Spatiotemporal Analysis of Land Surface Temperature and Vegetation Changes in Duhok District, Kurdistan Region, Iraq

Abstract


Introduction
The spatiotemporal distributions of land surface temperature (LST) are crucial in land surface energy studying and vegetation cover on the global and regional scale (Wang et al., 2015).LST data is commonly utilized in vegetation monitoring, as well as evapotranspiration, urban, climate, soil moisture, and environmental sciences (Fayech andTarhouni, 2021, Guha, 2021).The earth's temperature is raised by undergoing various environmental changes including slow and rapid changes due to natural factors and human activities (Balcerzak, 2018, Zimmerman, 2020).
Scientists, planners, and policymakers have been paying attention to rising LST, which has been recognized as one of the most critical worldwide environmental challenges in recent decades (Shivanna, 2020, Tortell, 2020, Watts et al., 2020, Alazawi and Matouq, 2021).United Nations and humanitarian agencies are widely focusing on climate change, especially rising land surface temperature in the world (Campbell et al., 2018).Environmental Justice Foundation (EJF) in the last report mentioned that climate change's consequences led to migrant people around the world.Every year since 2008, roughly 21.7 million people have been relocated as a result of extreme weather conditions (Felli, 2013).
Iraq suffers from temperature raising due to natural factors such as drought, insufficient rainfall, and desertification.Iraq's temperature is rising at a rate that is 2 to 7 times higher than the global average, the minimum temperature (0.48-1.17 °C/decade) is rising faster than the maximum temperature (0.25-1.01 °C/decade) (Salman et al., 2017).As a result, natural vegetation, food production, and land fertility have decreased in Iraq (Mzuri et al., 2021a).The Duhok district in the Kurdistan region of Iraq (KRI) is affected by increasing the LST due to a variety of natural and human factors.
On a worldwide scale, many strategies have been developed to reduce temperature and mitigate its harmful repercussions.An example of such methods increasing the vegetation ratio by planting trees to reduce the air temperature and cool the environment (Mzuri et al., 2021b).The benefits of vegetation are reduced use of energy, improved air quality, and decreased the emission of greenhouse gases.Another method to minimize the air temperature is to reduce vegetation removal or land conversion to protect water resources and preserve natural landscapes (Faour et al., 2016, Wang et al., 2015).Globally, many agreements have been conducted to minimize air temperature.The most well-known is the Paris accord, adopted by 196 parties in 2015.It seeks to clarify the threat of climate change by keeping a global air temperature rise well below 2°C.The accord also intends to employ technology to assist countries in dealing with the climate change effects (Obergassel et al., 2016).
Recently, a conference of parties 26 (COP26) took place in Glasgow and 197 states have been involved.The goal is to cut carbon emissions until they reach net zero in 2050 and to adhere to the Paris agreement by decreasing the 2°C of temperature (Wenger and Firm, 2021).As a result, much research has been conducted around the world to better understand the impact of temperature rise and the factors that drive it (Joshi et al., 2020, Odjugo, 2010, Yihui et al., 2007).
GIS and remote sensing data would be the most appropriate approaches for analyzing the spatiotemporal distribution of temperature and vegetation and understanding their interrelationship.This can be accomplished by utilizing data acquired from satellite images, specifically LST and the Modified Soil Adjusted Vegetation Index (MSAVI2).
The LST, often known as the land's skin temperature, is a key measurement to evaluate surfaceatmosphere interactions and energy transfers between the ground and the atmosphere (Connors et al., 2013, Ibrahim and Rasul, 2017, Urban et al., 2013).LST derived from satellite imagery is frequently used to monitor urban climate and analyze the state of the environment to ensure human sustainability (Aryal et al., 2021).The MSAVI2 is a typical vegetation index, obtained from spectral band reflectance measurements in the red and near-infrared bands (Qi et al., 1994).The MSAVI2 corrects reflections in locations where there is a lot of soil exposed.For vegetative area delineation in bare soil areas, the MSAVI2 index surpasses the NDVI (Mzuri et al., 2021b, Gaznayee andAl-Quraishi, 2020).
Ibrahim and Rasul, 2017 used LST and vegetation index to assess the land surface temperature in Duhok.Vegetation index and LST were used to monitor thermal islands effectors in Ramadi using multitemporal Landsat satellite images (Hurat, 2020).MSAVI2 index was used to identify the spatiotemporal status of drought in Duhok (Gaznayee and Al-Quraishi, 2020).Alemu, 2019 used vegetation index and LST to analyze the spatiotemporal change in the watershed of Andessa in Ethiopia.
The main objectives of this study are to (1) analyze the spatiotemporal relationship and distributions between LST and MSAVI2 in the Duhok district; (2) spatially identify the MSAVI2 and LST distribution within the study area; and (3) to determine the relation between LST and air temperature.To achieve these goals, the article aims to answer the following research questions: (1) what is the relationship between LST and MSAVI2?(2) what is the spatial distribution of LST and MSAVI2 within the Duhok district?(3) what is the relation between LST and air temperature?By achieving these research questions, the extent and distribution of LST and MSAVI2 within the study area have been identified, and setting a useful framework for local government decision-making and remedial action.

