A Remotely Sensed Assessment of Surface Ecological Change over the Gomishan Wetland, Iran

Due to the excessive use of natural resources in the contemporary world, the importance of ecological and environmental condition modeling has increased. Wetlands and cities represent the natural and artificial strategic areas that affect ecosystem conditions. Changes in the ecological conditions of these areas have a great impact on the conditions of the global ecosystem. Therefore, modeling spatiotemporal variations of the ecological conditions in these areas is critical. This study was aimed at comparing degrees of variation among surface ecological conditions due to natural and unnatural factors. Consequently, the surface ecological conditions of Gomishan city and Gomishan wetland in Iran were modeled for a period of 30 years, and the spatiotemporal variations were evaluated and compared with each other. To this end, 20 Landsat 5, 7, and 8, and 432 Moderate Resolution Imaging Spectroradiometer (MODIS), monthly land surface temperature (LST) (MOD11C3) and normalized difference vegetation index (NDVI) (MOD13C3) products were utilized. The surface ecological conditions were modeled according to the Remote Sensing-based Ecological Index (RSEI), and the spatiotemporal variation of the RSEI values in the study area (Gomishan city, Gomishan wetland) were evaluated and compared with each other. According to MODIS products, the mean of the LST and NDVI variance values for the study area (Gomishan city, Gomishan wetland) were obtained to be 6.5 °C (2.1, 12.1) and 0.009 (0.005, 0.013), respectively. The highest LST and NDVI temporal variations were found for Gomishan wetland near the Caspian Sea. According to Landsat images, Gomishan wetland and Gomishan city have the highest and lowest temporal variations in surface biophysical characteristics, respectively. The mean RSEI for the study area (Gomishan city, Gomishan wetland) was 0.43 (0.65, 0.29), respectively. Additionally, the mean Coefficient of Variation (CV) of RSEI for the study area (Gomishan city, Gomishan wetland) was 0.10 (0.88, 0.51), respectively. The surface ecological conditions of Gomishan city were worse than those of the Gomishan wetland at all dates. Temporal variations in the surface ecological conditions of Gomishan wetland were greater than those of the study area and Gomishan city. These results can provide useful and effective information for environmental planning and decision-making to improve ecological conditions, protect the environment, and support sustainable ecosystem development.


Data
In this study, MODIS products and Landsat images were used (Table 1). Landsat 5,7, and 8 images, including Level 1 Terrain (Corrected) (L1T), were used with WRS_PATH = 164 and WRS_ROW = 035. These imagery are available on the https://www.usgs.gov/. Landsat imagery was selected in such a way that (1) the maximum cloud cover of the study area at satellite overpass time was a maximum of 10% and (2) their dates were close enough to each other so that seasonal changes did not affect the results of various analyses. The range of Day of Year (DOY) values of used Landsat images was less than 60 days. These images have the appropriate resolution to highlight the trend and surface coverage features.
In addition to Landstat data, monthly LST (MOD11C3) and NDVI (MOD13C2) MODIS/Terra The eastern and northeastern regions of the wetland are affected by arid and semi-arid climates due to the vast expanse of flat lands, lack of heights, distance from the southern forests, and proximity to the Turkmen deserts. Moving from the south to the border, the intensity of this type of climate increases. The study area geographical location is shown in Figure 1.

Data
In this study, MODIS products and Landsat images were used (Table 1). Landsat 5,7, and 8 images, including Level 1 Terrain (Corrected) (L1T), were used with WRS_PATH = 164 and WRS_ROW = 035. These imagery are available on the https://www.usgs.gov/. Landsat imagery was selected in such a way that (1) the maximum cloud cover of the study area at satellite overpass time was a maximum of 10% and (2) their dates were close enough to each other so that seasonal changes did not affect the results of various analyses. The range of Day of Year (DOY) values of used Landsat images was less than 60 days. These images have the appropriate resolution to highlight the trend and surface coverage features. In addition to Landstat data, monthly LST (MOD11C3) and NDVI (MOD13C2) MODIS/Terra products were used every month from the beginning of 2000 to the end of 2017. The spatial resolution of these products is 5000 m. MODIS products were downloaded from https://lpdaac.usgs.gov/.

