Using InVEST to evaluate water yield services in Shangri-La, Northwestern Yunnan, China

Study area

Shangri-La City is in the northwestern part of Yunnan Province, China, between 26°52′–28°52′N and 99°20′–100°19′E and has a total area of 11,600 km² (Fig. 1). The city forms a core part of the World Natural Heritage site “Three Parallel Rivers” and falls within a fragile and sensitive area of the ecosystem in the hinterland of the Hengduan Mountains. The city has also been designated a biodiversity hotspot of global significance by the World Wildlife Fund (WWF) (Yang & Wang, 2015). The climate of Shangri-La City is affected by the monsoon originating from the southwestern Indian Ocean. The wet and dry seasons extend from June to October and November to May, respectively. Shangri-La City has a complex and diverse topography with the local climate characterized by high spatial variability (Liu et al., 2021b). Other climate characteristics of the region include a high atmospheric transparency, increased daytime solar radiation, and rapid increases in temperature, resulting in a large diurnal temperature range. The Jinsha River is the dominant river system in Shangri-La City, with the 13 main tributaries being the Shuoduogang, Dongwang, Gangqu, Jiren, Langdu, Liangmei, Annan, Tangman, Anle, Maidi, Niru, Xinglong, and Baishui rivers. The average surface water and groundwater resources in the study area over many years are 5 billion m³ and 2 billion m³, respectively (Li & Wang, 2016). Shangri-La City is rich in soil resources, mainly dark brown soil, brown soil, red soil, alpine shrub meadow soil, brown dark coniferous forest soil, alpine cold desert soil, and alluvial soil (Chen et al., 2019). Most areas of Shangri-La City are subalpine and alpine landforms with elevations exceeding 3,000 m and the region contains 43 species of major forest trees, including 10 species of coniferous forests and 33 species of broadleaf forests (Liu et al., 2018; Yu et al., 2019; Zhang, Wang & Liu, 2020).

The InVEST water yield model

Water yield refers to the amount of water produced per unit area over a given period. The InVEST water yield model runs on a gridded map and is an estimation method based on the Budyko hydrothermal coupling equilibrium hypothesis in which water yield in an area is obtained by subtracting the total actual evapotranspiration per unit area from total precipitation (Budyko, 1974). The equations describing the water yield are:

(1) $$Y\left(x\right)=\left({\displaystyle \frac{1-AET\left(x\right)}{P\left(x\right)}}\right)\times P\left(x\right)$$

where $Y\left(x\right)$, $AET\left(x\right)$, and $P\left(x\right)$ are the total annual water yield (mm), the annual actual evapotranspiration (mm), and the average annual precipitation (mm) of grid unit x, respectively. ${\displaystyle \frac{AET\left(x\right)}{P\left(x\right)}}$ is the ratio of actual evapotranspiration to average annual precipitation (Zhang et al., 2004) and is calculated as:

(2) $${\displaystyle \frac{AET\left(x\right)}{P\left(x\right)}=1+{\displaystyle \frac{PET\left(x\right)}{P\left(x\right)}-{\left[1+{\left({\displaystyle \frac{PET\left(x\right)}{P\left(x\right)}}\right)}^w\right]}^{1/w}}}$$

(3) $$PET\left(x\right)=K_c\left(x\right)\times E{T}_0\left(x\right)$$

(4) $$w\left(x\right)={\displaystyle \frac{AWC\left(x\right)\times Z}{P\left(x\right)}+1.25}$$

(5) $$AWC\left(x\right)=min\left(MaxSoilDept{h}_x,RootDept{h}_x\right)PAW{C}_x$$

In Eqs. (2) to (5), $PET\left(x\right)$ is the potential evapotranspiration of grid unit x, $K_c\left(x\right)$ is the reference crop evapotranspiration coefficient, $E{T}_0\left(x\right)$ is the reference crop evapotranspiration (mm d⁻¹), $w\left(x\right)$ is a nonphysical empirical fitting parameter of natural climate-soil properties, $AWC\left(x\right)$ is the available water content of plants and is used to determine the amount of water stored and released by the soil for vegetation growth, calculated as the difference between the field water holding capacity and the wilting point, Z is the Zhang parameter, $MaxSoilDept{h}_x$ is the maximum soil depth, and $RootDept{h}_x$ is the root depth.

Correlation coefficient (R) and significance test

The Pearson correlation coefficient (R) is a measurement of the degree of linear correlation between variable, and used to quantify the correlation between water yield and precipitation and between water yield and PET for Shangri-La City from 1974 to 2015 at a pixel level. An R ≥ 0 indicates a positive correlation, whereas an R < 0 indicates a negative correlation, and the closer the absolute value of R is to 1, the stronger the correlation.

