Assessing the Spatial Pattern of Irrigation Demand under Climate Change in Arid Area

Studying the pattern of agricultural water demand under climate change has great significance for the regional water resources management, especially in arid areas. In this study, the future pattern of the irrigation demand in Hotan Oasis in Xinjiang Uygur Autonomous Region in Northwest China, including Hotan City, Hotan County, Moyu County and Luopu County, was assessed based on the general circulation models (GCMs) and the Surface Energy Balance System model (SEBS). Six different scenarios were used based on the GCMs of BCC_CSM1.1, HadGEM2-ES and MIROC-ESM-CHEM under the Representative Concentration Pathway (RCP) 4.5 and RCP 8.5. The results showed that the method integrating the GCMs and SEBS to predict the spatial pattern was useful. The irrigation demand of Hotan Oasis will increase in 2021–2040. The annual irrigation demand of Hotan City is higher, with 923.2 and 936.2 mm/a in 2021–2030 and 2031–2040, respectively. The other three regions (Hotan County, Moyu County and Luopu County) are lower in the six scenarios. The annual irrigation demand showed a spatial pattern of high in the middle, low in the northwest and southeast under the six scenarios in 2021–2040. The study can provide useful suggestions on the water resources allocation in different regions to protect water resources security in arid areas.


Introduction
Climate change has become one of the most important global environmental challenges in the 21st century. The negative impact of climate change on the water resource system has happened in almost all regions of the world [1,2]. Currently, there are many studies about the impact of climate change on hydrology and water resources in many areas [3][4][5][6][7]. Some previous studies showed that climate change was the main driver of the runoff change, especially in areas with less human activities [8][9][10]. For arid desert areas, agricultural water demand mainly came from the runoff of the river, which was supplied by the snowmelt of glacier. Therefore, studying the pattern of agricultural water demand under climate change is of great significance for water resources management.
At present, there are many studies on the spatial pattern of agricultural water resources. The data of those studies' sources were mainly collected through the Water Resources Bulletin from the government. In addition, the used analytical methods were mainly spatial analysis, which combined mathematical statistics with Geographic Information Systems (GIS) [11]. The spatial analysis of the water resources could not be carried out in the area, where there was no Water Resources Bulletin. In addition,

Study Area and Data
The Hotan Oasis, a part of the Hotan River Basin, located at the southwest of Tarim Basin in the south of Xinjiang Uygur Autonomous Region in China, was studied in the research ( (Figure 1). The Oasis includes the parts of Hotan City, Hotan County, Moyu County and Luopu County. The Hotan River flows for 1127 km with a mean annual water discharge of 4.478 × 109 m 3 . The river passes through the Taklimakan Desert from south to north and then flows into the Tarim River. The river, which comes from the Yulongkasi River and the Kalakasi River, is the second biggest river on the northern slope of the Kunlun Mountains of southern Xinjiang.
The remote sensing data in the study are based on data products of moderate-resolution imaging spectroradiometer (MODIS) mounted on Terra satellite, which was launched on 18 December 1999. The data include that the 8-day, 500-m MODIS surface reflectance product (MOD09A1), the 16-day, 1000-m MODIS land surface temperature product (MOD11A2) and the 16-day, 500-m MODIS Normalized Difference Vegetation Index (NDVI) (MOD13A1). The serial number of the regional grid in this study area is h24v06 with the sinusoidal projection. Elevation data are from the National Aeronautics and Space Administration (NASA) with a spatial resolution of 30 m. In addition, the hourly observation data of air pressure, sea level pressure, temperature, wind speed, sunshine hours, humidity and horizontal visibility at Hotan Meteorological Station are customized from the Hourly Observation Data of China Surface Meteorological Station in National Meteorological Information Center. The remote sensing data in the study are based on data products of moderate-resolution imaging spectroradiometer (MODIS) mounted on Terra satellite, which was launched on December 18,1999. The data include that the 8-day, 500-m MODIS surface reflectance product (MOD09A1), the 16-day, 1000-m MODIS land surface temperature product (MOD11A2) and the 16-day, 500-m MODIS Normalized Difference Vegetation Index (NDVI) (MOD13A1). The serial number of the regional grid in this study area is h24v06 with the sinusoidal projection. Elevation data are from the National Aeronautics and Space Administration (NASA) with a spatial resolution of 30 m. In addition, the hourly observation data of air pressure, sea level pressure, temperature, wind speed, sunshine hours, humidity and horizontal visibility at Hotan Meteorological Station are customized from the Hourly Observation Data of China Surface Meteorological Station in National Meteorological Information Center.