Study Area
The study area is the Duhok district, which is located in the northwest of the Kurdistan Region Iraq.It covers an area of approximately 1000 km2 and it lies between latitudes 36°18' -37°20' N, and longitudes 42°20' -44°17' E (Fig. 1) at an altitude 433 to 1512 m above sea level.The study area consists of three subdistricts: Duhok, Zawita, and Mangish.In the last two decades, due to the impact of human activities such as resource extraction and urbanization significant land degradation has been noticed in the Duhok District (Mzuri et al., 2021a).This land degradation affected the growth of vegetation in the study area, especially in lower elevated areas.The soil in the Duhok district is defined as a none saline soil based on its chemical content (Buringh, 1960).The summer between June and September receives little or no rainfall, while the rainy season happens between November and March (Trigo et al., 2010).The maximum amount of precipitation is during December and January, and the temperature reaches the highest degree during July and August.The Mediterranean climate affects the weather of Duhok; the summers are usually dry and hot, with cold, wet winters.The annual precipitation averages between 500 and 1000 millimeters.The average temperature is 19.3°C to 21.2°C, the temperatures in summer ranging from 20°C to 37°C and temperatures in winter ranging from 0°C to 15°C.

Data Sets
Meteorological data, including air temperature, were gathered from 7 distinct weather stations in Duhok over 21 years .The Iraqi Kurdistan Region's General-Directorate of Meteorology and Seismology provided the air temperature data.Landsat satellite images (Landsat-5 TM 2001, Landsat-7 ETM 2011, and Landsat-8 OLI 2021) with a different spatial resolutions and 16 days' intervals were used in this study (table 1).Those images were obtained from the United States Geological Survey (USGS) website.The images that were used were taken in July.They are calculated to assess the LST and MSAVI2, and images are passed through image pre-processing for this purpose.To correct the satellite images from atmospheric noises and radiometric the Fast Line of Sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) function has been applied by using ENVI 5.3 program (Fig. 2).

Spatiotemporal Analysis of MSAVI2
The vegetation index employed in this study was MSAVI2.It was utilized to calculate the study area's spatiotemporal vegetation coverage.MSAVI2 was chosen because it is more effective in defining vegetated areas than other indices such as the enhanced vegetation index (EVI) and normalized difference vegetation index (NDVI), especially in bare soils.As shown in prior studies (Gaznayee andAl-Quraishi, 2020, Mzuri et al., 2021a), In the KRI, where most of the soil is exposed, this index is more significant for vegetation studies.MSAVI2 values range from -1 to +1 and are computed on a pixel basis as shown in equation ( 1) (Qi et al., 1994).
Where ρ is the spectral reflectance of the Red and NIR bands.MSAVI2 value greater than zero indicates an increasing the vegetation trend.Whereas, a value of less than zero indicates a decreasing in the trend of vegetation in the study area.

