Preamble

ARID ZONE RESEARCH Vol. 42 No. 7 Jul. 2025
Spatiotemporal Variation Patterns and Driving Factors of Evapotranspiration in the Aksu River Basin

YANG Chen¹,²,³, MA Bin⁴,⁵,⁶, HE Xuemin¹,²,³, HAO Zhe⁴,⁵,⁶, MA Yu⁴,⁵,⁶

¹College of Ecology and Environment, Xinjiang University, Urumqi, Xinjiang, China
²Key Laboratory of Oasis Ecology of Education Ministry, Xinjiang University, Urumqi, Xinjiang, China
³Xinjiang Jinghe Observation and Research Station of Temperate Desert Ecosystem, Ministry of Education, Jinghe, Xinjiang, China
⁴Urumqi Comprehensive Survey Center on Natural Resources, China Geological Survey, Urumqi, Xinjiang, China
⁵Observation and Research Station of Soil and Water Processes and Ecological Security of Oasis in the Headstream Area of the Tarim River, Urumqi, Xinjiang, China
⁶Field Observation and Research Station of Water Resources and Ecological Effect in Lower Reaches of Tarim River Basin, Urumqi, Xinjiang, China

Abstract

Evapotranspiration, as a critical component of the water cycle, is essential for water resource regulation and ecological protection, particularly in arid regions where it plays a major role in water consumption and redistribution. This study examines the Aksu River Basin using MOD16 evapotranspiration product data to systematically analyze the spatiotemporal variation patterns of actual evapotranspiration (AET) and potential evapotranspiration (PET), and to explore their influencing factors, providing a scientific basis for regional water resource management and ecological conservation.

The results indicate that: (1) MOD16 product data show strong consistency with ET₀ data (R² = 0.8133, p < 0.01), demonstrating that the product accuracy meets the requirements for analyzing evapotranspiration distribution in the Aksu River Basin. (2) The multi-year average AET and PET are 168.36 mm and 1569.03 mm, respectively. AET exhibits an overall increasing trend, while PET shows a decreasing trend. The spatial distributions of AET and PET differ markedly, with opposite trends observed across the region. (3) Over the past 22 years, AET in the Aksu River Basin has increased significantly, primarily in cultivated land, forestland, and oases, whereas PET has decreased overall but increased near oasis edges and river channels. AET demonstrates poor stability while PET remains relatively stable. Both show potential future trend reversals, with 56% of the region exhibiting anti-persistence for AET and 89% for PET. (4) Changes in AET and PET are intrinsically linked to climatic factors, with wind speed and relative humidity identified as the primary drivers influencing regional variations. This research provides an important reference for water resource management and scientific utilization in arid regions.

Keywords: evapotranspiration; water cycle; spatiotemporal variation; water resource management; Aksu River Basin

1. Introduction

1.1 Study Area Overview

The Aksu River Basin is located on the southern slopes of the Tianshan Mountains and the northwestern edge of the Tarim Basin, forming the largest headwater region of the Tarim River. This study focuses primarily on the oasis portion of the basin (Figure 1), covering an area of approximately 1.5 × 10⁴ km² within the geographic range of 78°89′–81°87′E and 40°17′–41°54′N. The region is characterized by a warm temperate extreme continental arid desert climate, with a multi-year average temperature of 10.7°C and annual precipitation of 50.4 mm. Soil types consist primarily of poplar forest soils (Tukay soils), and the dominant vegetation includes poplar, tamarisk, Haloxylon, and reed.

1.2 Data Sources and Processing

1.2.1 Remote Sensing Data

The MOD16 global terrestrial evapotranspiration product includes actual evapotranspiration (AET), latent heat flux (LE), potential evapotranspiration (PET), and potential latent heat flux (PLE). The data are derived using the Penman-Monteith formula based on an algorithm developed in 2005. This study utilized monthly and annual AET and PET data from 2001 to 2022 (MOD16A2). The data were processed using the MODIS Reprojection Tool (MRT) for batch mosaicking, projection conversion, and transformation into WGS-1984 GeoTiff format.

