Simulation of Actual Evapotranspiration and Its Multiple-Timescale Attribution Analysis in the Upper Reaches of the Jinsha River, China

Abstract

For quantifying the contribution rates of climatic variation and anthropogenic activities on the actual evapotranspiration at multiple timescales in the URJR, based on the monthly meteorological and hydrological data of the URJR, this study first used the BG mutation method to determine mutation years of runoff depth to divide the research period into base and variation periods. Then, the ABCD hydrological model was used to simulate the runoff variation process during the base period and the variation period, and the actual evapotranspiration data at the monthly scale was simulated. Finally, a multiple-timescale Budyko model was applied for quantitatively computing the impacts of climatic variation and anthropogenic activities on multiple-timescale actual evapotranspiration in the URJR. The results demonstrated the following: (1) The mutation years of runoff depth at the Batang and Shigu hydrological stations were 1988 and 1987. (2) The actual evapotranspiration at multiple timescales (quarterly and monthly) in the upper reaches of the Jinsha River all presented a significant increase (p < 0.01), with a growth rate ranging from 0.02 mm/a to 0.22 mm/a. (3) Climatic variation is dominant factor leading to actual evapotranspiration growth at multiple timescales (quarterly and monthly) at the Batang Hydrological Station, with a contribution ranging from 0.71 mm to 8.50 mm. (4) Human activities are dominant factors leading to actual evapotranspiration growth at multiple timescales (quarterly and monthly) at the Shigu hydrological station, with a contribution ranging from 0.60 mm to 9.62 mm.

1. Introduction

Actual evapotranspiration is a key hydrological process that connects soil moisture, vegetation growth, and atmospheric circulation. Climate variation and human activities are driving mechanisms affecting hydrological processes [1,2,3,4]. Global warming can alter factors such as temperature, humidity, and precipitation in the Earth system and thereby affect evapotranspiration [5,6,7]. Human activities have had a significant impact on actual evapotranspiration through various means such as altering land use patterns and water resource management [8,9,10]. In recent years, climate change and human activities, as two core driving forces, have had extremely significant and complex impacts on global and regional actual evapotranspiration, profoundly altering the water and energy balance of the land surface [11,12,13,14]. Actual evapotranspiration variation may affect the productivity and stability of ecosystems through the coupled processes of water, energy, and nutrient cycling [15,16,17,18]. Therefore, accurately assessing the actual evapotranspiration changes and their driving mechanisms is a prerequisite for predicting and responding to future regional ecological risks and formulating scientific adaptive management strategies (such as the scale and layout of vegetation restoration), which are of great significance for protection of the ecological environment in river basins and efficient utilization of water resources.

Given the complexity of the actual evapotranspiration evolution process [19,20], many methods were employed to evaluate and quantify contributions of different variables to actual evapotranspiration changes [21,22,23]. Budyko found that energy and water are the main factors determining actual evapotranspiration [24], which was widely employed for quantitative analysis of the driving factors of actual evapotranspiration [25,26]. Ji et al. [27] simulated actual evapotranspiration in the Ganjiang River and analyzed contributions of different factors to actual evapotranspiration variation using the Budyko model. The results indicated that climate change was the dominant factor leading to the growth in the actual evapotranspiration of Ganjiang. Yu et al. [28] estimated the actual evapotranspiration in the middle reaches of the Heihe River from 2006 to 2015 based on the improved Budyko empirical model and quantitatively analyzed the contributions of different factors to the change in actual evapotranspiration using the elastic coefficient method. The results showed that the change in precipitation was the main cause of the change in actual evapotranspiration. Li et al. [29] explored the influences of various elements on evapotranspiration in the Wei River based on the Budyko framework and found that climate change played a leading role in the alteration of evapotranspiration, while the impacts of human activities gradually increased. Most existing studies have investigated the variation characteristics and attribution analysis of actual evapotranspiration on an annual scale [30,31,32], while few researchers have analyzed the influences of different factors on the changes of actual evapotranspiration at multiple timescales (seasons and months) within a year.