Digital Number (DN) Conversion to Spectral Radiance (𝑳 𝝀 )
For analysis of the satellite images, spectral data is stored as a digital number (DN) that must be converted to reflectance values.The Digital Number (DN) of a pixel in an image represents a numerical value that describes the brightness of that pixel.Equation (2) was used to convert each pixel of the Landsat satellite images to Spectral Radiance ( λ ) from Digital Numbers (DN) (Tahri et al., 2015).Table 2 represents the values of Landsat parameters.
where, Lλ: spectral radiance, Lmax and Lmin: maximum and minimum radiance detected by the sensor, Qcalmax and Qcalmin: maximum and minimum quantized calibrated pixel value in DN, Qcal: the pixel DN value.

Spectral Radiance Conversion to Brightness Temperature
Thermal band data can be converted to top of atmosphere brightness temperature from spectral radiance by using the thermal constants in the MTL file using equation (3) (Chander et al., 2009, Oguro et al., 2011): where: T: temperature brightness (Kelvin) Lλ: top of atmosphere (TOA) spectral radiance ( /  2 ×  × ) K1 and K2: Band-specific thermal conversion constants obtained from metadata.

Land Surface Temperature from Brightness Temperature
LST was calculated in degrees Celsius using Land Surface Emissivity and the at-satellite brightness temperature (Dang et al., 2020) as shown in Equation ( 4).

Validation of the LST
This part aims to verify the LST values derived from Landsat data.Due to the absence of observed LST data, the results were validated using data from seven meteorological stations.The monthly average temperature of the seven meteorological stations in the study area in July (the satellite images acquisition month) of 2001, 2011, and 2021 were utilized for the results verification.To estimate the spatial variations for air temperature stations in the study area, kriging interpolation method has been applied.kriging is the widely used method for the spatial estimation of air temperature and is influenced by the variability of the distances between the selected station and its neighboring stations (Jo et al., 2018).To find the relation between LST and air temperature, the linear equation was used.By fitting the equation to the observed data, it calculates the relationship between variable (x) and time (y) (equation 5).

𝑦(𝑥) = 𝑎 + 𝑏𝑥 (5)
where a is the intercept of the line and b is the line slope.

MSAVI2 Analysis
The spatiotemporal distribution of MSAVI2 values in the study area from Landsat satellite images is seen in Fig. 3.The values of MSAVI2 ranged from -0.067 to 0.58 in the year 2001, and the values of MSAVI2 for the year 2011 ranged from -0.052 to 0.52 whereas the MSAVI2 values ranged from -0.15 to 0.8 for the year 2021.The highest value of MSAVI2 was noticed on the southeastern, northeastern, and eastern parts of the study area.However, southern, southwestern, and northwestern parts showed low values of MSAVI2.By comparing the temporal distribution of MSAVI2 values at three different times (2001, 2011, and 2021), it is found that the maximum values of MSAVI2 decreased over the study period.The classes of MSAVI2 for the years 2001, 2011, and 2021 are shown in table 3. The comparison of each year's class revealed a significant shift in vegetation cover over the study period of 21 years.As a result, from table 3, around 72.76% in 2001 and 62.48% in 2011 of the study area were covered by MSAVI2 values between 0.1 and 0.2.Whereas, in 2021, 48.24% of the study area was covered by the same MSAVI2 class.These findings demonstrated that there was a dynamic change in vegetation cover from one class to another in the study area.The spatiotemporal changes in vegetation cover are likely to be affected by some external factors, such as land degradation, the intensity of rainfall, and soil erosion in the study area (Mzuri et al., 2021a).This trend is found in the eastern part of the study area in the sub-district of Zawita, where vegetation is abundant, as evidenced by high MSAVI2 values (> 0.3).Low MSAVI2 values (< 0.1) are found in the study area's southern and southwestern regions.

