Drought monitoring and its effects on vegetation and water extent changes using remote sensing data in Urmia Lake watershed, Iran

Drought is a climatic, regional, and occasional phenomenon with three dimensions: intensity, durability, and spatial extent (Palmer 1965; Rossi et al. 1992). The drought hazard is among the natural disasters that affect people's lives and the environment, with continuing and severe consequences (Mishra & Desai 2005). In the past decades, droughts have had devastating effects on ecology, water resources management, and environmental processes (Rojas et al. 2011) and need to be recorded and analyzed extensively. Spatial and temporal monitoring of drought is a complicated process, and it has been developed over some decades (Nalbantis 2008; Diaz et al. 2020). Drought indicators such as water anomalies are used to monitor and identify water scarcity and dry spells (Diaz et al. 2020). The spatial, temporal, and severity (hardness) of drought can be determined using these indicators (Benedetti et al. 1994). Drought hazards can influence meteorological and hydrological components in different spatial and temporal scales.

There are several indicators to assess meteorological drought. These indicators include the Standardized Precipitation Index (SPI), the Standardized Precipitation–Evapotranspiration Index (SPEI), the Soil Moisture Anomaly (SMA), the Drought Severity Index (DSI), the Palmer Drought Severity Index (PDSI), etc. The SPI is a widely used index to characterize meteorological drought on a range of timescales. On short timescales, the SPI is closely related to soil moisture, while at longer timescales, the SPI can be related to groundwater and reservoir storage. The SPI can be compared across regions with markedly different climates. It quantifies observed precipitation as a standardized departure from a selected probability distribution function that models the raw precipitation data. The raw precipitation data are typically fitted to a gamma or a Pearson Type III distribution, and then transformed to a normal distribution (Yihdego et al. 2019). There are advantages to using this index. It uses precipitation only; and can characterize drought or abnormal wetness at different timescales which correspond with the time availability of different water resources (e.g., soil moisture, snowpack, groundwater, river discharge, and reservoir storage). The more comparable across regions with different climates than the Palmer Severity Drought Index (PDSI), the less complex to calculate than the PDSI (Hao & Singh 2015).

In terms of platforms, the advantages of satellite-based remote sensing include high spatial resolution, which makes possible the extraction of long-time data series of consistent and comparable data, which can be cost-effective. Furthermore, some satellite platforms have free access to visible and multispectral data, such as Landsat 7-8. However, there are two main problems with these platforms for precision agriculture applications, which are related to the per-pixel resolution (30 m² per pixel for Landsat and 500 m² for MODIS) and the orbit period (16 days for Landsat and 26 days for SPOT). More recently, pixel resolution has been increased by newer satellites, such as WorldView-2 and -3 (Digital Globe, Longmont, CO, USA). WorldView-2 was the first commercial high-resolution satellite to provide eight spectral sensors in the visible to near-infrared range. Along with the four typical multispectral bands: blue (450–510 nm), green (510–580 nm), red (630–690 nm), and near-infrared (NIR) (770–895 nm), each sensor is narrowly focused on a particular range of the electromagnetic spectrum that is sensitive to a particular feature from the ground or a property of the atmosphere. However, images from this platform can be cost-prohibitive for long-time data series studies (Xue & Su 2017). That's why using MODIS images looks so much better because of its easy access, larger coverage, higher bandwidth, and simpler processing. But if the area is small and the time-series period is less, it is better to use data with higher spatial and spectral accuracy (Bannari et al. 1995).