As a key water source area and ecological barrier in the Yangtze River, the water cycle process of the Jinsha River, especially the variation characteristics of actual evapotranspiration (ET), directly affects the total amount and spatio-temporal distribution of water resources in the lower reaches of the Yangtze River. Under the dual background of global warming and the intensification of regional human activities, the water cycle process in the Jinsha River is undergoing complex and profound changes. Therefore, many scholars have analyzed the water cycle process and its influencing factors in the Jinsha River [33,34]. Zhang et al. assessed the contributions of climate change and human activities to runoff variation of four subregions in the Jinsha River by the Budyko-based elasticity method [35]. Wang et al. applied the relative importance analysis method to analyze the causes of runoff changes from the perspectives of early runoff, rainfall, snowfall, evapotranspiration, and soil water content [36]. Zhang et al. integrated extreme gradient boosting, convolutional neural networks, long short-term memory, and informer to quantify cumulative and offsetting effects of runoff depth response to climatic variation in a large reservoir group in the Jinsha River [37]. Previous studies have focused on analyzing the influencing factors of runoff changes, while there are relatively few studies on the attribution analysis of multiple-timescale (quarterly and monthly) actual evapotranspiration changes in the Jinsha River.

Therefore, taking the upper reaches of the Jinsha River Basin (URJR) as the research area, this study constructed a multi-method coupling framework of BG segmentation, an ABCD hydrological model, and a Budyko model to quantitatively analyze the contribution rates of climate factors and human activities to the actual evapotranspiration at multiple timescales through the following steps: (1) We first used the B-G mutation segmentation algorithm to determine the mutation years of runoff depth to divide the research period into the base period and the variation period. (2) The ABCD hydrological model was used to simulate the runoff variation process during the base period and the variation period, and the actual evapotranspiration data at the monthly scale was simulated. (3) A multi-timescale Budyko model was constructed to quantitatively analyze the impacts of climate change and human activities on the actual evapotranspiration at multiple timescales (quarterly and monthly). This research can provide reference suggestions for the rational allocation and regulation of water resources, agricultural irrigation utilization, and ecological environment protection in the URJR.

2. Research Region and Data

The Jinsha River is located in the upper reaches of the Yangtze River, originating from the Geladandong Snow Mountain in the Tanggula Mountains of Qinghai Province. Its drainage area ranges from 24° N to 36° N and from 90° E to 105° E. The drainage area of the Jinsha River is approximately 387,500 square kilometers, accounting for 26% of the Yangtze River. The terrain of the basin is complex, mainly characterized by high mountains and deep valleys, and there is a significant difference in altitude between the upstream and downstream areas. The distribution of river systems is complex. Rivers are mainly replenished by rainfall, supplemented by groundwater and meltwater. This study takes the basin above Shigu Station of the Jinsha River as the research area, which is called the upper reaches of the Jinsha River (URJR) (Figure 1). The URJR is located in the southeast of the Qinghai-Xizang Plateau, covering an area of 214,000 square kilometers. The main land cover type in the basin is grassland, followed by bare land. The main soil type is sandy loam. The URJR has a semi-humid and semi-arid climate. Due to the blocking effect of the Hengduan Mountains, the precipitation is relatively low. The annual average temperature is 1.5 °C, and the annual average precipitation is 452.6 mm.

The monthly runoff depth data of Batang Hydrological Station and the Shigu Hydrological Station in the URJR from 1967 to 2016 were obtained from the Yangtze River Water Resources Commission. Climate station data of the URJR were obtained from the China Meteorological Administration (http://www.cma.gov.cn/, accessed on 1 January 2025). The dataset contains daily meteorological data from 1967 to 2016 at each station in the study area. Based on the monthly precipitation data from 1967 to 2016 at the meteorological stations in the study area, the monthly precipitation in the URJR was obtained by using the kriging interpolation method. The reference evapotranspiration values of all meteorological stations from 1967 to 2016 were calculated using the Penman Monteith formula, and then the monthly reference evapotranspiration of the URJR was obtained through kriging interpolation.

3. Approaches

3.1. Bernaola–Galván (BG) Segmentation Algorithm

The Bernaola–Galván (BG) segmentation algorithm is a mutation point detection method based on recursive segmentation and statistical hypothesis testing, which was proposed in 2001 [38]. This method is a data-driven algorithm that does not require the pre-setting of the number of mutation points and can adaptively identify significant changes in statistical features in time series. This algorithm can calculate the empirical p-value by using sampling techniques, effectively overcoming the influence of sequence autocorrelation and significantly improving the robustness and reliability of the detection results. In addition, the recursive segmentation mechanism of this algorithm can reveal the non-stationary and multi-level structural features of the time series layer by layer, further enhancing the adaptability and analytical ability of the method. This algorithm was initially used for DNA sequence analysis. Due to its good universality and robustness, it has been widely applied in fields such as climatology, hydrology, finance, and neuroscience [39,40,41,42]. In view of the excellent performance of the BG algorithm in handling non-stationary time series and nonlinear mutation tests, this study adopts this method to detect mutation points in the runoff data of two hydrological stations in the URJR.