LST Analysis
The spatial pattern of LST for the years 2001, 2011, and 2021 in the study area can be seen in fig. 4. The highest and the lowest temperatures are indicated by the red to the green color scheme in Fig. 5. Also, table 4 represents the LST statistics for the mentioned years.LST readings in 2001LST readings in , 2011LST readings in , and 2021 as highest values were 48.84°C, 51.61°C, and 51.71°C, respectively; while lower values were 26.25°C, 28.33°C, and 27.99°C, respectively.The findings demonstrated that the LST variation in the study area has risen during the last two decades.Whereas LST varied from 48.84°C to 26.25°C with a mean of 41.35°C for the year 2001.The value of LST is varied from 51.71°C to 27.99°C with a mean equal to 44.11°C for the year 2021.The smallest difference between the highest and lowest LST was computed in 2001, while the largest difference was obtained in 2011.
The variability of LST was examined by using the F test for the years 2001, 2011, and 2021.Each pair passed the test significance at the level P<0.01.To demonstrate the extent of LST in the area of study, boxplots have been used and shown in Fig. 5.The year 2011 had the most variability in LST, with a standard deviation of 2.88, followed by 2001 with a standard deviation of 2.69, and 2021 with a standard deviation of 2.71.The average LST increases with positive rates, which is shown in the boxplots.It was reported within 21 years (2001-2021) of the study period, the minimum temperature level has raised 1.74°C with an average of 0.08°C per year.The average temperature has raised 3.06°C with an increasing rate of 0.15°C yr -1 in the specified period.The spatiotemporal variations of LST revealed that the barren lands were demonstrated to the maximum temperature (about 51°C).The barren lands are widely covered in the western, northwestern, and central parts of the study area.
Another factor that led to increased LST is increasing built-up areas, which made the climate warmer, especially during summer.Furthermore, the study area would suffer from the warmer local climate conditions due to a lack of interest in planting and increasing vegetation cover and landscapes in the area.

Analysis of the Relationship between MSAVI2 & LST
The relation between MSAVI2 and LST was estimated by the correlation coefficient method.1522 points were produced using the 50 X 50 fishnet tool in ArcMap software to retrieve values from MSAVI2 and LST raster data.The relationship was calculated for the years 2001, 2011, and 2021 (Table 5).In all three years, the findings demonstrate that MSAVI2 has always a negative relation with LST (Fig. 6).Other previous researchers have also presented similar results (Alemu, 2019, Aryal et al., 2021, Guha, 2021).As a result, places with less vegetation cover have higher LST and inversely.Studies have shown that by lowering the land surface temperature, a vegetated surface can considerably contribute to good quality of life and human relief (Alavipanah et al., 2015, Orimoloye et al., 2018).

Air Temperature and LST Correlation
According to statistical analysis, there is a strong relation between ground-based air temperature and satellite LST.The result showed that the value of r2 is 0.912 which is an indicator that shows the strong relationship between LST and air temperature (Fig. 7).