SEBS Model
The SEBS model includes two physical processes: surface radiation balance and surface energy balance. Its basic principle is energy balance: where λE is the turbulent latent heat flux (λ is the latent heat of vaporization and E is water vapour

SEBS Model
The SEBS model includes two physical processes: surface radiation balance and surface energy balance. Its basic principle is energy balance: where λE is the turbulent latent heat flux (λ is the latent heat of vaporization and E is water vapour flux density), W/m 2 ; Rn is net radiation, W/m 2 ; G is the soil heat flux, W/m 2 ; and H is the sensible heat flux, W/m 2 [17]. The SEBS model program was used by the Integrated Land and Water Information System (ILWIS 3.8.5), which integrates remote sensing technology and geographic information systems developed by the Faculty of Geo-information Science and Earth Observation (ITC) in the Netherlands. It has been applied in many studies with good applicability [30]. ILWIS can provide data preprocessing, parameter extraction and model calculation for the SEBS model. Specifically: MODIS data products have been used to calculate the land surface physical parameters, such as land surface albedo, emissivity, NDVI, and land surface temperature, combined with the meteorological data, then put all the data into the model to calculate, and finally generate, the spatial distribution map of daily evapotranspiration.
In order to verify the reliability of the model, the Penman-Monteith model was used based on the previous studies of the verification method of the model [31,32], for that there was no eddy covariance observation systems in the studied area. Although the value of the evapotranspiration which is directly calculated by eddy covariance observation systems is more actual, the value of the water and heat fluxes between vegetation and atmosphere can be measured by the eddy covariance observation systems, and the evapotranspiration can be calculated according to the water balance then. The reference crop evapotranspiration (ET 0 ) was calculated using the Penman-Monteith model recommended by FAO-56, then the actual crop evapotranspiration (ET) obtained by considering the local crop coefficient, which from a previous research [33], was to compare the results calculated using SEBS model.
Currently, there have been a few studies about the annual evapotranspiration pattern from the daily evapotranspiration pattern [34], while the previous research only studied the daily evapotranspiration pattern. In this study, the spatial distribution of annual evapotranspiration is obtained by interpolation and summation method, used the daily evapotranspiration data with 16 days interval. The formula is as follows: where ET is annual evapotranspiration, mm; ET i and ET i+1 are daily evapotranspiration on day i and day i + 1, respectively, mm; ∆T is the interval of days between two adjacent diurnal evapotranspiration (i.e., 16 days); i is the number of the daily evapotranspiration(1, 2, 3, . . . , n).

Prediction of the Future Climate
The fifth phase of the Coupled Model Intercomparison Project (CMIP5) is the most recent coordinated experiment of global climate modeling [35]. There are currently 65 general circulation models (GCMs) from 28 different research institutes, which are widely used in the field of climate change research [36,37]. CMIP5 has four greenhouse gas emission scenarios: Representative Concentration Pathway 2.6, RCP 4.5, RCP 6.0 and RCP 8.5. In this study, RCP 4.5 and RCP 8.5 and three GCMs ( Table 1) that performed well in China were selected to study [38,39]. Therefore, six scenarios (BCC_4.5, BCC_8.5, HADG4.5, HADG8.5, MIROC_4.5 and MIROC_8.5) were studied in this research. This study includes data on meteorological factors for the 21st century (2006-2100) under the RCP4.5 and RCP8.5 scenarios. The direct output results of climate models have large systematic deviations. Therefore, when using GCMs climate model data, downscaling is necessary. A statistical downscaling method can establish the statistical relationship between large-scale climatic factors and regional climatic factors by using multi-year observation data. In addition, this method can test the relationship based on independent observation data. Then, the relationship is applied to the output of large-scale climatic information from GCMs to predict future scenarios of regional climate change (e.g., temperature and precipitation). In this study, the equidistant quantile-based mapping method was used for the statistical downscaling method. This method is a new deviation correction method based on the quantile method proposed by Li [40]. Based on the former observed data, such as precipitation and temperature, at three meteorological stations (i.e., Hotan, Pishan and Yutian stations) in the Hotan River Basin, a statistical downscaling model was established by using an equidistant quantile-based mapping method. In the study area, the future meteorological data was generated from the output of the three GCMs by the downscaling model.