1.2.2 Meteorological Data

Meteorological data from the Aksu, Awati, and Alar stations for the period 2001–2022 were obtained from the National Meteorological Science Data Center and Urumqi Meteorological Bureau. Parameters included daily mean temperature, maximum temperature, minimum temperature, atmospheric pressure, wind speed, relative humidity, and sunshine duration. These data were used to calculate ET₀ values using the Penman-Monteith equation for validation of the remote sensing products.

1.3 Research Methods

1.3.1 Sen+MK Trend Analysis

The Theil-Sen median slope estimation and Mann-Kendall trend test were employed to analyze trends in AET and PET from 2001 to 2022. The Sen slope (β) is calculated as:

$$
\beta = \text{Median}\left(\frac{x_j - x_i}{j - i}\right) \quad \text{for } i < j
$$

where $x_j$ and $x_i$ represent the AET/PET time series values. A positive β indicates an increasing trend, while a negative β indicates a decreasing trend.

The Mann-Kendall test statistic S is defined as:

$$
S = \sum_{i=1}^{n-1}\sum_{j=i+1}^{n} \text{sgn}(x_j - x_i)
$$

where $\text{sgn}$ is the sign function. For large samples (n > 10), S approximates a normal distribution with mean zero and variance:

$$
\text{Var}(S) = \frac{n(n-1)(2n+5)}{18}
$$

The standardized test statistic Z is then calculated as:

$$
Z = \begin{cases}
\frac{S - 1}{\sqrt{\text{Var}(S)}} & \text{if } S > 0 \
0 & \text{if } S = 0 \
\frac{S + 1}{\sqrt{\text{Var}(S)}} & \text{if } S < 0
\end{cases}
$$

A two-tailed test at significance level α = 0.05 was used to determine trend significance (|Z| > 1.96). Trend classification criteria are presented in Table 1.

1.3.2 Coefficient of Variation

The coefficient of variation (Cᵥ) was used to assess the stability of AET and PET spatial patterns:

$$
C_v = \frac{\sigma}{\bar{x}}
$$

where σ is the standard deviation and $\bar{x}$ is the multi-year mean. Variability was classified as: very stable (Cᵥ < 0.1), stable (0.1 ≤ Cᵥ < 0.2), unstable (0.2 ≤ Cᵥ < 0.3), and very unstable (Cᵥ ≥ 0.3).

1.3.3 Hurst Analysis

Hurst analysis was applied to evaluate the persistence of AET and PET trends. For time series ETᵢ (i = 1, 2, ..., n), the mean series is:

$$
\overline{ET}(t) = \frac{1}{t}\sum_{i=1}^{t} ET_i \quad t = 1, 2, \ldots, n
$$

The cumulative deviation is:

$$
R(t) = \max_{1 \leq i \leq t}\left(\sum_{i=1}^{t}(ET_i - \overline{ET}(t))\right) - \min_{1 \leq i \leq t}\left(\sum_{i=1}^{t}(ET_i - \overline{ET}(t))\right)
$$

The standard deviation is:

$$
S(t) = \sqrt{\frac{1}{t}\sum_{i=1}^{t}(ET_i - \overline{ET}(t))^2}
$$

The Hurst exponent (H) is derived from the relationship $R(t)/S(t) = ct^H$. When H > 0.5, the series exhibits persistence; when H = 0.5, it shows randomness; and when H < 0.5, it demonstrates anti-persistence. Future trend classification criteria are shown in Table 2.

2. Results

2.1 MOD16 Data Validation

To assess the applicability of MOD16 data in the Aksu River Basin, ET₀ values were calculated using the Penman-Monteith equation for the Aksu, Awati, and Alar meteorological stations and compared with MOD16-PET data. Since both ET₀ and PET estimate potential evapotranspiration under given meteorological conditions without water limitation, they are directly comparable. The correlation analysis revealed a strong relationship (R² = 0.8133, p < 0.01) between Penman-Monteith simulated ET₀ and MOD16-PET (Figure 2), confirming that MOD16 data are suitable for analyzing spatiotemporal evapotranspiration patterns in the study area.

