Observed Impacts of Climate Variability on LULC in the Mesopotamia Region

: Remote sensing analysis techniques have been investigated extensively, represented by a critical vision, and are used to advance our understanding of the impacts of climate change and variability on the environment. This study aims to find a means of analysis that relies on remote sensing techniques to demonstrate the effects of observed climate variability on Land Use and Land Cover (LULC) of the Mesopotamia region, defined as a historical region located in the Middle East. This study employed the combined analysis of the Normalized Difference Vegetation Index (NDVI), Land Surface Temperature (LST), and two statistical analysis methods (Pearson Correlation Analysis, r ; Coefficient of Determination, R 2 ), which were applied using the Moderate Resolution Imaging Spectroradiometer data and observed surface meteorological data from 2000 to 2018. The resulting NDVI images show five LULC classes with NDVI values varying between − 0.3 and 0.9. Furthermore, changes in the classified LULC area were compared statistically to those in NDVI values, where a positive relationship was found. Also, when the LST values and temperature are more extreme, the NDVI values were found to be smaller, suggesting a decrease in the density of vegetation cover. A negative correlation was found through Pearson correlation analysis ( r = ∼− 0.64), indicating a direct effect of increased temperatures on LULC. Indeed, this negative relationship between NDVI and LST was proven using R 2 values, where a two-dimensional scatter plot analysis showed that R 2 ranges from 0.54 to 0.9. Ultimately, the results obtained from this study reveal changes that may have many prominent effects in the field of LULC classification, accelerating the implications of climate change and variability factors.

Many studies have attempted to use remote sensing data to address changes in LULC, some of which have focused on semi-arid and arid regions. Post-Classification Comparison (PCC), image differencing, principal components analysis, and vegetation indices are the most common techniques that have been adopted to monitor LULC changes using remotely sensed data [20]. Othman et al. [21] were able to classify LULC changes using data from the Moderate Resolution Imaging Spectroradiometer (MODIS) where it has been shown that the use of MODIS-Normalized Difference Vegetation Index (NDVI) produced good results when revealing LULC classes [21]. To minimize problems associated with multi-temporal images recorded under different atmospheric and environmental conditions, the PCC method [22] has been used to classify and detect other LULC classes using multi-temporal remote sensing data [23]. Land Surface Temperature (LST) is an essential parameter that is often used to describe physical processes ongoing in land surface classes on regional and global scales. It combines the results of all interactive operations between the land surface and atmosphere, such as the exchange of energy fluxes between water bodies and the atmosphere [24]. Land cover change analysis data based on LST retrieved from remote sensing products have been used in many studies related to drought monitoring and soil salinization. Many researchers have been using LST products in their studies. For instance, LST data have been used to estimate the effects of long-term climate change, especially when integrated with NDVI data used to monitor vegetation coverage. All these studies have proven that LST can be a good indicator of many climate-related changes [25][26][27]. In addition, MODIS-NDVI data were used to explore whether differences existed in Spain and the Balearic Islands through a temporal statistical trend analysis of vegetation cover. The analysis showed that the NDVI trends were negative or positive at different times [28]. As an extremely beneficial technique, a geographical information system plays a significant role in detecting changes in LULC fields. Its high-power capability can provide quantitative estimation, statistical analysis, and spatial distribution solutions for a variety of case studies related to LULC [29,30]. Distribution analysis applications have been used to produce sensitive models that can be used to analyze the location of LULC classes on the ground.
Remarkably, the present study used a particular combination of remote sensing data to find an optimal analysis procedure that could be used to allocate climate change effects on the LULC in the study area. In addition, this study aims to address the factors that cause these changes and to find possible land management solutions or proposals to limit these changes from an environmental perspective. Furthermore, the following goals are addressed to achieve the study objectives: (i) to provide recent and historical LULC maps for the Mesopotamia region and to reveal the changes in LULC cover classes; ii) to identify the effects of observed longterm climatic variability on LULC using a developmental procedure that depended on LST and statistical analyses.