Methods
The complete steps of this research are shown in Figure 2. In the first step, the trend of LST and NDVI spatiotemporal variations of the study area (Gomishan city, Gomishan wetland) from 2000 to 2017 were evaluated and compared using the monthly MODIS products. In the second step, spectral indices and methods were implemented to model the biophysical characteristics after pre-processing the Landsat imagery. Then, spatiotemporal variations of the Landsat-derived biophysical characteristics for the study area (Gomishan wetland and Gomishan city) were evaluated and compared with each other. In the third step, the land cover of the study area was classified using Landsat images. Then, the spatiotemporal changes of the different land covers were examined. In the fourth step, the main surface characteristics, including heat, greenness, dryness, and wetness, were combined based on RSEI, and surface ecological conditions were modeled using Landsat imagery. Then, the spatiotemporal variations for the study area (Gomishan wetland and Gomishan city) were evaluated and compared with each other.

Spatiotemporal Changes of LST and NDVI Using MODIS Products
In this study, the trend of LST and NDVI spatiotemporal variations in the study area was investigated using MODIS monthly LST and NDVI products. For this purpose, the temporal variations of the mean LST and NDVI in the study area for each month during the period from 2000 to 2017 were calculated and evaluated. The temporal variations of the mean LST for the study area, Gomishan city and the Gomishan wetland in different months were evaluated and compared with each other. To investigate spatiotemporal variations, the variance of LST and NDVI values in the region and pixel scales were calculated monthly and annually and the variance maps of LST and NDVI values in pixel scale were produced. Then, the spatiotemporal variations of LST and NDVI were compared for the study area, Gomishan city and the Gomishan wetland. High values of variance for one pixel, indicated instability in the surface characteristics of the pixel during the study period. The values of this parameter are zero and positive. The larger the variance of the surface characteristic values at different times, the larger the temporal variation of that surface characteristic. Azero value of this index in any pixel indicated no variations of surface characteristics in that pixel over time. Additionally, the Coefficient of Variation (CV) has been used to evaluate the temporal changes of the average surface characteristics at the study area, the Gomishan wetland and city scales.

Spatiotemporal Changes of LST and NDVI Using MODIS Products
In this study, the trend of LST and NDVI spatiotemporal variations in the study area was investigated using MODIS monthly LST and NDVI products. For this purpose, the temporal variations of the mean LST and NDVI in the study area for each month during the period from 2000 to 2017 were calculated and evaluated. The temporal variations of the mean LST for the study area, Gomishan city and the Gomishan wetland in different months were evaluated and compared with each other. To investigate spatiotemporal variations, the variance of LST and NDVI values in the region and pixel scales were calculated monthly and annually and the variance maps of LST and NDVI values in pixel scale were produced. Then, the spatiotemporal variations of LST and NDVI were compared for the study area, Gomishan city and the Gomishan wetland. High values of variance for one pixel, indicated instability in the surface characteristics of the pixel during the study period. The values of this parameter are zero and positive. The larger the variance of the surface characteristic values at different times, the larger the temporal variation of that surface characteristic. Azero value of this index in any pixel indicated no variations of surface characteristics in that pixel over time. Additionally, the Coefficient of Variation (CV) has been used to evaluate the temporal changes of the average surface characteristics at the study area, the Gomishan wetland and city scales.

Spatiotemporal Changes of Surface Biophysical Characteristics Using Landsat Imagery
Landsat images were used to more accurately compare spatiotemporal variations in the biophysical characteristics of Gomishan wetland and Gomishan city. Therefore, the variance of the Landsat-derived biophysical characteristics of the Gomishan wetland and Gomishan city were calculated and compared.
Extraction of Surface Biophysical Characteristics from Landsat Images Radiometric conversions and atmospheric correction of Landsat satellite images were performed for the pre-processing. The details of Landsat images pre-processing are provided in [44][45][46].
Landsat reflective and thermal bands have been used to calculate surface biophysical characteristics. In this study, 9 widely used spectral indices, presented in Table 2, were used to extract surface biophysical characteristics. These indices include LST, impervious surfaces, vegetation, moisture, and salinity information.

Land Cover Classification
Land cover classes in the study area were determined based on information obtained from field surveys and visual interpretation of Landsat RGB (red, green and blue) images. These classes include urban, green space, water, and bare soil lands. In this study, the maximum likelihood method was used for image classification. In this method, after evaluating the probabilities in each class, the pixels are assigned to the class with the highest probability [57].
Training and test datasets are needed to supervise the classification of satellite imagery and evaluate land cover classification results. In this study, training and test datasets were collected through visual interpretation of false color composite images, field survey-derived information, and Google Earth images. For each land cover class, more than 300 training samples were selected for classification, and 200 test samples were selected for evaluation. To assess the accuracy of the land cover maps prepared from the classification, Kappa coefficient and overall accuracy criteria were used. After preparing the land cover maps, the area of each class and the amount of its change during the period of Remote Sens. 2020, 12, 2989 8 of 24 1987 to 2017 were calculated and evaluated. Finally, the CV was used to compare the temporal changes of the Gomishan wetland and city.