3.2. ABCD Hydrological Model

The ABCD hydrological model is an aggregated model based on water balance, which divides a river basin into four water reservoirs, soil moisture, groundwater, surface water, and snow cover, and describes the water exchange among these reservoirs through simple nonlinear relationships. This model has a simple structure, few parameters, and relatively low requirements for input data. The ABCD model was widely applied in the simulation of hydrological processes in river basins [43,44,45,46,47,48,49]. The Nash efficiency coefficient (NSE) and relative error (RE) are selected as the evaluation indexes. If the absolute value of RE between runoff simulation and observation values is less than 20%, and the NSE is greater than 0.6, it is considered that the established hydrological model can be applied to simulate the runoff of the study area.

3.3. Multiple-Timescale Budyko Model

The multiple-timescale Budyko model was put forward in 2013 [50] and was widely applied in the quantitative analysis of the contribution rate of climate and human factors to runoff variation [51,52,53,54,55]. The expression of the seasonally scaled Budyko model is as follows:

$${\displaystyle \frac{ET}{P-\Delta S}}={\left[1+{\left({\displaystyle \frac{ETp}{P-\Delta S}}-\phi \right)}^{-\omega }\right]}^{-{\displaystyle \frac{1}{\omega }}}$$

(1)

P represents precipitation, ET represents actual evapotranspiration, ET_p represents potential evapotranspiration, ΔS is the variation of storage water, $\phi$ is the lower bounds of aridity indexes, and ω is the underlying surface parameter.

$$\Delta ET=E{T}_2-E{T}_1$$

(2)

$$ETh=E{T}_2-E{T}_2’$$

(3)

$$E{T}_c=\Delta ET-E{T}_h$$

(4)

The actual evapotranspiration values during the base period and the mutation period are ET₁ and ET₂, respectively. The simulated actual evapotranspiration during the mutation period is ET₂′. ET_h represents the change in actual evapotranspiration caused by human activities, and ET_c represents the change in actual evapotranspiration caused by climate change.

4. Result and Analysis

4.1. Mutation Analysis of Runoff Depth in the URJR

In this study, the B-G segmentation mutation analysis method was adopted to conduct mutation tests on the Batang (Figure 2) and Shigu (Figure 3) hydrological stations in the URJR. When using the BG segmentation algorithm, it is generally believed that when P_Tmax is greater than 0.5, it is considered that the year corresponding to P_Tmax is the mutation year. From Figure 2, we found that the P_Tmax of the Batang hydrological station in 1998 was 0.65541, indicating that 1998 was the mutation year of runoff depth at the Batang hydrological station. From Figure 3, we found that the P_Tmax of the Shigu hydrological station in 1987 was 0.51721, indicating that 1987 was the mutation year of runoff depth at the Shigu hydrological station.

4.2. Runoff Simulation

Based on results of mutation analysis, the research period (1967–2016) of the Batang hydrological station was divided into base (1967–1998) and variation periods (1999–2016), and the research period (1967–2016) of the Shigu hydrological station was divided into base (1967–1987) and variation periods (1988–2016). Then, the ABCD hydrological model was employed to simulate runoff variation processes of the Batang and Shigu stations during base and variation periods. Actual evapotranspiration values at the monthly scale of the two hydrological stations were simulated, providing a basis for the subsequent multi-timescale attribution analysis of actual evapotranspiration. Parameters and evaluation indexes of ABCD model in the URJR are shown in

Table 1. Parameters and evaluation indexes of the ABCD model in the URJR.