Study Area
The historical land of the Middle East, situated between the Tigris and Euphrates river basins, is called Mesopotamia. Agriculture first emerged in Mesopotamia and then spread to neighboring lands, such as Ancient Egypt. Early agricultural operations enabled the residents of Mesopotamia to settle in the area and abandon their previously nomadic lives [31,32]. It is located from 30 • -39 • N to 40 • 00 -50 • 54 E covering an area of about 496,290 km 2 , including much of Iraq and the eastern parts of Syria as well as small portions of western Turkey and Iran (Fig. 1). Sand and dust storm (SDS) events create challenging environmental phenomenon for humans in the Mesopotamia region, where there has been an increase in SDSs with varying levels of intensity and frequency, especially in the past decade [2,33]. In addition, LULC changes have been occurring in this region primarily because of SDSs. These environmental catastrophes have been described in many international forums dealing with climate change phenomena and are categorized as uncontrollable transboundary environmental phenomena that cause a great deal of damage to human property and lives [34,35]. Al-Hayani et al. [36] concluded that SDSs contains various microorganisms, including Mycobacterium, Brucella, Bacillus, and many others. Furthermore, SDSs have a major economic effects. For example, a recent study by Al-Hayani et al. [37] showed that dust events in Kuwait City in 2001-2014 have caused $28,180 worth in damage per day to the oil export sector.

Remote Sensing Data
In recent years, remote sensing has become an impressive technique based on its robust methods and reliable products. The MODIS on board Terra, an Aqua platform available for largescale terrestrial applications, has 36 spectral bands ranging in wavelength from 0.4 to 14.4 µm with varying spatial resolutions having two bands at 250 m, five bands at 500 m, and 29 bands at 1 km. The MODIS product data are distributed in the Hierarchical Data Format (HDF), and each file contains multiple layers of data. Out of seven bands used for land-surface applications (bands 1 to 7) from a total of 36 bands, only bands 1 and 2 (representing the red and near-infrared parts of the spectrum) were used in this NDVI analysis and study; cloud-free images were selected. Furthermore, of the 36 bands, only six bands deal with land surface temperature/emissivity (Bands 20 to 23 and Bands 31 to 32). In this study, bands 31 and 32 were used for the analysis of LST [38]. Because the study area covers an extensive region, 76 satellite images with two types of MODIS data acquired in July were retrieved from the United State Geological Survey (USGS) website representing from 2000 to 2018 [39]. The downloaded images included 36 images of NDVI type MOD13Q1 and 38 images of LST type MOD11A1 Version 6, where every two images cover the entire study area. A subset and mosaic process was applied to produce 19 scenes of each image type in the study area. These images have specific information, as described below: 1) MOD13Q1 data provides two essential vegetation layers generated every 16 days as a Level 3 product at a 250 m spatial resolution. The band's width fluctuated between 620 and 876 µm with spectral radiance at 21.8 and 24.7 W/m 2 -µm-sr for bands 1 and 2, respectively. 2) Authorized validation of MOD11A1 Version 6 was obtained by the Terra satellites by various in situ measurements [40]. These data can be described as having a daily temporal resolution generalized by a split-window algorithm under clear-sky conditions at about 1 km (actually 0.928 km) spatial resolution with nominal data array dimensions equal to 1200 rows by 1200 columns gridded in a sinusoidal map projection. The band's width fluctuated between 10.78 and 12.27 µm with spectral radiance at 9.55 and 8.94 W/m 2 -µm-sr) for bands 31 and 32, respectively. 3) In the temporal evaluation of MODIS-NDVI results, Landsat 5 Thematic Mapper, Landsat 7 Enhanced Thematic Mapper+, and Landsat 8 Operational Land Imager were taken simultaneously. Landsat satellite images have 30 m spatial resolution with a 16-18 days temporal resolution, and associated spectral bands had wavelength differences between 0.433 and 2.35 nm. Landsat images feature a higher spatial and temporal resolution than MODIS images and provide high-quality multispectral data for land surfaces since 1972.

Normalized Difference Vegetation Index (NDVI)
The NDVI is the most extensively used remotely sensed vegetation index and is used for detecting and investigating LULC and climatic phenomena, such as drought and desertification. For LULC studies, NDVI supplies a measure of the vegetation cover, associated with standing biomass volume. The main challenge related to using NDVI is that, even though biomass and productiveness are intimately related in some systems, the consequences of NDVI conditions can be generally based on changes of vegetation cover across LULC and ecosystem types [43]. The NDVI is a quantified vegetation remote sensing index that measures the differences between the red and near-infrared spectra reflected or observed from the vegetation cover. It is mostly used to discriminate between vegetated and unvegetated areas; it can also be used to evaluate the health status of vegetation [44]. The NDVI is represented mathematically using Eq. (1) as follows: where NIR and RED reflect the near-infrared and red range of the spectrum, respectively. This formula generates a value between −1 and +1. The values from −1 to 0 represent unvegetated areas such as naturally bare, abandoned, or urbanized land, and values greater than 0 to +1 represent existing vegetation in the area under study [45]. In other words, a high NDVI value is observed when low reflectance is observed in the red channel and high reflectance is observed in the NIR channel, and vice versa. Based on the example provided by [46] and using MODIS13Q1 products, in the present study the NDVI was extracted from each image through multi-layer MODIS data for each frame, and the mosaic process of two frames was applied for each year (Fig. 2). From NDVI results validated via Landsat satellite imagery, the final NDVI images and LULC were classified into five classes, where the NDVI range of values for each class was estimated and was used in the statistical analysis of this study.