2.2 Spatiotemporal Characteristics of AET and PET

2.2.1 Interannual Variation

Annual AET fluctuated between 102.44 mm and 228.09 mm, with a multi-year average of 168.36 mm. The most pronounced fluctuations occurred in 2004 (-22.94%) and 2010 (37.99%), reflecting significant variations in water availability. PET ranged from 1480.65 mm to 1659.83 mm, averaging 1569.03 mm, with the largest variations in 2004 (7.55%) and 2010 (-10.79%). The substantial difference between AET and PET indicates that the region experiences severe arid conditions and water scarcity (Figure 3).

2.2.2 Spatial Distribution

The spatial distribution of multi-year average AET and PET shows contrasting patterns (Figure 4). AET decreases from north to south, ranging from 0 to 344.3 mm, while PET increases from north to south, ranging from 1074.79 to 1919.17 mm. High AET values correspond to areas with dense vegetation cover, whereas high PET values occur in sparsely vegetated regions. Blank areas in the MOD16 product represent deserts, water bodies, unused land, and built-up areas where the algorithm cannot retrieve valid data due to limitations in using leaf area index to represent soil moisture conditions.

2.3 Dynamic Change Characteristics

2.3.1 Interannual Trends

Sen+MK trend analysis reveals that AET increased significantly across 94.96% of the basin, primarily in cultivated land, forestland, and oases, while only 3.57% of built-up areas showed significant decreases. Conversely, PET decreased across 83.25% of the region but increased significantly near oasis edges and river channels (9.39%). The area of AET increase far exceeds that of decrease, while PET shows the opposite pattern (Figure 5; Table 3).

2.3.2 Interannual Variability

The coefficient of variation analysis indicates that AET is highly unstable, with unstable and very unstable areas comprising 38.86% of the basin, while very stable areas account for only 0.16%. In contrast, PET exhibits remarkable stability, with very stable areas covering 92.85% of the region and stable areas covering 3.56%, totaling nearly 96.41% of the study area. Unstable areas represent only 0.03% of PET distribution, with no very unstable areas observed (Figure 6; Table 4).

2.3.3 Persistence Characteristics

Hurst analysis shows that AET has a mean H value of 0.44, with 56% of pixels exhibiting anti-persistence (H < 0.5), suggesting future trend reversals. These anti-persistent areas are concentrated in regions with currently significant AET increases, indicating that growth trends may cease or reverse. PET has a mean H value of 0.41, with 89% of the region showing anti-persistence, while only 11% demonstrates persistence (Figure 7).

2.4 Influencing Factors

Correlation analysis between meteorological factors and AET/PET reveals distinct relationships (Table 5). AET is positively correlated with temperature, wind speed, vapor pressure, and sunshine duration, but negatively correlated with relative humidity. Specifically, AET shows a significant positive correlation with wind speed (r = 0.817, p < 0.01) and a significant negative correlation with relative humidity (r = -0.606, p < 0.01). PET, however, is positively correlated with temperature, relative humidity, and sunshine duration, but negatively correlated with wind speed and vapor pressure. Notably, PET exhibits a significant negative correlation with wind speed (r = -0.446, p < 0.05). These results demonstrate that wind speed and relative humidity are the primary drivers of AET and PET variations in the Aksu River Basin.

3. Discussion

As the core hydrological unit of the Tarim River system, the Aksu River Basin provides the most significant perennial water supply to the mainstream and serves as a critical ecological barrier against desertification. Despite previous research on PET using various methods, systematic studies on long-term spatiotemporal dynamics and sustainability predictions remain limited. This study addresses this gap using MOD16 data and meteorological observations.

The increasing AET trend reflects multiple factors, including enhanced snowmelt from global warming and improved water use efficiency through advanced irrigation techniques (e.g., drip irrigation), which reduce water stress during growing seasons. The decreasing PET trend reveals an "evaporation paradox" consistent with findings in other arid regions. Spatially, the contrasting north-south patterns of AET and PET can be explained by the complementary relationship theory: in water-sufficient areas (farmland, oases), AET approaches PET, while in water-limited areas (deserts, gobi), reduced AET leads to increased PET due to enhanced land-atmosphere interactions.