Modeling the Surface Ecological Conditions
In this study, RSEI was used to model the surface ecological conditions [15,26]. To model the surface ecological conditions based on this index, the main information of the surface characteristics, including surface heat, greenness, dryness, and wetness, were combined with each other (Equation (1)). To assessing the surface ecological conditions based on RSEI, first, the values of the used indices including NDVI, wetness, LST and brightness were standardized between 0 and 1 to reduce the effect of climatic and meteorological conditions and co-scaling of information obtained from different indicators and sensors on the results of the RSEI. RSEI = PCA(NDVI, Wetness, LST, Brightness) (1) PCA was utilized here for identifying the relative importance of NDVI, wetness, LST and brightness variables. The eigenvalue of the first component of PCA (PC1) integrates most characteristics of all variables, and therefore a surface ecological conditions map was built with PC1 in this study [5,15,26,34]. The value of the RSEI index is standardized between 0 and 1. RSEI values of 1 and 0 indicate the worst and best surface ecological conditions, respectively. Finally, the RSEI was classified according to 5 classes of surface ecological conditions, which were very good (0-0.2), good (0.2-0.4), acceptable (0.4-0.6), bad (0.6-0.8), and very bad (0.8-1) [14,15]. After preparing classified RSEI maps, the mean of RSEI and the area of each surface ecological condition class during the period of 1987 to 2017 were calculated. Variance was used to evaluate the temporal changes of surface ecological conditions at the pixel scale and CV was used to evaluate and compare the temporal changes of the average RSEI and the area of surface ecological conditions classes at the scale of Gomishan wetland and city. High values of variance and CV indicate high temporal changes in surface ecological conditions.