Retrieving LST from the MODIS Data
The incident solar radiation received on earth is the ultimate energy source for plant life and largely determines the climatic conditions of earth's habitats. The amount of solar energy reaching the earth's surface is a significant observation related to the surface energy balance and governs many diverse surface processes. The LST provides a good indicator of long-wave and incoming solar irradiation that affects all the LULC classes. When soil, vegetation, water, or urban area, and so on have a different response for incoming solar radiation, any changes in the local surface temperature can be detected [47]. Several methods can be used to retrieve LST-based remote sensing data such as the split-window, temperature/emissivity separation, or single-channel methods [48,49]. In the present study, LST was retrieved automatically, where several steps had to be completed based on the single-channel method, as presented by [47]. Here, LST type (MOD11A1 Version 6) data from 2000 to 2018 (July) that covered the entire study area with a mosaic subset was adopted to obtain the final scene in the study area. The LST data were extracted in degrees Kelvin; the LST units obtained were converted to degrees Celsius using Eq. (2): where DN represents the digital numbers of LST data (Fig. 3) [50].

Statistical Analysis
The histograms of all NDVI and LST images were divided into 10 break values (Fig. 4). Two-dimensional scatter plot analysis was prepared to establish the relationship among the three variables: NDVI, LST, and the temperature distribution. Standardized statistical analyses were carried out to show the relationship between these variables with LULC changes in the study area during the studied period [51]. Two-dimensional scatter plot analysis and Pearson Correlation Coefficient (r) analysis was applied to find the statistical relationship between NDVI and LST for each LULC class.