There are several indicators in the study of drought and vegetation using remote sensing, including the Ratio Vegetation Index (RVI), the Difference Vegetation Index (DVI), the Perpendicular Vegetation Index (PVI), the Vegetation Index (GVI), the Normalized Difference Vegetation Index (NDVI), the Atmospherically Resistant Vegetation Index (ARVI), the Soil Atmospheric Impedance Vegetation Index (TSARVI), the Transformed Soil-Adjusted Vegetation Index (TSAVI), the Soil-Adjusted Vegetation Index (SAVI), the Enhanced Vegetation Index (EVI), etc. The use of these combined indices, which include NDVI, EVI, TVDI, and Normalized Difference Water Index (NDWI) Dubai, is very useful for checking the amount of vegetation and humidity in the area. For these reasons, such as the TVDI, it uses temperature and vegetation at the same time to check the humidity. This index forms a regression relationship between temperature and vegetation to identify wet and dry areas. Vegetation indices, which include EVI and NDVI, identify the vegetation using the reflection rates of red, near-infrared, and green bands. The NDWI is also a moisture and plant index.

Although meteorological and hydrological drought studies have received more attention in recent years, the extension potential for the drought risk has not yet been extensively studied (Gu et al. 2020). Many studies conducted the drought analysis using meteorological indicators and remote sensing data over different climatic conditions. In this regard, Hadian et al. (2017) investigated the effect of drought using satellite images on vegetation and land use around the Maharloo Lake in Fars province. They observed that the highest amount of reduction in vegetation canopy cover and destruction of natural lands has occurred in three different periods (9-, 13-, and 23-year periods). The changes in vegetation cover of the Isfahan province were studied by Safari Shad et al. (2017) by applying MODIS satellite images from 2000 to 2008. They used the NDVI and showed that changes in the NDVI had no significant correlation with agricultural and meteorological drought. Luo et al. (2016) investigated the seasonal variations of the Lake Taihu in China and its impacts on the vegetation regions using Landsat satellite imagery. Their results showed that air temperature directly impacted vegetation growth and growth retardation. Also, the soil water content/moisture assessment can be considered a crucial variable of the soil–atmosphere system, as it affects various processes and has multi-direction feedback within the climate system.

Meanwhile, soil moisture is of significant consequence for the global water–soil–atmosphere cycles. The occurrence of drought events is in close relation to water components, affecting the available water content to vegetation. The Temperature–Vegetation–Dryness Index (TVDI) is a widely used satellite-derived index in monitoring soil water content in different studies.

The relationship of the drought index (SPI) with the Difference Vegetation Index (DVI) and the NDVI (obtained from MODIS sensor images) were examined by Saberi et al. (2018) in West Azerbaijan province, Iran. According to their findings, the SPI had the maximum correlation with the NDVI to determine the dry-period changes. Also, Saberi et al. (2019) investigated the drought in Lake Urmia by analyzing the correlation between the SPI and the NDVI. The results highlighted that the short-term drought indicator in that basin exhibited a remarkable correlation with the monthly NDVI. Zhang et al. (2020) explored the effects of drought on runoff changes in the Ebinur Lake watershed using the NDVI and the NDWI. Their findings indicated that the lake water levels were associated with changes in many variables, and the vegetation in the region experienced significant changes in response to the drought with a significant correlation. In the study of Shamshirband et al. (2020) using machine learning models, Zhao et al. (2018) using GRACE and Wu & Chau (2013) using Modular Soft Computing methods drought and hydrology have been studied.

In the above review, many different studies have analyzed drought in various fields. These studies have examined only one or at most one drought index of distance meteorological drought. In this study, we seek to determine the extent and impact of drought in this area using various indicators. The Urmia region has always been considered due to the many changes that have taken place in its water zone. This is because the water of the Lake Urmia and the amount of drought in this area has a great impact on the agriculture of the two neighboring provinces, especially the province of West Azerbaijan. The Urmia region, especially with the Lake Urmia, has abundant agriculture and fertile soil. The lake is located in the middle of a closed catchment so that all surface and groundwater overflows from the surrounding areas to the lake. The special position of the basin in terms of geology, high evaporation rate, and continuous accumulation of salts in it has led to the transformation of the lake into an extremely saline one. The lake is surrounded by a number of freshwater satellite lagoons. The complex of these wetlands has created an important ecological area around the lake.