Prediction of Irrigation Demand
In this study, the grids of canal system were included in the farmland in the simulation results, for that the spatial resolution of evapotranspiration calculated by SEBS was 500 m. So, the canals were counted as the farmland and the results of evapotranspiration of the farmland included two aspects: the evapotranspiration of the canal system and the evapotranspiration of actual farmland. While, according to the water balance in Hotan Oasis: the water intake is equal to the sum of evapotranspiration, net percolation of the field and percolation of the canals, the formulas are as follows: where Q is the estimation of irrigation demand from the Hotan River; E is the calculation of evapotranspiration of farm land by SEBS, mm; q 1 is the net percolation in the field (i.e., percolation deducting phreatic evaporation), mm; q 2 is percolation in canal system, mm; k is the proportion coefficient of evapotranspiration to irrigation demand. The irrigation demand under the scenarios in every region in the future can be calculated as follows: where k is the proportion coefficient of evapotranspiration to irrigation demand; k is the variation ranges of the estimated evapotranspiration from present to future; E is the calculation of evapotranspiration of farm land in each region at present by SEBS, mm.

Model Validation
The daily evapotranspiration pixel values of the farmland simulated by SEBS in 2015, which was located at the meteorological station, were compared with those calculated by Penman-Monteith. Simulated and calculated daily evapotranspiration at Hotan station in 2015 were compared in the study, because the other two stations (Pishan and Yutian) were not located in Hotan Oasis. The results are shown in the Figure 2. Figure 2 shows that the simulated values from SEBS are smaller than calculated values from Penman-Monteith on the same day. This is due to other land use types such as roads mixed in farm land pixels (The pixel size is 30 m × 30 m), which results in smaller evaporation values. Through statistical analysis of the simulated and calculated values on the same day, the error range of those values is between 5% and 20%. Therefore, the evapotranspiration estimated by SEBS is reasonable. ISPRS Int. J. Geo-Inf. 2020, 9, x FOR PEER REVIEW 6 of 13  Figure 2 shows that the simulated values from SEBS are smaller than calculated values from Penman-Monteith on the same day. This is due to other land use types such as roads mixed in farm land pixels (The pixel size is 30 m × 30 m), which results in smaller evaporation values. Through statistical analysis of the simulated and calculated values on the same day, the error range of those values is between 5% and 20%. Therefore, the evapotranspiration estimated by SEBS is reasonable.

Spatial Distribution of the Evapotranspiration
The spatial distribution of evapotranspiration in Hotan Oasis in 2015 is shown in Figure 3. The individual white pixels are missing values. The evapotranspiration of Hotan city and Hotan County in the middle of the oasis is higher than that of Moyu County in the northwest and Luopu County in the southwest. Figure 3 shows a spatial pattern of high in the middle, low in the northwest and southeast. According to the type of land use analyzed by ArcGIS 10.2 (Figure 4), the proportion of unused land in Lop and Moyu counties is too large, and the unused land is mainly desert. The annual evapotranspiration of desert area is obviously smaller than that of other land types, which leads to this pattern. The highest evapotranspiration of the pixel is 1448 mm, and the lowest one is 21 mm. In addition, it can be seen from the Figure 3 that the annual evapotranspiration of farm land is high, and the spatial distribution of evapotranspiration in river area is consistent with the shape of rivers.

Spatial Distribution of the Evapotranspiration
The spatial distribution of evapotranspiration in Hotan Oasis in 2015 is shown in Figure 3. The individual white pixels are missing values. The evapotranspiration of Hotan city and Hotan County in the middle of the oasis is higher than that of Moyu County in the northwest and Luopu County in the southwest. Figure 3 shows a spatial pattern of high in the middle, low in the northwest and southeast. According to the type of land use analyzed by ArcGIS 10.2 (Figure 4), the proportion of unused land in Lop and Moyu counties is too large, and the unused land is mainly desert. The annual evapotranspiration of desert area is obviously smaller than that of other land types, which leads to this pattern. The highest evapotranspiration of the pixel is 1448 mm, and the lowest one is 21 mm. In addition, it can be seen from the Figure 3 that the annual evapotranspiration of farm land is high, and the spatial distribution of evapotranspiration in river area is consistent with the shape of rivers.   .

Spatial Pattern of the Evapotranspiration of Farm Land
The evapotranspiration of farm land in different regions in 2015 were calculated by Arcgis 10.2 ( Table 2). In Hotan Oasis, the mean value of evapotranspiration in Hotan City was the highest (901.94 mm/a). This was due to the transpiration of the paddy crop rice planted in Hotan City, which increased the value. On the other hand, the irrigation infrastructure is the best one in all the regions in Hotan Oasis. While, the region is close to Hotan River and has high irrigation guarantee capability. There was little difference among the other three regions for the mean value. Hotan County is the smallest, mainly because the farm land is scattered in the region and there is a low irrigation guarantee capability.