Stations	Periods	a	b	c	d	Periods	NSE	RE/%
Batang	Base period	0.70	390	0.60	0.50	Calibration period (1967–1982)	0.80	−9.31
						Validation period (1983–1998)	0.84	−3.39
	Variation period	0.75	400	0.53	0.50	Calibration period (1999–2007)	0.80	8.34
						Validation period (2008–2016)	0.84	−3.92
Shigu	Base period	0.60	432	0.57	0.64	Calibration period (1967–1977)	0.87	−3.24
						Validation period (1978–1987)	0.88	4.47
	Variation period	0.70	400	0.50	0.40	Calibration period (1988–2002)	0.83	4.25
						Validation period (2003–2016)	0.86	−7.02

. The Nash coefficients (NSE) of the Batang and Shigu stations for both base and variation periods are more than 0.8, and the relative errors (REs) of the Batang and Shigu stations for both base and variation periods are less than 10%, indicating that the simulated runoff results are reliable. Therefore, the ABCD hydrological model can well simulate the monthly runoff changes of the Batang and Shigu stations for both base and mutation periods. Monthly runoff simulation results of the Batang station in base (1967–1998) and mutation periods (1999–2016) are displayed in Figure 4. Monthly runoff simulation results of the Shigu station in base (1967–1987) and mutation periods (1988–2016) are displayed in Figure 5. In conclusion, the ABCD hydrological model can simulate high-precision actual evapotranspiration data on a monthly scale

4.3. Trend Analysis of Multi-Timescale Actual Evapotranspiration in the URJR

Based on the actual evapotranspiration data on a monthly scale from 1967 to 2016 simulated in Section 4.2, we analyzed the temporal variation of multiple-timescale (quarterly and monthly) actual evapotranspiration at the Batang and Shigu hydrological stations (

Table 2. Trend analysis of actual evapotranspiration changes at various hydrological stations.

Time	Batang		Shigu
	β (mm/a)	Z	β (mm/a)	Z
January	0.02	2.52 **	0.03	3.87 **
February	0.03	2.68 **	0.04	3.89 **
March	0.03	3.65 **	0.04	4.11 **
April	0.05	3.84 **	0.05	4.59 **
May	0.11	4.30 **	0.13	5.20 **
June	0.12	3.78 **	0.17	5.23 **
July	0.16	3.34 **	0.22	5.37 **
August	0.12	3.32 **	0.21	5.42 **
September	0.09	2.59 **	0.16	4.35 **
October	0.06	2.97 **	0.08	4.34 **
November	0.04	2.78 **	0.05	4.03 **
December	0.03	2.59 **	0.04	4.37 **
Spring	0.06	4.77 **	0.08	5.35 **
Summer	0.14	4.20 **	0.20	5.87 **
Autumn	0.06	3.03 **	0.09	4.73 **
Winter	0.03	3.70 **	0.04	5.16 **

Note: ** indicates significant at the 0.01 level.

). From Table 2, the actual evapotranspiration values at the Batang station in spring, summer, autumn, and winter all showed significant growth (p < 0.01), with growth rates of 0.0591 mm/a, 0.1358 mm/a, 0.0560 mm/a, and 0.0274 mm/a, respectively. The actual evapotranspiration values at the Shigu hydrological station in spring, summer, autumn, and winter all showed significant growth (p < 0.01), with growth rates of 0.0783 mm/a, 0.1995 mm/a, 0.09 mm/a, and 0.0373 mm/a, respectively. From a monthly scale perspective, the monthly actual evapotranspiration values at the Batang hydrological station all showed significant growth at the 0.01 level. The months with relatively large growth rates were June, July, and August, which increased significantly, with growth rates of 0.1223 mm/a, 0.1553 mm/a, and 0.1231 mm/a, respectively. The monthly actual evapotranspiration values at the Shigu hydrological station all showed significant growth at the 0.01 level. The months with relatively large growth rates were June, July, and August, which increased significantly, with growth rates of 0.1734 mm/a, 0.2231 mm/a, and 0.2050 mm/a, respectively.

4.4. Multiple-Timescale Attribution Analysis of Actual Evapotranspiration in the URJR

For calculating contributions of climate and anthropic factors on multiple-timescale actual evapotranspiration, parameters of the multiple-timescale Budyko model in the base period need to be fitted (

Table 3. Fitting parameters and evaluation indexes of the Budyko model in the base period.