The purpose of this study was to monitor the meteorological drought index and its effects on the vegetation and water areas in the Lake Urmia basin, Iran, using Moderate Resolution Imaging Spectroradiometer (MODIS) images and meteorological data. In this study, between 2000 and 2020, a 2-year interval and an average annual time series were used. The reason for not using the images of newer years such as 2022 is the lack of all images of this year to average the drought for this year. The SPI, NDVI, EVI, NDWI, and TVDI were derived from MODIS data. Also, the correlation analysis was utilized to explore the interdependency among the meteorological and vegetation indices.

The Lake Urmia watershed is located in northwest Iran with 51,876 km². The study area lies between 35 °40′ to 38 °30′ N latitude and 47 °53′ to 44 °07′ E longitude (Figure 1). The altitude of the study area ranges from 1,258 to 3,656 m above the mean sea level. The watershed of the Lake Urmia is part of the Great Plateau of Armenia and is heterogeneous in terms of terrain complexity (plain (24%) and mountainous (65%)) and climatic regimes. The mountainous regions of the area are steep and prone to soil erosion and highly flooding. The area of the lake is approximately 11% (Godarzi et al. 2015; Salahi et al. 2017). The water level of the Urmia Lake is on average 1,280 m above sea level, which is the lowest elevation of the basin. Due to its geographical and mountainous position, this basin is regarded as the wettest in Iran. The land uses in the basin include aquifers, pastureland, and agricultural lands ((Heydari Tasheh Kabood et al. 2020). The minimum and maximum temperatures observed in the Urmia Lake basin during the study period are on average 4.3 and 17.7 °C, respectively, with an average annual precipitation of 359.1 mm. The average discharge of the basin is 18.2 m³/s. The geographical location of Lake Urmia and the studied meteorological stations are shown in Figure 1.

The data used in this study include MODIS sensor images aboard the Terra satellite corresponding from 2000 to 2020 (with 2-year time intervals). Several MODIS images were used (Table 1). The satellite image data were obtained from the National Aeronautics and Space Administration (NASA) website (https://search.earthdata.nasa.gov/search). The high spectral resolution, wide swath width, and the specific calibration techniques in the MODIS sensor result in higher accuracy in estimating NDVI, NDWI, EVI, and TVDI (Hulka 2008). The images of June (MOD09A1) were used to obtain NDWI. For this reason, this type of data is used to calculate this index because it is not readily available and must be calculated. But other indicators are readily available. Meanwhile, the available MODIS products (MOD13A3, MOD11A2) were used to calculate the annual average of NDVI, EVI, and TVDI.

Also, the meteorological data were obtained from Iran Meteorological Organization (http://www.irimo.ir/). Monthly precipitation data from eight synoptic stations of the Lake Urmia watershed, which covers all the study regions, were used from 2000 to 2020 to obtain 1-month and 3-month SPI. The locations of synoptic stations are shown in Figure 1. These stations had complete statistical data and non-missing and homogeneous.

The reason for examining remote sensing drought as a multi-year time series is because remote sensing drought does not occur in 1 day and occurs in time series, and also our emphasis on the meteorological drought index is long in time and there is no drought in series. It does not happen for a short time. If it is created, it is cross-sectional. Of course, it is important to note that meteorological data are easier to analyze, so it is given in a lower time series.

The SPI values were calculated based on the data of meteorological stations, and the NDVI, EVI, NDWI, and TVDI were extracted from satellite data to monitor the drought and changes in the lake water area during 2000–2020 as described as follows:

Since the NDWI effectively measures moisture content, it is often compared with the NDMI, also known as NDWI GAO. In fact, there is a major difference in how the two are calculated and used. The NDMI makes use of the NIR–SWIR (near-infrared and short-wave infrared) combination to enhance the presence of water in the leaves of plants. The NDWI, on the other hand, is calculated using the GREEN–NIR (visible green and near-infrared) combination, which allows it to detect subtle changes in the water content of the water bodies.