Prediction of Evapotranspiration
The trend of the evapotranspiration of Hotan Station under the six scenarios can be obtained using the Penman-Monteith model in 2020-2090 ( Figure 5), according to the meteorological data of future climate prediction mentioned above. The results show that, in the future, the evapotranspiration will increase under the six scenarios. The increasing trend of MIROC_RCP8.5 is the strongest and the BBC_RCP4.5 is the slowest.

Prediction of Evapotranspiration
The trend of the evapotranspiration of Hotan Station under the six scenarios can be obtained using the Penman-Monteith model in 2020-2090 ( Figure 5), according to the meteorological data of future climate prediction mentioned above. The results show that, in the future, the evapotranspiration will increase under the six scenarios. The increasing trend of MIROC_RCP8.5 is the strongest and the BBC_RCP4.5 is the slowest.

Spatial Pattern of Irrigation Demand
According to the previous studies [41,42], it can be concluded that q 1 and q 2 accounts for 34.67% of the irrigation demand in Hotan Oasis, so the proportion coefficient of evapotranspiration calculated by SEBS to irrigation demand is 65.32% (i.e., k = 65.32%). The dissipative hydrological model for Hotan Oasis [43] was used in previous studies, and every part of water consumption was calculated by the model. Hence

Discussion
A distributed hydrological model can be used to make the spatial pattern of agricultural water demand [44]. However, due to the special geographical environment in arid areas, the distributed hydrological model cannot be used. Hence, this study suggested a new method based on SEBS to address the previous problem. The annual evapotranspiration of the whole Hotan Oasis in 2015 was 866.04 mm calculated by SEBS model, which was close to that in another desert area [26]. The k coefficient was directly calculated, based on previous studies, in this study. The question of whether this coefficient can be obtained by statistical analysis of field experiments also needs further study. In addition, in this study, the county in Hotan Oasis is taken as the research object to study the pattern of irrigation demand. The pattern of irrigation demand in smaller regions such as towns can also be calculated using the suggested method.
The global climate change influence on Xinjiang has been obvious science the 1980s, where the temperature tread has been rising [45], while the trend of temperature was increased especially from the 21st Century in Hotan Oasis [46]. Therefore, the predictions of the evapotranspiration were very likely to happen, while some studies used the GCMs climate model data in Hotan Oasis [47].
In the past decades, the area of farm land in Hotan Oasis has always been increased [48], and the area of farm land will probably increase further with the enhancement of human activities, in the future. In addition, the irrigation demand will increase because of the increase in farm land area. At the same time, the future irrigation demand will also increase under the condition of climate change, according to this study. Therefore, the future agricultural water demand of the oasis will increase significantly. It is necessary for the government to formulate more stringent reclamation plans, as well as to construct more irrigation projects for improving Irrigation Water Utilization Efficiency.

Discussion
A distributed hydrological model can be used to make the spatial pattern of agricultural water demand [44]. However, due to the special geographical environment in arid areas, the distributed hydrological model cannot be used. Hence, this study suggested a new method based on SEBS to address the previous problem. The annual evapotranspiration of the whole Hotan Oasis in 2015 was 866.04 mm calculated by SEBS model, which was close to that in another desert area [26]. The k coefficient was directly calculated, based on previous studies, in this study. The question of whether this coefficient can be obtained by statistical analysis of field experiments also needs further study. In addition, in this study, the county in Hotan Oasis is taken as the research object to study the pattern of irrigation demand. The pattern of irrigation demand in smaller regions such as towns can also be calculated using the suggested method.
The global climate change influence on Xinjiang has been obvious science the 1980s, where the temperature tread has been rising [45], while the trend of temperature was increased especially from the 21st Century in Hotan Oasis [46]. Therefore, the predictions of the evapotranspiration were very likely to happen, while some studies used the GCMs climate model data in Hotan Oasis [47].
In the past decades, the area of farm land in Hotan Oasis has always been increased [48], and the area of farm land will probably increase further with the enhancement of human activities, in the future. In addition, the irrigation demand will increase because of the increase in farm land area. At the same time, the future irrigation demand will also increase under the condition of climate change, according to this study. Therefore, the future agricultural water demand of the oasis will increase significantly. It is necessary for the government to formulate more stringent reclamation plans, as well as to construct more irrigation projects for improving Irrigation Water Utilization Efficiency.

Conclusions
The spatial pattern of irrigation demand in the six scenarios had been predicted using the SEBS model, based on the analysis of the spatial pattern of evapotranspiration in Hotan Oasis. The mean value of evapotranspiration in Hotan City was the highest, which was 901.94 mm/a, while the lowest was Hotan County with 854. 41