Discussion
This study has analyzed the spatiotemporal relation between LST and MSAVI2 in three different years in the Duhok district.It depicts the potential of remote sensing and GIS techniques in assessing and monitoring the LST and MSAVI2 throughout the study period.The results showed that the average LST has risen significantly over the last two decades.The mean LST in 2001 was 41.35°C, rising to 44.11°C in 2021, causing an increase of 2.76°C over two decades.The highest observed LST was recorded in 2011, even though the mean LST in 2001 was lower.Retrieving LST from several satellite image data sources improves consistency and minimizes the possibility of getting excessive findings.
Furthermore, the correlation between LST and MSAVI2 variation was investigated using linear regression.The dependent variable was LST, while the independent variables were MSAVI2 and air temperature.The negative and positive coefficients in the equation indicate that the independent variable is involved in the change in LST.The determination coefficient (R 2 ) reflects the model's prediction ability.If the confidence level of the regression models was significant (p < 0.05), the models were considered valid models.
Moreover, the final LST and MSAVI2 maps demonstrated the spatial distribution of LST and vegetation cover (MSAVI2) in the area of study.The results indicated that the east of the Duhok district is covered by dense vegetation, whereas LST is rare.The vegetation cover decreased in the west, north, and central parts of the study area, which led to a raise the LST in this area.However, the LST is also high in the south part of the study area, due to the human activities in this area.
The validation processes for LST maps were obtained by using air temperature data of seven weather stations for July of all three years.Using a linear regression equation, a similar approach was used to validate LST in the Andassa watershed in Ethiopia's Blue Nile Basin (Alemu, 2019).However, if available, it would be ideal to compare our findings to previously created LST maps of the study area.
The limitation of the correlation between LST and vegetation cover (MSAVI2) is varying with season, location, and the type of vegetation.Many efforts have been made to describe this relationship using a variety of biophysical and geographic aspects such as topography, moisture content, and vegetation cover.Solar radiation was found to be the most important element driving the relationship between LST and MSAVI2 during the study period, while other biophysical variables play a lesser role (Karnieli et al., 2010, Wang et al., 2016, Chen et al., 2021).Most previous researchers (Alemu, 2019, Fayech and Tarhouni, 2021, Guha, 2021) have used NDVI and EVI to find the relationship between vegetation and LST.However, in this study MSAVI2 has been used.This is due to the fact that the MSAVI2 index surpasses NDVI for delineation of the vegetation in bare soil areas, where much of the soil is exposed (Gaznayee andAl-Quraishi, 2020, Mzuri et al., 2021a).Moreover, some other indices like NDBI, NDWI, and EVI could also be used with LST to find out the relationship.This aspect could be considered for future work.
Finally, the study's findings on the correlation between LST and vegetation cover in the Duhok district point to an opportunity to reduce LST while increasing vegetation cover by providing the necessary data to local governments and researchers.As a result, the final LST and MSAVI2 maps help urban planners better comprehend future sustainable urban growth and can be used for land development decision-making in this area.Overall, our findings point to a long-term strategy for expanding vegetation in the Duhok district to mitigate future temperature rises.

Conclusions
Remote sensing and GIS are becoming highly relevant in a variety of sectors, including environmental sciences and ecological studies.Three Landsat satellite images from 2001, 2011, and 2021 were used to evaluate the spatiotemporal change of LST and vegetation cover in the Duhok district.The main conclusions are as below: • During the study period, the LST has been raised in the Duhok district.The average LST has been increasing with a rate of 0.15°C yr -1 .• The findings of this analysis also revealed that between 2001 and 2021, there was a decrease in vegetation cover (MSAVI2).Spatially, the areas that human activities are very high, the vegetation declined.• In all three years of the study, correlation analysis revealed a negative relation between LST and MSAVI2.• According to the findings of this study, degradation of vegetation, as well as the intensification and continuous increase of LST, can disrupt and destroy the biosphere and ecology of the Duhok district.• LST north, west, and northwest of the district showed higher LST, according to the findings.The LST was likewise higher in the southern part of the study area, but it was lower near water bodies.Moreover, the minimum LST in 2001 was 26.25 °C, rising to 28.33 °C in 2011, and reaching 51.71 °C in 2021.• Large barren lands and expanding built-up areas resulted in higher LSTs and warmer local temperature conditions, particularly during the summer.

Fig. 7 .
Fig.7.The relation between the mean temperature and LST

Table 1 .
Properties of the used Landsat satellite images

Table 2 .
The parameters values from metadata of Landsat satellite images

Table 3 .
The different classes of MSAVI2 distribution

Table 4 .
LST summary statistics for all three years.

Table 5 .
Regression parameters for LST relation with MSAVI2 in all three years