Results and Discussion
According to [50], the region of Mesopotamia has suffered from two major problems, increasing salinization and dune movements, which play a leading role in decreasing quality and availability of agricultural lands. These problems are the main results of climate change in the study area based on extensive evidence that the findings of this study have confirmed. The NDVI images at a 16-day composite were produced by MODIS13Q1 products covering the study area from 2000 to 2018. The NDVI levels were estimated using the NDVI rescale method to improve visual interpretation. As a result, it was found that the NDVI values ranged from −0.3 to 0.9. The highest value (0.9) represents vegetation, whereas the lowest value (−0.3) represents water; the NDVI of barren lands fall in between (Fig. 2). The NDVI was used to analyze the four main LULC classes, including vegetation, agricultural lands, barren lands, and water. The areas of each class and the NDVI value for each year are summarized. According to Tab. 1, the region is generally covered by barren land (∼80%), open water (2%), and agricultural areas (∼7%). Although no significant changes were observed in the overall assessment of the area during the study period, changes were especially observed in open water areas when these areas are examined in detail (Fig. 5). In addition, the most noticeable differences can be seen in the vegetation where vegetation cover varied between 4% and 11%. To visualize the relationship between LULC and NDVI in the study area, the changes in each class area were estimated as a percentage for each studied year (Fig. 6). Therefore, changes in LULC classes can be defined. These changes in the region can be compared, evaluated, and attributed to several factors, such as the effects of climate change as a connection to the retreat of the classes of water bodies. The increase in the barren land class that has been caused by human activity is an example of an increase in the class of the agricultural land area.  Fig. 7. Note that all parameters are in a relationship that binds them to a specific format. When the LST and temperature are high, the NDVI values are low, indicating a decrease has occurred in the density of the vegetation cover as was observed in the period from 2003 to 2012. However, this also occurred in 2014, where this decrease in the density of the vegetation cover proves that climate change has been exacerbated and as a result caused desertification in the study area. Conversely, an increase in NDVI values can be observed when the LST and temperature decrease as seen during the period 2000 to 2003 and for 2012, 2016, and 2018, suggesting that the density of vegetation cover had increased. This may have been caused by the high rates of rainfall in those years or by the increase in the agricultural area used by humans. This analysis method can be considered a suitable and straightforward solution for the assessment of the effects of various drought factors in the region.  2000  8279  2  403428  81  35945  7  48645  10  496299  2001  9944  2  399345  80  37634  8  49377  10  496300  2002  10720  2  375527  76  55113  11  54938  11  496300  2003  9986  2  401673  81  30972  6  53668  11  496300  2004  11101  2  395304  80  31773  6  58120  12  496299  2005  10024  2  387599  78  41714  8  56964  11  496300  2006  10332  2  386633  78  43475  9  55859  11  496299  2007  10332  2  397580  80  32528  7  55859  11  496299  2008  8274  2  402711  81  30877  6  54437  11  496299  2009  8275  2  403428  80  35946  7  48649  11  496300  2010  9369  2  396073  80  32879  7  57978  12  496299  2011  8882  2  395353  80  32686  7  59379  12  496299  2012  8645  2  399739  81  31052  7  56866  10  496300  2013  9986  2  397640  80  31760  6  56912  12  496299  2014  10084  2  412250  83  20561  4  53404  11  496299  2015  8914  2  411015  83  20341  4  56029  11  496299  2016  8275  2  403428  80  35942  7  48645  11  496299  2017  9944  2  399345  80  37634  7  49377  11  496300  2018  10720  2  375527  80  55113  7  54938 11 496300  As mentioned in many articles, water bodies and vegetation cover are often influenced by surface temperatures where this relationship can be simulated as a linear relationship [3,6,7,28,51]. In the present study, the total number of the metadata matrices of the NDVI and LST images were used to study the relationship between NDVI and LST to find the correlation between these two for the LULC classes (Fig. 8). This analysis shows a negative correlation existed between NDVI and LST for all LULC classes in the study area. In other words, the increase in surface temperature adversely affects the other LULC classes. The strength of the obtained negative correlation between the NDVI and the LST was analyzed using analyzing the Pearson correlation coefficient (r). This correlation varies between −1 (representing a perfect negative correlation), zero (represents no linear correlation), and +1 (representing a perfect positive correlation) [52], where the average between the NDVI and the LST Pearson Correlation Coefficient (r) was −0.64, and this proved a negative correlation existed between them. The Coefficient of Determination (R 2 ) showed an excellent fitting negative relationship between the LST and the NDVI for all classes ranging from 0.54 to 0.9. Fluctuation in LST values indicates that these values directly affect the decrease or increase in LULC classes. Previous studies have analyzed the effects of climate change on LULC in some parts of Mesopotamia, as mentioned by [53] where some changes in the Mesopotamian marshland occurred during the period 1988-2017; the authors found that dramatic changes in vegetation cover occurred. As a large part of the region of Mesopotamia, all parts of Iraq suffer from major effects of climate change. The spatial extent of the desert area in the western part of Iraq increased to 2,287,000 km 2 between 1990 and 2017 (studied period), as mentioned by [54]. Furthermore, Yahya et al. [5] found a reduction in annual runoff values in the studied period from 1992 to 2017, where the drought satiation fluctuated between moderate to extreme drought and increased in the desertified area of northwestern Iraq. According to the above, only some parts of the study area and corresponding study subject were analyzed. Our research provides a new vision of how climate change affects the entire study area through the interplay of several remote sensing and statistical analysis methods, which have been efficient in finding potential relationships between the variables under study and in achieving the primary goal of the study.

Conclusions
The complicated situation regarding climate variability in the region of Mesopotamia must be studied with caution. The variable changes in temperature, precipitation, and increase in SDS events should not be underestimated; this change in climate behavior would largely eliminate some species or change the composition of a combination of other species present in the study area. In the present study, MODIS products that have been shown to be efficient to many environmental applications, such as the detection of vegetation change, were used to assess the relationship between the LULC classes and the possibility of climate change occurring over time. Statistical analyses were done on the LST and NDVI to find any possible connection between increasing surface temperature and decreasing vegetation cover, reflecting the likelihood of increasing drought factors affecting the study area over time. Based on the present study's results, the LULC categories should be monitored with caution, especially concerning the increase and decrease in the area of these classes over time. The use of higher resolution images would allow a more detailed analysis of the changes. Periodic monitoring of these classes will help land managers to develop the study area in several ways and reduce the opportunity for the occurrence of direct negative effects of erosion-related factors by increasing the spatial extent of vegetation cover at the expense of the barren land class, and thus, reduce the effects of potential desertification factors in the study area. Immediate solutions must be developed in the region of Mesopotamia in the hope of reducing this phenomenon, because the cultivation of windbreaks requires planting lines of unique carefully selected species of fast-growing trees in areas corresponding to prevailing surface winds. It is hoped that these windbreaks would stop the movement of dunes and reduce the risk of sand and dust storms events, especially in the western and southwestern parts of Mesopotamia.