Data sources and processing

This study integrated remote sensing, a geographic information system (GIS), and the InVEST model. The data used in the study included precipitation, potential evapotranspiration, plant available water content, soil depth, land use/land cover (LULC) data, the biophysical parameters, and the Zhang parameter. In addition, the ENVI 5.3 (https://www.l3harrisgeospatial.com/Software-Technology/ENVI), ArcGIS 10.5 (https://www.esri.com/en-us/arcgis/products), SPSS 22 (https://www.ibm.com/products/spss-statistics), and InVEST (https://naturalcapitalproject.stanford.edu/software/invest) software packages were used in the study to reprocess and analyze data. The spatial resolution of all input data used in the InVEST model were unified to 30 m × 30 m.

The Zhang parameter

The Zhang parameter is an important input parameter used to estimate water yield in the InVEST model and characterizes the seasonal distribution of precipitation within a range of [0, 10]. The Zhang parameter was set to 3.8 after several test corrections as being closest to the results published in the Shangri-La Water Resources Bulletin.

Analysis of the InVEST model input data

The input data of the InVEST water yield model included annual precipitation, potential evapotranspiration (PET), plant available water capacity (PAWC), soil depth, and LULC. The average precipitation values in each region of Shangri-La City in 1974, 1989, 2003, and 2015 were 777.52 mm, 865.32 mm, 823.84 mm, and 733.48 mm, respectively, whereas the average PET values were 1,100.80 mm, 973.01 mm, 1,026.06 mm, and 1,024.88 mm, respectively. The input data for 2015 were taken as a case study to analyze the changes in the above parameters (Fig. 2). As shown in Fig. 2A, precipitation in 2015 varied spatially from 598.33 mm to 881.51 mm. The precipitation showed a highly uneven spatial distribution, with high rainfall in the south, low rainfall in the central regions, and slightly higher rainfall in the north. The annual rainfall values in Jinjiang and Hutiaoxia in 2015 were relatively high at 859.63 mm and 832.87 mm, respectively. Fig. 2B shows that the annual PET in 2015 ranged from 923.63 mm to 1,194.55 mm across the different regions. The annual PET values of Jinjiang and Hutiaoxia were also relatively high at 1,137.14 mm and 1,130.47 mm, respectively. The PET decreased gradually from south to north, showing a generally positive relationship to precipitation. Fig. 2C shows that the PAWC decreased with decreasing elevation, with high PAWC mainly distributed in areas such as Gezan, Dongwang, and Wujing. Conversely, soil depth decreased with increasing elevation, with low soil depth mainly distributed in areas such as Geza, Jiangtang, and Dongwang (Fig. 2D). The LULC in 2015 mainly comprised mixed forest, coniferous forest, alpine shrub, and alpine meadow, accounting for 35.86%, 32.17%, 8.96%, and 8.45% of the total area of the study area, respectively (Fig. 2E).

Characteristics of spatiotemporal variation of water yield

Figure 3 shows that water yield was generally higher in the north and south and lower in the central region. The ranking of the different simulation years in terms of spatially averaged annual water yield for all villages and towns was 1989 > 2003 > 1974 > 2015, with water yields of 316.05, 266.42, 221.27, and 199.06 mm, respectively. Except Dongwang Township which showed the largest annual water yield in 2003 of 314.68 mm, the annual water yields of other townships were in general consistent with that of Shangri-La City. The average water yields in Hutiaoxia, Jinjiang, and Shangjiang across the different years were relatively high. Except in 1989, the annual water yields of Jiantang Town, Wujing Township, and Luoji Township were relatively low. Taking Gezhan and Dongwang as examples in Fig. 2, the highest water yields were in areas with high PAWC, low soil depth, and alpine meadows.

The changes in water yield were classified as a significant decrease (>−30 mm), a decrease (−30 to −10 mm), a non-significant change (−10 to 10 mm), an increase (10–30 mm), and a significant increase (>30 mm). According to the above categorization, the average water yield from 1974 to 2015 from the southeast to the northwest showed significant decreases, decreases, non-significant changes, increases, and significant increases across areas of 3,840, 2,619, 2,584, 924, and 1,646 km², accounting for 33.0%, 22.6%, 22.3%, 8.0%, and 14.1% of the total area of Shangri-La City, respectively (Fig. 3E).