NDVI and LST Spatiotemporal Variations of the Study Area Using MODIS Products
The LST means of the study area were 12.0, 13.9, 18.6, 24.7, 31.8, 35.5, 36.5, 37.5, 34.1, 28.2, 19.3, and 12.8 • C in the months of January, February, March, April, May, June, July, August, September, October, November, and December, respectively ( Figure 3). The temporal variations the LST mean for the study area vary in different months. On a monthly scale, the highest and lowest standard deviation (SD) of LST means for the study area were April (2.5 • C) and February (1.3 • C), respectively.
The LST means for the study area in winter, spring, summer, and autumn were 14.9, 30.7, 36.1, and 20.2 • C, respectively. The highest and lowest coefficient of variation (CV) of the LST were for winter (0.08) and summer (0.04), respectively. Additionally, the LST mean for the study area varies for different years. The highest and lowest LST means were for 2010 (27. Figure 4). Additionally, the monthly mean of NDVI for the region varies for different years. The highest and lowest NDVI means were for 2010 (0.24) and 2001 (0.14), respectively. As with LST, the CVs of the monthly NDVI mean in April (0.06), March (0.06), February (0.05) and May (0.05) were higher than in other months, while the lowest CV of the monthly NDVI mean was in September (0.01). The highest and lowest CV of NDVI were for spring (0.19) and summer (0.09), respectively. The LST means for the study area in winter, spring, summer, and autumn were 14.9, 30.7, 36.1, and 20.2 °C, respectively. The highest and lowest coefficient of variation (CV) of the LST were for winter (0.08) and summer (0.04), respectively. Additionally, the LST mean for the study area varies for different years. The highest and lowest LST means were for 2010 (27. Figure 4). Additionally, the monthly mean of NDVI for the region varies for different years. The highest and lowest NDVI means were for 2010 (0.24) and 2001 (0.14), respectively. As with LST, the CVs of the monthly NDVI mean in April (0.06), March (0.06), February (0.05) and May (0.05) were higher than in other months, while the lowest CV of the monthly NDVI mean was in September (0.01). The highest and lowest CV of NDVI were for spring (0.19) and summer (0.09), respectively.   The LST means for Gomishan city and the Gomishan wetland in different months in the period of 2000-2017 is shown in Figure 5. The trend of LST mean temporal variations in these two parts of the study area between 2000 and 2017 was similar. The difference between the LST means of the wetland and the city of Gomishan were high in some dates. However, in a significant number of months in different years, the LST means for Gomishan city and the Gomishan wetland were almost equal.  The LST means for Gomishan city and the Gomishan wetland in different months in the period of 2000-2017 is shown in Figure 5. The trend of LST mean temporal variations in these two parts of the study area between 2000 and 2017 was similar. The difference between the LST means of the wetland and the city of Gomishan were high in some dates. However, in a significant number of months in different years, the LST means for Gomishan city and the Gomishan wetland were almost equal. m spatial resolution.
The LST means for Gomishan city and the Gomishan wetland in different months in the period of 2000-2017 is shown in Figure 5. The trend of LST mean temporal variations in these two parts of the study area between 2000 and 2017 was similar. The difference between the LST means of the wetland and the city of Gomishan were high in some dates. However, in a significant number of months in different years, the LST means for Gomishan city and the Gomishan wetland were almost equal. The LST variance values of each pixel for each month are shown in Figure 6. Areas with high (red) and low (green) variance values have high and low temporal variations of LST, respectively. The spatial distribution of LST variance varies in different months and areas. The mean of LST variance values of the study area were 3 (1), 4 (2), 8 (6), 11.6 (8.1), 7.1 (6.0), 5.5 (5.5), 5.9 (4.9), 7.5 (5.7),  The LST variance values of each pixel for each month are shown in Figure 6. Areas with high (red) and low (green) variance values have high and low temporal variations of LST, respectively. The spatial distribution of LST variance varies in different months and areas. The mean of LST variance values of the study area were 3 (1), 4 (2), 8 (6), 11.6 (8.1), 7.1 (6.0), 5.5 (5.5), 5.9 (4.9), 7.5 (5.7), 6.1 (4.1), 4.3 (3.0), 6.9 (4.1), and 4.6 (2.5) • C January through December, respectively. The mean LST variance values for winter, spring, summer, and autumn were 5.5, 8.1, 6.5, and 5.3 °C, respectively. The LST of the Gomishan wetland changed more in June, July, August, and September than in other months, which indicates high temporal variations in the wetland's water level during these months. In general, the temporal variations of LST in the northern half of the region were greater than in the southern half. The temporal variations of LST in the eastern part of the region, which includes the Caspian Sea, were lower than in other areas.
The NDVI variance values for the study area were calculated separately for different months and are shown in Figure 7 in pixel scale. The results show that the temporal variation of NDVI for the margins of the southeastern areas of the region with agricultural lands and the Caspian Sea were high in January, March, February, April, and May. However, for June, July, August, September, October, November, and December, high levels of NDVI variance were limited to the Caspian Sea margins. The highest and lowest NDVI variance values were in March (0.023) and August (0.008), The mean LST variance values for winter, spring, summer, and autumn were 5.5, 8.1, 6.5, and 5. • C, respectively. The LST of the Gomishan wetland changed more in June, July, August, and September than in other months, which indicates high temporal variations in the wetland's water level during these months. In general, the temporal variations of LST in the northern half of the region were greater than in the southern half. The temporal variations of LST in the eastern part of the region, which includes the Caspian Sea, were lower than in other areas.
The NDVI variance values for the study area were calculated separately for different months and are shown in Figure 7 in pixel scale. The results show that the temporal variation of NDVI for the margins of the southeastern areas of the region with agricultural lands and the Caspian Sea were high in January, March, February, April, and May. However, for June, July, August, September, October, November, and December, high levels of NDVI variance were limited to the Caspian Sea margins. The highest and lowest NDVI variance values were in March (0.023) and August (0.008), respectively. Additionally, the spatial distribution of NDVI variance varies in different months and locations.
Remote Sens. 2020, 12, x FOR PEER REVIEW 13 of 25 respectively. Additionally, the spatial distribution of NDVI variance varies in different months and locations. The LST and NDVI variance maps of each pixel were calculated annually, and the results are shown in Figure 8. The mean (SD) values of the LST and NDVI variance of the study area were 6.54 (4.04) °C and 0.009 (0.006), respectively. The correlation coefficient (r) between the LST and NDVI variance values of the study area is 0.78, which indicates the spatial consistency of LST and NDVI temporal variations in the study area. The mean of the LST (NDVI) variance for Gomishan city and the Gomishan wetland were 2.1 (12.1) °C and 0.005 (0.013), respectively. Most of the temporal variations in LST and NDVI for the region were related to the Gomishan wetland that is bordered by the Caspian Sea. The results showed that during the period of 2000 to 2017, the temporal variations in the Gomishan wetland's LST and NDVI were greater than those in Gomishan city. The LST and NDVI variance maps of each pixel were calculated annually, and the results are shown in Figure 8. The mean (SD) values of the LST and NDVI variance of the study area were 6.54 (4.04) • C and 0.009 (0.006), respectively. The correlation coefficient (r) between the LST and NDVI variance values of the study area is 0.78, which indicates the spatial consistency of LST and NDVI temporal variations in the study area. The mean of the LST (NDVI) variance for Gomishan city and the Gomishan wetland were 2.1 (12.1) • C and 0.005 (0.013), respectively. Most of the temporal variations in LST and NDVI for the region were related to the Gomishan wetland that is bordered by the Caspian Sea.
The results showed that during the period of 2000 to 2017, the temporal variations in the Gomishan wetland's LST and NDVI were greater than those in Gomishan city.