	Batang				Shigu
Time	Parameters		Evaluation Indexes		Parameters		Evaluation Indexes
	Ω	φ	NSE	RE%	ω	φ	NSE	RE%
January	0.85	0.83	0.86	0.04	0.72	0.83	0.83	0.04
February	0.89	0.78	0.95	0.05	0.77	0.96	0.94	0.19
March	0.87	0.20	0.75	0.09	0.72	0.13	0.70	−0.02
April	0.81	0.20	0.78	0.16	0.70	0.15	0.71	−0.02
May	0.80	0.40	0.77	0.86	0.70	0.15	0.72	−0.01
June	0.86	0.09	0.72	−0.03	0.79	0.14	0.73	−0.01
July	0.97	0.15	0.79	0.10	0.89	0.23	0.75	−0.38
August	1.10	0.31	0.73	−0.07	0.92	0.22	0.71	0.01
September	0.97	0.23	0.75	0.12	0.82	0.26	0.70	−4.07
October	0.96	0.47	0.74	0.63	0.82	0.41	0.75	0.07
November	0.85	0.49	0.93	0.27	0.71	0.33	0.93	−0.05
December	0.84	0.72	0.89	0.14	0.68	0.48	0.91	0.06
Spring	0.82	0.30	0.82	0.34	0.71	0.32	0.77	−0.01
Summer	0.96	0.20	0.85	0.46	0.84	0.16	0.85	0.01
Autumn	0.87	0.20	0.77	−0.52	0.76	0.23	0.72	−0.02
Winter	0.87	0.94	0.88	0.73	0.76	0.99	0.88	0.01

). From Table 3, the NSE values of the Batang and Shigu stations were more than 0.7, and relative errors were less than 5%, indicating that the constructed multiple-timescale Budyko model could be applied for quantitatively analyzing the impact degree of climatic variation and anthropic factors on multiple-timescale actual evapotranspiration in the URJR (

Table 4. Attribution analysis of multiple-timescale actual evapotranspiration.

Time	Batang		Shigu
	ETc/mm	ETh/mm	ETc/mm	ETh/mm
January	0.79	0.35	0.42	0.65
February	0.89	0.21	0.56	0.62
March	0.79	0.14	0.91	0.95
April	1.10	0.15	0.88	1.12
May	1.22	1.28	1.77	1.99
June	2.14	1.38	2.35	2.89
July	3.74	1.14	2.51	3.38
August	2.62	1.73	2.39	3.36
September	1.93	0.84	0.57	4.07
October	0.71	0.69	0.74	1.69
November	0.76	0.39	0.94	0.81
December	0.74	0.36	0.35	0.60
Spring	3.10	1.57	3.55	4.06
Summer	8.50	4.25	7.25	9.62
Autumn	3.41	1.92	2.25	6.57
Winter	2.42	0.92	1.33	1.88

).

From Table 4, we found that climatic variation is dominant factor leading to actual evapotranspiration growth at multiple timescales (quarterly and monthly) at the Batang hydrological station. Climatic variation has led to increases of 3.10 mm, 8.50 mm, 3.41 mm, and 2.42 mm in actual evapotranspiration at the Batang Hydrological Station in spring, summer, autumn, and winter, respectively. From a monthly scale perspective, both climate change and human activities have a positive increasing effect on the actual evapotranspiration in the upper reaches of the Jinsha River from January to December. Moreover, the effect intensity of climate change is more than that of human activities. The months when climate change has a greater impact on the actual evapotranspiration are June, July, and August, with increases of 2.14 mm, 3.74 mm, and 2.62 mm, respectively. The months when anthropogenic activities have a greater impact on actual evapotranspiration are May, June, July, and August, with the actual evapotranspiration increasing by 1.28 mm, 1.38 mm, 1.14 mm, and 1.73 mm, respectively. The area above the Batang Hydrological Station is a typical plateau cold region basin. Its hydrological processes are mainly driven by glacial meltwater and precipitation. The ecological environment is primitive and fragile, and the influence of human activities is weak.

From Table 4, we found that human activity is the dominant factor leading to actual evapotranspiration growth at multiple timescales (quarterly and monthly) in the Shigu hydrological station. From a seasonal-scale perspective, anthropogenic activities have resulted in growths of 4.06 mm, 9.62 mm, 6.57 mm, and 1.88 mm in actual evapotranspiration at the Shigu station in spring, summer, autumn, and winter, respectively. From a monthly scale perspective, both climatic variation and anthropogenic activities have a positive increasing effect on actual evapotranspiration at the Shigu station from January to December. Moreover, the effect intensity of anthropogenic activities is more than that of climatic factors. The months when human activities have a greater impact on the actual evapotranspiration are June, July, August, and September, with actual evapotranspiration increasing by 2.89 mm, 3.38 mm, 3.36 mm, and 4.07 mm, respectively. The basin between the Batang Hydrological Station and the Shigu Hydrological Station is one of the river sections with the richest hydropower resources, the most perilous terrain, the greatest engineering challenges, and the most unique ecological environment on the entire Jinsha River and even the main stream of the Yangtze River. This section of the river basin is the core area of the key power supply points of China’s “West-to-East Power Transmission” strategy, where a series of giant hydropower stations have been built.