Green refers to the green band, and NIR stands for the near-infrared band. In MODIS data, NIR, RED, and Green represent band 1, band 2, and band 5, respectively. These index values vary between −1 and +1, theoretically values of water bodies are larger than 0.5. Vegetation has much smaller values, which results in distinguishing vegetation from water bodies easier. Built-up features have positive values between 0 and 0.2. The NDWI values correspond to the following ranges:

• 0.2–1: water surface,

• 0.0–0.2: flooding, humidity,

• − 0.3 to 0.0: moderate drought, non-aqueous surfaces,

• − 1 to −0.3: drought, non-aqueous surfaces.

NIR and RED reflect the near-infrared and red bands, respectively, and these index values vary between −1 and +1, theoretically (Saeidi et al. 2014). The negative value of the NDVI specifies the presence of barren land and water, while the dense green canopy results in high NDVI values because of a high concentration of chlorophyll with low reflectance in the red band (Pradhan & Lee 2010).

In Equation (7), the NDVI is the vegetation index obtained from Equation (4). Also, the coefficient a and the slope line b are determined using the least-squares fitting between temperature and vegetation. The values of this moisture index are between zero and one variable where the more it moves toward one, the wetter the area and the closer it gets to zero, the drier the area.

Error estimation and accuracy estimation are usually based on statistical parameters extracted from the error matrix. The error matrix is also called the Confusion Matrix. The result is a pixel-by-pixel comparison of known pixels (in terrestrial reality) with the corresponding pixels in the results. The label of each known pixel is compared to the corresponding pixel label. Then, the same results are added together and labels that do not match are also calculated (Fathy et al. 2011).

In cases where the real (observed) agreement approaches one and the probable agreement approaches zero, the value of K approaches one, which is ideal (Fathy et al. 2011).

Consequently, the NDVI, EVI, NDWI, and TVDI were calculated by ArcGIS software from MODIS data. The drought condition was characterized by using estimation of SPI during the available recorded period (2000–2020) in the study area. Then, the correlation between drought and the satellite-derived indices was examined by using the Pearson correlation technique.

The SPI variations values at 1-month and 3-month scales in the meteorological stations of the study area have been shown in Figure 3.

The results of SPI (Figure 3) for eight synoptic stations in two short periods (1 and 3 months) showed that a dry period occurred in the Lake Urmia basin, and the amount of precipitation was less than average in all corresponding years (2000, 2002, 2004, 2006, 2008, 2010, 2012, 2014, 2016, 2018, and 2020). In terms of specific stations, the meteorological stations in Urmia, Oshnaviyeh, and Tabriz exhibited more fluctuations than other stations in the entire region. Frequent changes in 1-month SPI (1 month) and 3-month SPI (3 months) almost fluctuated similarly between the values of ±2. However, 3-month SPI showed fewer variations in some stations due to its geographic characteristics. Generally, it can be interpreted that 1-month SPI and 3-month SPI identified the almost years as dry-spell over the available time series.

The results of changes in NDWI for different 2-year time intervals have been shown in Figures 4 and 5. The results revealed that the maximum area of Lake Urmia was recorded in 2000 (with an area of 4,950 km²). Meanwhile, the minimum area was measured in 2014 with a value of 1,750 km². According to the results, it was evident that the lake area has continuously decreased from 4,950 km² (2000) to 1,750 km² (2014) and in 2020 increased to 2,850 km².

However, a significant decline exists in the year 2014. The accuracy of the NDWI from MODIS images during the dry period was assessed using the kappa coefficient and overall accuracy measures (O.A) based on Google Earth images and the areas of the lake that the Urmia Lake Rehabilitation Organization has reported. The results of the accuracy assessment are given in Table 3. According to the accuracy values (Table 3), the accuracy of NDWI values varies between 0.86 and 0.95. The highest accuracy belonged to 2006 at the level of 0.95, and the lowest was recorded as 0.86 for the year 2000.