Spatiotemporal Variations in Biophysical Characteristics Using Landsat Images
The mean of surface biophysical characteristics for the study area (Gomishan city and the Gomishan wetland) was calculated using Landsat imagery (Figure 9). The mean of surface biophysical characteristics was different in various land cover classes. Additionally, the difference between the means of surface biophysical characteristics for different locations in the study area (the Gomishan wetland and Gomishan city) was high. Due to water cover, the average amount of LST, NDVI, NDBI, brightness, albedo and SI1 in the Gomishan wetland is less than other areas.
The SD of the mean values of LST, NDVI, NDWI, NDBI, brightness, albedo, wetness, NDSI, and SI1 for the study area (Gomishan city and the Gomishan wetland) were 3.7

Spatiotemporal Variations in Biophysical Characteristics Using Landsat Images
The mean of surface biophysical characteristics for the study area (Gomishan city and the Gomishan wetland) was calculated using Landsat imagery (Figure 9). The mean of surface biophysical characteristics was different in various land cover classes. Additionally, the difference between the means of surface biophysical characteristics for different locations in the study area (the Gomishan wetland and Gomishan city) was high. Due to water cover, the average amount of LST, NDVI, NDBI, brightness, albedo and SI1 in the Gomishan wetland is less than other areas.

Spatiotemporal Variations in Biophysical Characteristics Using Landsat Images
The mean of surface biophysical characteristics for the study area (Gomishan city and the Gomishan wetland) was calculated using Landsat imagery (Figure 9). The mean of surface biophysical characteristics was different in various land cover classes. Additionally, the difference between the means of surface biophysical characteristics for different locations in the study area (the Gomishan wetland and Gomishan city) was high. Due to water cover, the average amount of LST, NDVI, NDBI, brightness, albedo and SI1 in the Gomishan wetland is less than other areas.
Temporal variations of the biophysical characteristics were calculated on a pixel scale, and the results are presented in Figure 10. The biophysical characteristics temporal variation was heterogeneous. Most of the biophysical characteristics temporal variations in the study area are in the Gomishan wetland.
The mean surface biophysical characteristics variance values for the Gomishan wetland, Gomishan city, and the study area were calculated ( Table 3). The temporal variation in surface biophysical characteristics was higher in the Gomishan wetland and lower in Gomishan city. Additionally, the highest and lowest temporal variations in surface biophysical characteristics were related to NDVI and albedo, respectively. The mean surface biophysical characteristics variance values for the Gomishan wetland, Gomishan city, and the study area were calculated ( Table 3). The temporal variation in surface biophysical characteristics was higher in the Gomishan wetland and lower in Gomishan city. Additionally, the highest and lowest temporal variations in surface biophysical characteristics were related to NDVI and albedo, respectively.