In addition, the correlations between precipitation and water yield were mainly positive, with an average R of 0.83 and with the area with R ≥ 0.8 accounting for ~77.56% of the total study area (Fig. 4A). In contrast, the correlations between PET and water yield were mainly negative (Fig. 4B), with an average R of −0.52 and the area with R ≤ −0.6 accounting for ~63.61% of the study area. Figure 4C shows the significance of correlation between water yield and precipitation, with the results indicating that 71.48% of the study area passed the significance test of $P\le 0.1$ (including P = 0.01, P = 0.05 and P = 0.1). Areas showing a highly significant correlation (P = 0.01) accounted for 15.36% of the total area, with these areas mainly distributed in the central areas of Shangri-La City, whereas the areas showing significant correlations of P = 0.05 accounted for 44.82% of the total area and were widely distributed. Figure 4D shows the spatial distribution of the significance of the correlations between water yield and potential evapotranspiration, indicating that only 4.534% of the study area passed the significance test of $P\le 0.1$, with these areas sporadically distributed in each region.

Spatial division of water yield importance

This study determined the grades of importance of the water yield ecosystem service functions in Shangri-La City according to the “China Ecological Protection Red Line–Ecological Function Red Line Demarcation Technical Guide (trial)”: (Grade I) generally important area (<100 mm); (Grade II) moderately important area (100–200 mm); (Grade III) highly important area (200–300 mm) and; (Grade IV) extremely important area (≥300 mm). Taking 2015 as an example (Fig. 5), the areas of Shangri-La City falling within these different grades in decreasing order were: Grade II (5,053 km²), Grade III (2,427 km²), Grade I (2,349 km²), and Grade IV (1,802 km²), accounting for 43.36%, 20.90%, 20.22%, and 15.52% of the total area, respectively. Areas falling in Grade I were mainly distributed in the east and west of Jiantang, Gezan, and Xiaozhongdian. Those in Grade II were mainly distributed in Naxi, Luoji, the northern part of Sanba, and the western part of Gezan. Those in Grade III were mainly distributed in central Dongwang, Jinjiang, Hutiaoxia, and parts of Sanba and Naxi. Areas falling in Grade IV were mainly distributed to the west of Jiantang, central Xiaozhongdian, central Gezhan, and northwestern Dongwang and Hutiaoxia. Since the areas falling within Grade IV have the highest water yield function value, these areas should be prioritized for protection.

Effect of precipitation and evapotranspiration on water yield

Changes in climatic factors affect water yield mainly by affecting precipitation and PET (Li et al., 2021b). The water yield of Shangri-La City generally showed significant changes corresponding to changes in precipitation and evapotranspiration, with the effect of precipitation on water yield far exceeding that of PET. Many past studies have similarly shown that a decrease in precipitation results in corresponding decreases in water yield (Lian et al., 2020) and that precipitation is considered to be an important factor affecting changes in water yield (Li et al., 2021b; Yang et al., 2021). Evapotranspiration is a key component of the terrestrial water cycle connecting the atmosphere, vegetation, and soil, and represents the sum of vegetation and soil evaporation (Yin et al., 2021).

Effect of LULC on water yield

LULC change is an important environmental issue that affects ecosystem services, and particularly water ecosystem services (Liu et al., 2021a). LULC can change the hydrological flux due to changes in the physical properties of the catchment surface, soil, and vegetation, including properties such as surface roughness, albedo, infiltration capacity, root depth, vegetation index, and stomatal conductance (Wei et al., 2021). The results indicated that the average water yield of shrubs (alpine, subalpine, and valley shrubs) exceeded that of meadows (alpine meadows, subalpine meadows, and valley grassland) and forests (deciduous broad-leaved, evergreen broad-leaf, coniferous, and mixed forests). Subalpine shrubs obtained the highest average water yield over 1974, 1989, 2003, and 2015 of 356.70, 525.27, 407.27, and 327.98 mm, respectively (Fig. 6A). Forest land obtained the lowest average water production since the growth of natural forest land requires the consumption of relatively large quantities of water, the evapotranspiration coefficient of natural forest land is relatively large, and the evapotranspiration capacity of forest is relatively strong. Although shrubland requires more water than grassland, the evapotranspiration coefficient of shrubland is lower than that of grassland. Therefore, the average water yield capacity of shrubland exceeded that of grassland. The average water yield of all natural/semi-natural vegetation types over time first increased and then decreased, reaching a peak in 1989.

There are differences in evapotranspiration, litter water retention, soil water content, and canopy interception among different vegetation types. Correspondingly, the different vegetation types show obvious differences in water yield capacity (Nahib et al., 2021; Yang et al., 2021). Figure 6B shows the total water production among different vegetation types, indicating that although the average water yield of forests is relatively small, the forest area plays a major role in the water yield ecological service function in Shangri-La City given that forest occupies the largest area. In general, forest land showed the highest total water yield across the four periods of 1974, 1989, 2003, and 2015, accounting for approximately 59.69%, 61.80%, 58.21% and 55.75% of the water yield of all vegetation types, respectively. The results of this study showed that LULC is an important factor regulating changes in the water yield ecosystem service function, similar to the conclusion of Guo et al. (2021).