The maps of the NDVI values for different 2-year time intervals have been shown in Figure 6. According to the NDVI maps, the NDVI values in 2000, 2002, 2004, and 2006 were almost the same. The maximum NDVI values were between 0.8 and 1. While, in 2012 and 2014, the NDVI decreased to about 0.75. In other words, the changes in the NDVI in the last years of the study period, especially 2014–2018, have had more and decreasing changes. With an increase in the NDVI values in 2018 and 2020, the NDVI values of the last years reached a maximum value of 0.9.

Table 4 examines the area of each class of NDVI. In this way, it is between −0.2 and 0 class one, 0 and 0.3 class two, 0.3 and 0.5 class three, and 0.5 and 1 fourth class. According to the results, we have had many changes in the study of time series in different classes. As can be seen, in 2000, on the −0.2 to 0 floor, its area was 5,020 km². Also in 2020, the area of the same floor was 2,890 hectares. The highest dense vegetation was in 2000 and 2002, respectively, and the lowest dense vegetation was in 2014 and 2016, respectively.

This index is similar to that in the NDVI, except that it uses additional wavelengths to correct the oscillation error of the radiation angle in the NDVI calculation, which includes atmospheric conditions (distortions in the reflection of light by aerosol particles and signals that are emitted from the ground cover under vegetation) and by considering these factors, this index solves some problems. The reason for this high index in some years and vice versa is low in many hours of the time series due to drought and lack of rainfall in the region, as is evident in the results of the meteorological drought index.

The annual average of EVI values in different time intervals is shown in Figure 7. According to the EVI result, the highest value was observed in 2000 by 0.7, while its lowest value was observed in 2012 and 2014. As mentioned above, changes in the rate of this index have been greatly affected by dry conditions and its values have decreased significantly in 2014.

To obtain dense vegetation in the area using vegetation index, we have set values higher than 0.5 according to Professor Jensen's book as dense vegetation. The values of occupied area by dense vegetation cover for the Urmia Lake basin in consecutive 2-year time intervals have been shown in Figure 8. Obviously, a decreasing trend in the canopy cover was observed from 2000 to 2020. According to the results, dense vegetation was 2,850 km² in the year 2000. In 2012, the area of dense vegetation reached 1,700 km², and consider the amount of vegetation reached the lowest amount of 1,300 km² in 2014. An increasing trend has been observed following 2014.

The reason for using 3 years and bringing the different line coefficients has been the high number of changes this year, also in fact we have mentioned only these 3 years in the article just to show these coefficients. In principle, coefficients have been obtained for all years. Also, Figure 9 is just 2 years as an example. The relationship between the land surface temperature (T_S) and NDVI in the months of the 3 selected years is given in Table 5. According to the first month of 2002, the coefficient and slope line were −12.903 and 8.293, respectively (Figure 9 and Table 5), whereas, in the first month of 2014, the coefficient was equal to −20.172 and 26.985. Furthermore, in the first month of 2020, the coefficient was equal to −15.806, and the slope line was 24.014. According to the obtained line coefficients, it can be said that year 2002 underwent higher humidity than 2014 and 2020.

In the TVDI index, which is based on the amount of temperature and vegetation, we examine the amount of moisture. This index creates a regression relationship between vegetation and temperature to obtain moisture, which decreases with increasing vegetation and vice versa. The maps of the TVDI values during the study period are shown in Figure 10. The maximum soil moisture value was 1 and 0.9 in the years 2000–2006, respectively. In comparison, the lowest value of TVDI was obtained in 2014 with a maximum coefficient of 0.5. From 2008 to 2012, the maximum changes of TVDI were between 0.8 and 0.7. On the other hand, in 2016, 2018, and 2020, the maximum humidity increased compared to 2014 and reached 0.7 and 0.8. Based on the results, the number of changes in soil moisture corresponded to decreasing or increasing. In 2014, the reason for the lack of moisture was the lack of vegetation, rising temperatures, and more droughts. This indicator is in the wet at the bottom of the graph, because when the vegetation is high and the temperature is low, it is wet. These numbers in Table 5 are the same line coefficients in the negative and positive slopes. A full description of these two wet and dry edges is described above.