Land Cover Maps
Mean overall classification accuracy (kappa coefficient) of urban, bare land, green space and water classes on different dates was 90 (0.89), 92 (0.90), 93 (0.92) and 95 (0/93), respectively. The highest and lowest classification accuracy was for water and urban classes, respectively. The evaluation results showed that the accuracy of land cover maps obtained from the satellite images classification was acceptable [55,58]. Land cover maps for the study area are shown in Figure 11. The spatiotemporal variations of land covers in the study area were very high. In recent years, the water area in the Gomishan wetland has changed significantly, which is a function of the inward and

Land Cover Maps
Mean overall classification accuracy (kappa coefficient) of urban, bare land, green space and water classes on different dates was 90 (0.89), 92 (0.90), 93 (0.92) and 95 (0/93), respectively. The highest and lowest classification accuracy was for water and urban classes, respectively. The evaluation results showed that the accuracy of land cover maps obtained from the satellite images classification was acceptable [55,58]. Land cover maps for the study area are shown in Figure 11. The spatiotemporal variations of land covers in the study area were very high. In recent years, the water area in the Gomishan wetland has changed significantly, which is a function of the inward and outward movements of the Caspian Sea. Moreover, the amount of vegetation in the northern part of the study area has decreased significantly, which has led to an increase in the amount of bare lands in these areas. However, in the south of the study area, some of the bare lands have turned into green and agricultural lands. outward movements of the Caspian Sea. Moreover, the amount of vegetation in the northern part of the study area has decreased significantly, which has led to an increase in the amount of bare lands in these areas. However, in the south of the study area, some of the bare lands have turned into green and agricultural lands. The areas of different land cover classes in the study period are shown in Table 4. The area of Gomishan city has increased from 2.15 Km 2 in 1987 to 6.48 Km 2 in 2017. The highest and lowest green space areas were found in 1992 (715) and 1994 (323 Km 2 ). The CV of the area of urban, bare soil, green space, and water lands were 0.34, 0.11, 0.36, and 0.06, respectively. The results show that the spatial and temporal changes in the green space lands were greater than those in other land covers in the study area.  The areas of different land cover classes in the study period are shown in Table 4. The area of Gomishan city has increased from 2.15 Km 2 in 1987 to 6.48 Km 2 in 2017. The highest and lowest green space areas were found in 1992 (715) and 1994 (323 Km 2 ). The CV of the area of urban, bare soil, green space, and water lands were 0.34, 0.11, 0.36, and 0.06, respectively. The results show that the spatial and temporal changes in the green space lands were greater than those in other land covers in the study area. The temporal variation trend of water land area in the Gomishan wetland is shown in Figure 12. The temporal variations in the water land area of this wetland have changed during the study period. From 1987 to 1995, the changes in the water land area in the Gomishan wetland increased, but from 1995 to 2017, they decreased. The highest and lowest water land area in the Gomishan wetland were in 1995 (22.04) and 2016 (237.17 km 2 ), respectively. The CV of the water land area in the Gomishan wetland and Gomishan city area were 0.45 and 0.34, respectively, which indicates more changes in the water land area in the Gomishan wetland than in Gomishan city. The temporal variation trend of water land area in the Gomishan wetland is shown in Figure 12. The temporal variations in the water land area of this wetland have changed during the study period. From 1987 to 1995, the changes in the water land area in the Gomishan wetland increased, but from 1995 to 2017, they decreased. The highest and lowest water land area in the Gomishan wetland were in 1995 (22.04) and 2016 (237.17 km 2 ), respectively. The CV of the water land area in the Gomishan wetland and Gomishan city area were 0.45 and 0.34, respectively, which indicates more changes in the water land area in the Gomishan wetland than in Gomishan city.

Spatiotemporal Variations of the Surface Ecological Conditions
The RSEI means for the study area, Gomishan city and the Gomishan wetland, were 0.43, 0.65, and 0.29, respectively ( Figure 13). The surface ecological conditions of Gomishan city were worse than the surface ecological conditions of the study area and the Gomishan wetland. The CV of RSEI means for the study area, Gomishan city, and the Gomishan wetland were 0.10, 0.88, and 0.51, respectively. The surface ecological conditions of these lands have changed during the study period. Temporal variations in the surface ecological conditions of the Gomishan wetland were greater than in the study area and Gomishan city. The lowest and highest RSEI means were in 1995 (0.11) and 2015 (0.63), respectively.  1987 1989 1990 1991 1992 1994 1995 1996 1997 1998 1999