Trend analysis shows significant AET increases in vegetated areas, likely due to improved vegetation cover and ecosystem function, while PET decreases overall but increases near oasis edges, possibly influenced by radiation forcing and humidity changes. The high variability of AET compared to PET's stability reflects the impacts of water resource management policies, irrigation improvements, and land cover changes on actual water consumption, whereas PET remains more constrained by stable climatic conditions.

Hurst analysis reveals anti-persistence for both AET and PET, suggesting that current trends may reverse due to complex interactions among land use changes, water management policies, and climate factors. This instability poses challenges for agricultural production and water resource management. The identification of wind speed and relative humidity as primary drivers enhances understanding of evapotranspiration mechanisms and provides a scientific basis for future water management under climate change.

Future research should focus on the long-term impacts of climate change and human activities, particularly extreme events and land use changes, to provide more precise guidance for ecological protection and water resource management in the Aksu River Basin.

4. Conclusions

This study validated MOD16 evapotranspiration products using the Penman-Monteith equation and analyzed the spatiotemporal variations and trends of AET and PET in the Aksu River Basin from 2001 to 2022, along with their meteorological drivers. The main conclusions are:

1. Data Validation: MOD16-PET data correlate highly with Penman-Monteith simulated ET₀ values (R² = 0.8133, p < 0.01), confirming that MOD16 data are suitable for analyzing evapotranspiration patterns in the Aksu River Basin.

2. Interannual Variation: From 2001 to 2022, AET ranged from 102.44 to 228.09 mm (mean: 168.36 mm) and showed an increasing trend, while PET ranged from 1480.65 to 1659.83 mm (mean: 1569.03 mm) and exhibited a decreasing trend. Spatially, AET is higher in the north (vegetated areas) and lower in the south, whereas PET shows the opposite pattern, being higher in sparsely vegetated southern areas.

3. Dynamic Change Characteristics: Over the past 22 years, AET increased significantly across 94.96% of the basin, concentrated in cultivated land, forestland, and oases. PET decreased across 83.25% of the region but increased near oasis edges and river channels. AET is highly unstable (38.86% unstable or very unstable), while PET is predominantly stable (96.41% stable or very stable). The mean Hurst index for AET is 0.44, indicating anti-persistence in 56% of the region, while PET's mean H value of 0.41 shows anti-persistence in 89% of the area, suggesting potential future trend reversals.

4. Influencing Factors: Wind speed and relative humidity are the primary drivers of AET and PET variations. AET correlates significantly positively with wind speed (r = 0.817, p < 0.01) and negatively with relative humidity (r = -0.606, p < 0.01), while PET correlates significantly negatively with wind speed (r = -0.446, p < 0.05).

References

[1] Zhang Yuan, Jia Zhenzhen, Liu Shaomin, et al. Advances in validation of remotely sensed land surface evapotranspiration[J]. Journal of Remote Sensing, 2020, 24(8): 975-999.

[2] Li Xiaoyuan, Yu Deyong. Progress on evapotranspiration estimation methods and driving forces in arid and semiarid regions[J]. Arid Zone Research, 2020, 37(1): 26-36.

[3] Bastiaanssen W, Pelgrum H, Wang J, et al. A remote sensing surface energy balance algorithm for land (SEBAL): Part 2: Validation[J]. Journal of Hydrology, 1998, 212-213(1-4): 198-212.

[4] Jiang L, Islam S. A methodology for estimation of surface evapotranspiration over large areas using remote sensing observations[J]. Geophysical Research Letters, 1999, 26(17): 2773-2776.

[5] Rodell M, Houser P R, Jambor U, et al. Land data assimilation systems[J]. Bulletin of the American Meteorological Society, 2004, 85(3): 381-394.

[6] Martens B, Miralles D G, Lievens H, et al. GLEAM v3: Satellite based land evaporation and root zone soil moisture[J]. Geoscientific Model Development Discussions, 2016, 10(5): 1-36.

[7] Mu Q Z, Zhao M, Steven W. Improvements to a MODIS global terrestrial evapotranspiration algorithm[J]. Remote Sensing of Environment, 2011, 115(8): 1781-1800.