Effect of topographic factors on water yield

Elevation significantly affects the distribution of precipitation and vegetation types, thereby indirectly affecting the water yield capacity. Figs. 6C and 6D shows the effect of elevation on the water yield ecosystem services in Shangri-La City. The average water yield of each year initially decreased and then increased with increasing elevation. Except in 1989, the lowest average water yield occurred at an elevation range of 3,500–4,000 m, whereas the highest average water yield was distributed at an elevation of <2,000 m and the second highest water yield occurred at an elevation exceeding 4,000 m (Fig. 6C). Over the four periods, the highest water yields were mainly concentrated in areas with elevations exceeding 3,000 m, accounting for >80% of the total water yield. In contrast, relatively small water yields were evident at elevations <2,500 m, accounting for ~8.5% of the total water yield (Fig. 6D). There was an increasing trend followed by a decreasing trend in average water yield from 1974 to 2015. The change in average water yield with elevation was greatest in 1989, followed by 2003. Areas of low elevation are mainly distributed in the southern part of the study area characterized by precipitation far exceeding that in the central parts and areas with higher elevation, which results in these areas of low elevation having relatively high average water yield. The proportion of forest area decreased sharply in areas of extremely high elevation, resulting in weakened evapotranspiration and a corresponding increase in the average water yield.

Water yield initially gradually decreased and then gradually increased with increasing slope, with a minimum water yield in a slope range of 35° to 45° (Fig. 6E). The total water yield was affected by the area of each slope grade, with the total annual water yield from 1974 to 2015 showing a normal distribution across the different slope grades (Fig. 6F). The majority of water yield was concentrated between slope values of 5° to 45°, accounting for ~85% of the study area, whereas the remaining slope intervals contained a negligible water yield (Fig. 6F). In addition, within each slope grade, water yield initially increased and then gradually decreased over time, with a peak in 1989. In general, topographical factors indirectly affect the water yield ecosystem service function by affecting precipitation and vegetation types, as also reported by Lian et al. (2020).

Limitations and uncertainty of the model

The InVEST model simplifies the hydrological process in that it does not distinguish between surface and subsurface water. Therefore, many uncertainties persist in the model simulation (Lian et al., 2020; Liu et al., 2021a). The InVEST model requires two main input datasets, namely climate and LULC. These datasets may be prone to errors due to various uncertainties (Hoyer & Chang, 2014). Classification of landscapes contains subjective elements such as a lack of prior knowledge which may affect classification accuracy. Poor remote sensing image quality may also contribute to poor classification accuracy. These sources of uncertainty affect any ecological assessment (Hou, Burkhard & Müller, 2013). Precipitation data and evapotranspiration data were obtained through interpolation. In contrast, climate change is complex and difficult to simulate. Therefore, it difficult to obtain accurate meteorological data with high temporal and spatial resolutions. Model parameters, such as the evapotranspiration coefficient and maximum root depth, are derived from empirical data. Therefore, the accuracy of these data will affect the accuracy of model simulation to a certain extent, although the basic pattern of water yield should remain unchanged. Furthermore, although water yield is closely related to socioeconomic development and human activities, socioeconomic data and survey data are rarely used as model input data, thereby contributing to a certain degree of uncertainty (Lang, Song & Zhang, 2017). Despite some limitations, the InVEST model is widely regarded as a suitable tool for the evaluation of a variety of ecosystem services within catchments characterized by low data availability (Vigerstol & Aukema, 2011). Future research should conduct further field sampling in Shangri-La City to verify the water yield simulated by the InVEST model in the current study. On the other hand, previous studies have shown that the relatively simple InVEST water yield model can provide an accurate estimate of water yield provided that model input data and related parameters can represent the relevant spatiotemporal scales (Redhead et al., 2016).

Shangri-La City falls within a region containing a fragile and sensitive ecosystem. The region is also a valuable tourist destination in China. However, the development of tourism may result in various negative effects. For example, Shangri-La City is known for its beautiful scenery and attracts a large number of tourists to walk, rest, camp and barbecue on the grassland, whereas trampling on the grassland, throwing away garbage and pouring soda at will, all of which cause different degrees of damage to the grassland (Yang et al., 2019b). In order to further attract tourists and improve the hospitality capacity of the scenic spot, Shangri-la City has increased the construction of roads in the scenic spot. Due to the complexity of the geographical location and terrain of the study area, the process of road construction may damage the ecological environment on both sides of the road, resulting in the loosening and instability of mountains, thereby exacerbating the occurrence of geological disasters such as collapse, landslide and debris flow. In addition, future tourism development will also result in changes to the spatial distribution of land use types. Therefore, assuring sustainable future development of tourism and the economy requires a focus on intensive use of land and reasonable planning. In addition, there should be a move to transformation of traditional tourism to ecotourism to minimize the impact of land development on mountain water yield services.