Spatiotemporal Variations of the Surface Ecological Conditions
The RSEI means for the study area, Gomishan city and the Gomishan wetland, were 0.43, 0.65, and 0.29, respectively ( Figure 13). The surface ecological conditions of Gomishan city were worse than the surface ecological conditions of the study area and the Gomishan wetland. The CV of RSEI means for the study area, Gomishan city, and the Gomishan wetland were 0.10, 0.88, and 0.51, respectively. The surface ecological conditions of these lands have changed during the study period. Temporal variations in the surface ecological conditions of the Gomishan wetland were greater than in the study area and Gomishan city. The lowest and highest RSEI means were in 1995 (0.11) and 2015 (0.63), respectively.
Remote Sens. 2020, 12, x FOR PEER REVIEW 19 of 25 Figure 13. The RSEI mean for the study area, Gomishan wetland, and the Gomishan city.
The r between the water land areas in the Gomishan wetland and the RSEI means for the Gomishan wetland and the study area were −0.96 and −0.69, respectively ( Figure 14). With the advancement of the Caspian Sea and the increase in the water land area in the Gomishan wetland, the quality of surface ecological conditions in this region has improved.  8 1987 1989 1990 1991 1992 1994 1995 1996 1997 1998 1999  The r between the water land areas in the Gomishan wetland and the RSEI means for the Gomishan wetland and the study area were −0.96 and −0.69, respectively ( Figure 14). With the advancement of the Caspian Sea and the increase in the water land area in the Gomishan wetland, the quality of surface ecological conditions in this region has improved. Figure 13. The RSEI mean for the study area, Gomishan wetland, and the Gomishan city.
The r between the water land areas in the Gomishan wetland and the RSEI means for the Gomishan wetland and the study area were −0.96 and −0.69, respectively ( Figure 14). With the advancement of the Caspian Sea and the increase in the water land area in the Gomishan wetland, the quality of surface ecological conditions in this region has improved.  1987 1989 1990 1991 1992 1994 1995 1996 1997 1998 1999  A visual survey of classified RSEI maps shows that the spatial distribution of surface ecological conditions varies at different dates ( Figure 15). The areas covered with bare soil have poor surface ecological conditions, and the areas covered with water have excellent surface ecological conditions. With the retreat of water from the Gomishan wetland between 1987 and 2017, the surface ecological conditions of this part of the region have changed from excellent and very good to good. The areas of surface ecological condition classes were calculated, and the results are shown in Figure 16. Variations in the temporal and spatial dimensions of the study area's ecological conditions include both different class areas and different class types. Over the years, the surface ecological condition has worsened and most of the study area is now in the poor class. The CVs of the excellent, very good, good, fair, and poor surface ecological conditions' classes' areas were 0.09, 0.87, 0.77, 0.23, and 0.63, respectively. The maximum temporal variations among the classes of the surface ecological The areas of surface ecological condition classes were calculated, and the results are shown in Figure 16. Variations in the temporal and spatial dimensions of the study area's ecological conditions include both different class areas and different class types. Over the years, the surface ecological condition has worsened and most of the study area is now in the poor class. The CVs of the excellent, very good, good, fair, and poor surface ecological conditions' classes' areas were 0.09, 0.87, 0.77, 0.23, and 0.63, respectively. The maximum temporal variations among the classes of the surface ecological conditions were related to the Gomishan wetland. The areas of surface ecological condition classes were calculated, and the results are shown in Figure 16. Variations in the temporal and spatial dimensions of the study area's ecological conditions include both different class areas and different class types. Over the years, the surface ecological condition has worsened and most of the study area is now in the poor class. The CVs of the excellent, very good, good, fair, and poor surface ecological conditions' classes' areas were 0.09, 0.87, 0.77, 0.23, and 0.63, respectively. The maximum temporal variations among the classes of the surface ecological conditions were related to the Gomishan wetland.

Discussion
Modeling surface ecological conditions is critical. Past studies in modeling surface ecological conditions have focused on modeling the impact of human activity, including population growth