Figure 11 shows the correlation of NDWI, NDVI, EVI, and TVDI with 1-month SPI values. These inter-relationships can be interpreted as the drought's direct and significant impact on the satellite-derived indices. The correlation coefficients of NDWI, NDVI, EVI, and TVDI with 1-month SPI values were calculated to be 0.85, 0.9, 0.72, and 0.75, respectively, which indicates the meaningful effects of drought on different vegetation and water-related indices in the study area.

Likewise, the relationship of NDVI, NDWI, EVI, and TVDI with 3-month SPI is shown in Figure 12. Accordingly, the correlation of NDWI, NDVI, EVI, and TVDI with 3-month SPI was calculated to be 0.81, 0.88, 0.70, and 0.75, respectively. Most importantly, the correlation analysis of 1-month SPI and 3-month SPI meteorological drought indices showed a strong connection with obtained indices from remote sensing data and satellite imagery. An analogy between 1-month SPI and 3-month SPI indicates that the 1-month SPI had a higher correlation with NDVI, NDWI, EVI, and TVDI in the study area. Moreover, the highest value of correlation coefficient was observed among 1-month SPI and NDVI values.

Figures 13 and 14 show the correlation matrix of 1-month and 3-month SPI values over the meteorological stations in the study area. The presented scatterplot matrices in R quickly provide a sense of the correlations among the studied data.

The correlation coefficient, p-value of the variable ‘i’ and variable ‘j’ for each i–j combination were shown in the graph. Also, the histograms, kernel density overlays, absolute correlations, and significance asterisks (at 0.05, 0.01, and 0.001 levels) can make detailed interpretations.

Figure 15 shows the multiple correlation matrix of NDWI values with overall average values of other independent variables, including 1- and 3-month SPI, NDVI, EVI, and TVDI, based on the data obtained from different periods in the study area.

The Taylor diagram of the relationships of the NDWI with average values of other independent variables, including 1- and 3-month SPI, NDVI, EVI, and TVDI, is presented in Figure 16. The Taylor diagrams are a way of graphically summarizing the closeness of the pattern of some data set. This type of diagram presents concise statistical properties of how well patterns match each other based on their correlation, their root-mean-square (RMS) difference, and the value of their variances (Taylor 2001). Since the plotted values are derived from different timescales, and the different variables, including SPI values, NDWI, NDVI, EVI, and TDVI, have varying numerical values, the results are normalized. The normalized variances ratio indicates the model's relative amplitude and observed variations.

Figure 16 shows the different correlations of remote sensing and meteorological indicators. These indicators, especially the meteorological drought index, which indicates changes in rainfall in the region, clearly show the effects of reduced rainfall on the region, especially water resources and vegetation. Also, the soil surface moisture obtained in the remote sensing index confirms the results of the meteorological drought index. The results show this correlation between remote sensing indices and meteorological drought index, especially the NDWI and the soil surface moisture index.

According to the results, the TVDI values correlated with the NDWI and lower RMS difference values. This indicates that the TVDI can be considered a good measure to detect the variation of NDWI and the Urmia Lake water area. This inference made by people like Shashikant et al. (2021) and Xue & Su (2017) has been reviewed and approved. This finding means that the TVDI is a suitable measure to estimate the soil moisture meaningfully affected by vegetation cover, as reported in different studies. In this regard, Chen et al. (2015) reported that the TVDI could indicate soil moisture conditions under different tree species and different types of land cover from low to high percentage in the Laoshan forest in Nanjing, China. Yang et al. (2017) proved that the MODIS-derived improved TVDI (iTVDI) can be used as a more accurate measure to drought monitoring in Yunnan Province, China. The iTVDI had introduced as a new index to drought awareness in other areas.