[8] Fan Jianzhong, Li Dengke, Gao Maosheng. Spatiotemporal variations of evapotranspiration in Shaanxi Province using MOD16 products[J]. Ecology and Environmental Sciences, 2014, 23(9): 1536-1543.

[9] Song Shuo, Zhao Wanning, Li Shaoran, et al. Spatiotemporal characteristics and influencing factors of evapotranspiration in Yinchuan City of northwestern China based on MOD16[J]. Journal of Beijing Forestry University, 2024, 46(7): 18-26.

[10] Li Qing, Yang Pengnian, Peng Liang, et al. Study of the variation trend of evapotranspiration in the Yanqi Basin based on MOD16 data[J]. Arid Zone Research, 2021, 38(2): 351-358.

[11] Liu Jing, Liu Tiejun, Du Xiaofeng, et al. Simulation on spatiotemporal stability of evapotranspiration based on MOD16A2 in Mu Us Sandy Land[J]. Agricultural Research in the Arid Areas, 2020, 38(2): 243-250.

[12] Ailiya Ainiwaer, Yumsiti Halike, Maierdang Keyimu, et al. Spatiotemporal variation of evapotranspiration in the Tarim River Basin by using MOD16 products[J]. China Rural Water and Hydropower, 2018, 60(9): 79-84, 95.

[13] Kang Ligang, Cao Shengkui, Cao Guangchao, et al. Temporal and spatial changes of evapotranspiration in the Shaliu River Basin of Qinghai Lake[J]. Arid Zone Research, 2023, 40(3): 358-372.

[14] Yan Yuhui, Xue Baolin, Zhang Lufang, et al. Temporal and spatial distribution characteristics of evapotranspiration in the Heihe River Basin based on MOD16 product[J]. Water Saving Irrigation, 2019, 44(9): 85-92.

[15] Deng Xingyao, Liu Yang, Liu Zhihui, et al. Temporal-spatial dynamic change characteristics of evapotranspiration in arid region of Northwest China[J]. Acta Ecologica Sinica, 2017, 37(9): 2994-3008.

[16] Yao Xiaochen, Gao Fan, Han Fanghong, et al. Land use intensity change and its influence on evapotranspiration in Aksu River Basin from 2000 to 2020[J]. Arid Zone Research, 2024, 41(6): 951-963.

[17] Chen Yaning, Hao Xingming, Chen Yapeng, et al. Study on water system connectivity and ecological protection countermeasures of Tarim River Basin in Xinjiang[J]. Bulletin of the Chinese Academy of Sciences, 2019, 34(10): 1156-1164.

[18] Han Lu, Chen Jiali, Wang Jiaqiang, et al. Species composition, community structure, and floristic characteristics of desert riparian forest community along the mainstream of Tarim River[J]. Plant Science Journal, 2019, 37(3): 324-336.

[19] Zhang Shouhong, Liu Suxia, Mo Xingguo, et al. Assessing the impact of climate change on reference evapotranspiration in Aksu River Basin[J]. Acta Geographica Sinica, 2010, 65(11): 1363-1370.

[20] Wang Zhicheng, Fang Gonghuan, Zhang Hui, et al. Sensitivity analysis of crop water requirement to meteorological factors in Aksu Irrigation Area[J]. Desert and Oasis Meteorology, 2018, 12(3): 33-39.

[21] Luo Kaisheng, Tao Fulu. Hydrological modeling based on SWAT in arid Northwest China: A case study in Linze County[J]. Acta Ecologica Sinica, 2018, 38(23): 8593-8603.

[22] Wang Yining, Yang Miao, Wang Bing, et al. The evaporation paradox in Wudaogou area and its underlying mechanisms[J]. Journal of Irrigation and Drainage, 2020, 39(3): 126-133.

[23] Bouchet R J. Evapotranspiration réelle et potentielle, signification climatique[J]. International Association of Hydrological Sciences, 1963, 62: 134-142.

[24] Wang L, Wang J J, Ding J L, et al. Estimation and spatiotemporal evolution analysis of actual evapotranspiration in Turpan and Hami Cities based on multi-source data[J]. Remote Sensing, 2023, 15(10): 2565.

