (

**a**) the winter day (2016.2.2): (

**a-1**) 2

^{nd}-degree MPR, (

**a-2**) 3

^{rd}-degree MPR, (

**a-3**) 4

^{th}-degree MPR, (

**a-4**) GP, and (

**b**) the summer day (2016.8.4): (

**b-1**) 2

^{nd}-degree MPR, (

**b-2**) 3

^{rd}-degree MPR, (

**b-3**) 4

^{th}-degree MPR, and (

**b-4**) GP.

**Figure 6.**Detailed maps showing the downscaling steps for all models on the winter day (2016.2.2); (

**a**) 2

^{nd}-degree MPR, (

**b**) 3

^{rd}-degree MPR, (

**c**) 4

^{th}-degree MPR, and (

**d**) GP.

**Figure 7.**Detailed maps showing the downscaling steps for all models on the summer day (2016.8.4); (

**a**) 2

^{nd}-degree MPR, (

**b**) 3

^{rd}-degree MPR, (

**c**) 4

^{th}-degree MPR, and (

**d**) GP.

**Figure 8.**Histogram and fitted normal distribution of MODIS Chl-a for (

**a**) the winter day (2016.2.2) and (

**b**) the summer day (2016.8.4). $\mathsf{\delta}$ and $\mathsf{\theta}$ are the coefficient of variation (COV) and skewness coefficient, respectively. Note that the COV and skewness coefficient values for a normal distribution are zero. The skewness coefficient indicates that the distribution of the summer day data presents more non-normality than that of the winter day.

**Figure 9.**1:1 scatter plots of downscaled MODIS Chl-a and in situ data for all methods using two measured data at 0–15 cm depth; (

**a**) the winter day (2016.2.2) and (

**b**) the summer day (2016.8.4).

**Figure 10.**Sensitivity measure index (${\mu}_{}^{*}$) for the winter day (2016.2.2); (

**a**) 2

^{nd}-degree MPR, (

**b**) 3

^{rd}-degree MPR, (

**c**) 4

^{th}-degree MPR, and (

**d**) GP. Note that the higher values of ${\mu}_{}^{*}$ for a given parameter indicate higher sensitivity of the model to the changes of the parameter.

**Figure 11.**Sensitivity measure index (${\mu}_{}^{*}$) for the summer day (2016.8.4); (

**a**) 2

^{nd}-degree MPR, (

**b**) 3

^{rd}-degree MPR, (

**c**) 4

^{th}-degree MPR, and (

**d**) GP. Note that the higher values of ${\mu}_{}^{*}$ for a given parameter indicate higher sensitivity of the model to the changes of the parameter.

Band Name | Central Wavelength (µm) | Spatial Resolution (m) |
---|---|---|

B1-Coastal aerosol | 0.443 | 60 |

B2-Blue | 0.490 | 10 |

B3-Green | 0.560 | 10 |

B4-Red | 0.665 | 10 |

B5-Vegetation red edge | 0.705 | 20 |

B6-Vegetation red edge | 0.740 | 20 |

B7-Vegetation red edge | 0.783 | 20 |

B8-NIR | 0.842 | 10 |

B8A-Narrow NIR | 0.865 | 20 |

B9-Water vapor | 0.940 | 60 |

B10-SWIR-Cirrus | 1.375 | 60 |

B11-SWIR | 1.610 | 20 |

B12-SWIR | 2.190 | 20 |

MODIS | Sentinel-2A MSI | Chl-a Measurements |
---|---|---|

2016.2.2 | 2016.2.2 | 2016.2.1 |

2016.8.4 | 2016.8.4 | 2016.8.1 |

Date | Performance Statistics | Model | |||
---|---|---|---|---|---|

2^{nd} Degree MPR | 3^{rd} Degree MPR | 4^{th} Degree MPR | GP | ||

2016.2.2 | MAE | 0.108 | 0.144 | 0.150 | 0.108 |

(Winter) | MBE | −0.014 | −0.012 | −0.020 | 0.011 |

RMSE | 0.168 | 0.207 | 0.219 | 0.164 | |

R^{2} | 0.922 | 0.886 | 0.872 | 0.927 | |

2016.8.4 | MAE | 0.360 | 0.383 | 0.409 | 0.341 |

(Summer) | MBE | −0.056 | −0.044 | −0.068 | −0.030 |

RMSE | 0.562 | 0.542 | 0.595 | 0.527 | |

R^{2} | 0.732 | 0.753 | 0.704 | 0.763 |

^{2}is the coefficient of determination.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Mohebzadeh, H.; Yeom, J.; Lee, T.
Spatial Downscaling of MODIS Chlorophyll-a with Genetic Programming in South Korea. *Remote Sens.* **2020**, *12*, 1412.
https://doi.org/10.3390/rs12091412

**AMA Style**

Mohebzadeh H, Yeom J, Lee T.
Spatial Downscaling of MODIS Chlorophyll-a with Genetic Programming in South Korea. *Remote Sensing*. 2020; 12(9):1412.
https://doi.org/10.3390/rs12091412

**Chicago/Turabian Style**

Mohebzadeh, Hamid, Junho Yeom, and Taesam Lee.
2020. "Spatial Downscaling of MODIS Chlorophyll-a with Genetic Programming in South Korea" *Remote Sensing* 12, no. 9: 1412.
https://doi.org/10.3390/rs12091412