Discussion
Modeling surface ecological conditions is critical. Past studies in modeling surface ecological conditions have focused on modeling the impact of human activity, including population growth and urban physical expansion, on the spatiotemporal changes in surface ecological conditions [14,15,26]. Hence, variations in the surface ecological conditions of an area can be caused by natural and unnatural factors [7][8][9][10].
In this study, various spectral indices and methods were used to model spatiotemporal variations of the surface ecological conditions. Using a mathematical calculation between two or more reflective bands, spectral indices extract useful and effective information from the biophysical characteristics of the surface of original images. Thus, a target phenomenon is identified through spectral indices. In this study, impermeable surfaces, wetness and greenness are the three main components of information obtained from the spectral methods and indices to model the surface ecological conditions. Impermeable surfaces indicate information that includes bare soils and constructed lands, greenness indicates vegetation information, and wetness indicates water-related characteristics, soil moisture, plants, and constructed lands. In addition, LST was used as a biophysical index to assess spatiotemporal variations in surface ecological conditions. However, considering other surface characteristics, such as salinity in barren lands and vegetation health for agricultural and green space lands, can increase the accuracy of modeling the surface ecological conditions. RSEI was developed for the modeling of SES. This was developed solely using satellite data [15] by integrating surface dryness, wetness, greenness, and heat information and showed a strong correlation with SES [14,26]. This index is very effective in measuring the pressures of anthropogenic activities on the environment, changes in vegetation and water and their consequences. Overall the advantages of the RSEI can be summarized as: (a) visualizable, (b) scalable, (c) comparable at different scales and (d) customizable to minimize error or variation caused by other properties in weight definitions [15,26,59]. However, using RSUSEI and CEEI to model the surface ecological conditions and comparing their results with RSEI results can be useful.
One of the weaknesses of the previous studies is the modeling of temporal changes in surface ecological conditions. In previous studies, a few limited images have been used to model the temporal variation of surface ecological conditions [15,34,35]. To more accurately model the spatiotemporal variations of the surface ecological conditions, Landsat images and MODIS products were used. MODIS products have a low spatial resolution but a high temporal resolution, which is more suitable for modeling the temporal variation of ecological conditions. Consequently, Landsat images were used to more accurately model spatial variations of surface ecological conditions. Additionally, using radar images, hyperspectral images and high spatial resolution (synthetic aperture radar (SAR), Hyperion, AVIRIS, Sentinel 1 and 2, Worldview-3, etc.) might have even more potential.
The modeled biophysical characteristics of the study area, based on different spectral indicators, have many spatiotemporal variations during different months of the year due to the location of the Gomishan wetland and the surface cover of Gomishan city. These temporal variations are related to the increase in human activities, the increase in saline lands, the progression and regression of the Caspian Sea water level, and the spatiotemporal changes in the water level of the Gomishan wetland [43]. The results of the land cover classification show that the vegetation cover, the size of the city, and the area of the water levels of the Gomishan wetland have changed significantly over the past years. The Gomishan wetland is connected to the Caspian Sea, which directly affects its hydrological characteristics. Temporal variations in the ecological conditions of the Gomishan wetland are higher than those of the study area and Gomishan city. Additionally, the ecological conditions of the study area are affected by the spatiotemporal changes in the water level of the Gomishan wetland. The results indicated that the effect of natural factors in changing the surface ecological conditions can be greater than the effect of natural factors. Therefore, the implementation of detailed policies and programs to control and improve the conditions of natural areas, including wetlands, can be of great importance in preventing the destruction of ecological conditions of ecosystems. In this regard, the first program could be the development of Web Geographic Information System (GIS) to monitor changes in the ecological conditions of wetlands based on satellite information.

Conclusions
Studying the spatiotemporal variations of surface biophysical characteristics and surface ecological conditions is critical for solving the challenges of environmental degradation, improving the structure and performance of ecosystems, and maintaining ecosystem services. Therefore, this study compared the degree of variation between surface ecological conditions caused by natural factors and those caused by unnatural factors. In this regard, spatiotemporal variations in the ecological conditions of Gomishan city and the Gomishan wetland have been evaluated and compared with each other. The results show that the Gomishan wetland has a lower mean LST than Gomishan city in most years. Moreover, the highest and lowest differences between various level covers are related to NDVI and albedo.
The maximum variations between the surface biophysical characteristics of the study area are related to the Caspian Sea margins. The Gomishan wetland and Gomishan city have the highest and lowest variations in biophysical characteristics, respectively. Additionally, the CV in the water lands area in the Gomishan wetland is greater than in Gomishan city. The RSEI means for the study area, Gomishan city, and the Gomishan wetland during the study period are 0.43, 0.65, and 0.29, respectively. The ecological conditions of the study area, Gomishan city, and the Gomishan wetland changed during the study period. The surface ecological conditions of Gomishan city are worse than the ecological conditions of the study area and the Gomishan wetland. There were more changes to the surface ecological conditions of the Gomishan wetland than to those of the study area and Gomishan city. The most important factor that naturally affects the surface ecological characteristics of the Gomishan wetland is fluctuations in the water level of the Caspian Sea, which leads to a decrease and increase in the water level of the wetland.