5. Discussions

This study comprehensively utilized B-G mutation detection, ABCD, and a multi-timescale Budyko model to estimate actual evapotranspiration and conduct multiple-timescale attribution analysis on actual evapotranspiration in the URJR. The research results showed that human activities have a significant positive impact on actual evapotranspiration growth in the URJR. As an important hydropower base in China, the Jinsha River has built multi-level cascade reservoirs, significantly expanding the water surface area of the basin and thus leading to an increase in actual evapotranspiration [56]. In addition, the increase in agricultural irrigation and industrial and domestic water use can further intensify evapotranspiration in the basin. Land use/cover changes caused by human activities also significantly alter the underlying surface properties. The research showed that vegetation restoration is an important reason for the increase in actual evapotranspiration in the Yangtze River Basin [57]. Vegetation changes can directly affect the evapotranspiration process. Excessive afforestation may lead to a reduction in runoff due to the interception of rainfall by the canopy and the water consumption of vegetation transpiration [58]. Climate change also has a significant positive impact on the actual evapotranspiration growth in the URJR, which is similar to the existing research results [59,60]. Global warming has led to an increase in the saturated vapor pressure difference in the URJR, reducing frost and extending the evapotranspiration season. In addition, it accelerates melting of glaciers and snow, generating additional meltwater and increasing the total water supply in the basin, indirectly supporting evapotranspiration.

There are still some deficiencies in this study. (1) The ABCD hydrological model cannot accurately simulate the maximum runoff value. The inaccurate simulation of the maximum runoff value (or peak flood flow) is usually not caused by a single reason but rather the result of the accumulation and amplification of uncertainties in multiple links. It mainly stems from data input, model structure, and the complexity of nature itself. Minor errors in input data, especially rainfall, are amplified during the nonlinear hydrological response process. The simplification of the structure of hydrological models makes them inherently deficient when simulating extreme and nonlinear flood events. Hydrological models have difficulty capturing the complex dynamics of human activities and natural extreme conditions. (2) This study only used the ABCD hydrological model to simulate actual evapotranspiration in the URJR. All hydrological models are simplifications of complex natural systems. Hydrological models cannot fully capture the evolution process of actual evapotranspiration in the real environment, which results in some uncertainties. In the future, multiple hydrological models will be added for simulation, and combined with the existing actual evapotranspiration data products, the accuracy of actual evapotranspiration estimation in the study area can be improved to carry out subsequent research. (3) In this study, precipitation, potential evapotranspiration and soil water storage are classified as climatic factors, while all other factors are attributed to human activities. The impacts of connection and interaction between climatic factors and human activities on actual evapotranspiration were not taken into account. Therefore, in subsequent work, we will comprehensively consider the interaction between climatic factors and human activities and calculate the contribution of interaction to actual evapotranspiration.

6. Conclusions

For quantifying the contribution rates of climatic variation and anthropogenic activities on the actual evapotranspiration at multiple timescales in the URJR, based on the monthly meteorological and hydrological data of the URJR, this study first used the BG mutation method to determine the mutation years of runoff depth for dividing the research period into base and variation periods. Then, the ABCD hydrological model was used to simulate the runoff variation process during the base period and the variation period, and the actual evapotranspiration data at the monthly scale was simulated. Finally, a multiple-timescale Budyko model was applied for quantitatively computing the impacts of climatic variation and anthropogenic activities on multiple-timescale actual evapotranspiration in the URJR.

The results demonstrated the following: (1) The mutation years of runoff depth at the Batang and Shigu hydrological stations were 1988 and 1987. (2) The actual evapotranspiration values at multiple timescales (quarterly and monthly) in the upper reaches of the Jinsha River all presented a significant increase (p < 0.01), with a growth rate ranging from 0.02 mm/a to 0.22 mm/a. (3) Climatic variation is the dominant factor leading to actual evapotranspiration growth at multiple timescales (quarterly and monthly) at the Batang Hydrological Station, with a contribution ranging from 0.71 mm to 8.50 mm. (4) Human activities are a dominant factor leading to actual evapotranspiration growth at multiple timescales (quarterly and monthly) in the Shigu hydrological station, with a contribution ranging from 0.60 mm to 9.62 mm.