[25] Zhang Qiaofeng, Liu Guixiang, Yu Hongbo, et al. Temporal and spatial dynamic of evapotranspiration based on MOD16A2 in recent fourteen years in Xilingol Steppe[J]. Acta Agrestia Sinica, 2016, 24(2): 286-293.

[26] Wu Hongyue, Du Lingtong, Qiao Chenglong, et al. Water consumption of ecosystem driven by evapotranspiration evolution in Ningxia oasis plain[J]. Journal of Soil and Water Conservation, 2023, 37(3): 172-180, 189.

[27] Allen R, Pereira L, Raes D, et al. Crop Evapotranspiration: Guidelines for Computing Crop Water Requirements[M]. Irrigation and Drainage Paper No. 56, FAO, 1998.

[28] Yu Wenjun, Zhao Lin, Li Yanzhong, et al. Spatial-temporal variation of evapotranspiration based on the complementary relationship principle and its influencing factors on the Qinghai-Tibet Plateau[J]. Acta Ecologica Sinica, 2024, 44(12): 5024-5039.

[29] Yin Xiaowei, Wu Yiping, Zhao Wenzhi, et al. Drought characteristics and sensitivity of potential evapotranspiration to climatic factors in the arid and semi-arid areas of Northwest China[J]. Hydrogeology & Engineering Geology, 2021, 48(3): 20-30.

[30] Ma Yali, Niu Zuirong, Sun Dongyuan. Relationship between changes in spatial and temporal patterns of potential evapotranspiration and meteorological factors in the Hexi Corridor[J]. Arid Land Geography, 2024, 47(2): 192-202.

[31] Chahine M T. The hydrological cycle and its influence on climate[J]. Nature, 1992, 359(6394): 373-380.

[32] Moran M S, Rahman A F, Washburne J C, et al. Combining the Penman-Monteith equation with measurements of surface temperature and reflectance to estimate evaporation rates of semiarid grassland[J]. Agricultural and Forest Meteorology, 1996, 80(2-4): 87-105.

[33] Qiu Xinfa, Zeng Yan, Liu Changming. A study on actual evaporation from non-saturated surfaces[J]. Progress in Geography, 2003, 22(2): 118-124.

[34] Cao Xue, Ke Changqing. Classification of high-resolution remote sensing images using object-oriented method[J]. Remote Sensing Information, 2006, 21(5): 27-30, 51, 73.

[35] Liu Yang, Yu Entao, Yang Jianjun, et al. Characteristics of spatial and temporal variation of actual evapotranspiration in the arid region of Northwest China from 1960 to 2019[J]. Research of Soil and Water Conservation, 2021, 28(6): 75-80, 89.

[36] Zhou Yanzhao, Zhou Jian, Li Yan, et al. Simulating the evapotranspiration with SEBAL and modified SEBAL (M-SEBAL) models over the desert and oasis of the middle reaches of the Heihe River[J]. Journal of Glaciology and Geocryology, 2014, 36(6): 1526-1537.

[37] Adilai Wufu, Yusupujiang Rusuli, Reyila Kader, et al. Spatiotemporal distribution and evolution trend of evapotranspiration in Xinjiang based on MOD16 data[J]. Geographical Research, 2017, 36(7): 1245-1256.

[38] Wang Ranran, Lv Guanghui, He Xuemin, et al. Spatial-temporal evolution and driving factors analysis of oasis evapotranspiration in the middle reaches of the Keriya River Basin[J]. Bulletin of Surveying and Mapping, 2024, 70(3): 31-36, 139.

[39] Song Jia, Xu Changchun, Yang Yuanyuan, et al. Temporal and spatial variation characteristics of evapotranspiration and dry climate in Xinjiang based on MODIS16[J]. Research of Soil and Water Conservation, 2019, 26(5): 210-214, 221, 2.

[40] Guo Xiaotong, Meng Dan, Jiang Bowu, et al. Spatiotemporal change and influencing factors of evapotranspiration in the Huaihe River Basin based on MODIS evapotranspiration data[J]. Hydrogeology & Engineering Geology, 2021, 48(3): 45-52.