Also, the SPI values in 1-month and 3-month timescales are suitable indicators to interpretations on water area variations over the study area. Similar findings have been reported in Kubicz's (2018) study and Wang et al. (2018). According to Kubicz (2018), the evaluation of current water deficits in surface watercourses and groundwaters is complex. The application of SPI to monitor hydrological and hydrogeological drought requires assessing the level of the correlation between drought indices in the study area. Wang et al. (2018), through applying correlation analysis, showed that 3- and 6-month timescales were the most suitable measures for water area response in the Poyang Lake Basin, China. They also pointed out that the lake level variations to meteorological drought provide a reasonable scientific basis for surface water resource management in similar basins. According to the interpretation of the Taylor diagram, the NDVI and EVI having lower correlation coefficients and higher RMS values are the last explanatory variables in the variation of the Urmia Lake level. In the above results, according to research and meteorological drought indicators and remote sensing, 2000 was the wettest year and 2014 was the driest. The numerous droughts and reduced rainfall and water levels in the region in 2014 has caused the current situation.

As a prolonged climatic phenomenon and severe natural disaster, drought occurs almost everywhere and significantly impacts vegetation cover and water bodies. Therefore, this study explored the effects of drought on vegetation changes and water areas (waterbody area) in the Lake Urmia watershed for 20 years (2000–2020). In this regard, the SPI drought index and NDWI, NDVI, EVI, and TVDI were calculated and analyzed. The results of the SPI in two timescales (1 and 3 months) identified the drought years. The NDWI results also emphasized that 2000 and 2014 had experienced the maximum and minimum areas for the waterbody, respectively. Indeed, these changes are due to changes in vegetation cover in the study area. As reflected by NDWI, the lake was at its maximum water area (in 2000), and at the same time, the highest vegetated areas were identified by NDVI and EVI maps.

In contrast, the minimum areas of the lake and vegetation cover were derived in 2014 based on the calculated indices. The soil moisture in the region was also assessed using temperature and vegetation index (TVDI). To a great extent, changes in TVDI values were directly related to the areas of waterbody and canopy with the maximum and minimum values in 2000 and 2014, respectively. The correlation of SPI-1 and SPI-3 with remote sensing indices confirmed that the SPI displayed the most and most negligible correlations with NDVI and EVI, respectively. The higher correlation in SPI-1 rather than SPI-3 indicated the effect of drought on vegetation changes in the study area. It is suggested to use NDVI and NDWI to record and detect the changes in water bodies and canopies for proper monitoring and appropriate decision-making in the region. Comparatively, our outcomes were in agreement with studies by Zhang et al. (2020), Saberi et al. (2018), RifatAhmed & Akter (2017), and Nanzad et al. (2019).

According to results, the TVDI amounts with a higher correlation with the NDWI, and lower RMS difference values are a suitable measure to assess the changes in the water area of the Urmia Lake. The TVDI has reasonably corresponded with soil moisture and also vegetation cover. The 1-month and 3-month SPI values are suitable indicators to predict the water area variations over the study area, which has been proved by the results of the correlation analysis at a significant level. The NDVI and EVI measures are meaningful indices in interpreting variation of the Urmia Lake level. Using multiple indices and satellite images provides a comprehensive quantitative assessment of the relevant variable on the Urmia Lake water area variation. In this study, MODIS data has been used to study the drought of the region that has low spatial accuracy, and it is recommended to use images with high spatial accuracy. In this research, the interval between 2 years and the average of that year has been studied. It is recommended to study the drought as a monthly average. It is also recommended that hybrid drought indicators be used to improve outcomes.

All relevant data are included in the paper or its Supplementary Information.