Response of net primary production to land use and climate changes in the middle‐reaches of the Heihe River Basin

Abstract Net primary production (NPP) supplies matter, energy, and services to facilitate the sustainable development of human society and ecosystem. The response mechanism of NPP to land use and climate changes is essential for food security and biodiversity conservation but lacks a comprehensive understanding, especially in arid and semi‐arid regions. To this end, taking the middle‐reaches of the Heihe River Basin (MHRB) as an example, we uncovered the NPP responses to land use and climate changes by integrating multisource data (e.g., MOD17A3 NPP, land use, temperature, and precipitation) and multiple methods. The results showed that (a) land use intensity (LUI) increased, and climate warming and wetting promoted NPP. From 2000 to 2014, the LUI, temperature, and precipitation of MHRB increased by 1.46, 0.58°C, and 15.76 mm, respectively, resulting in an increase of 14.62 gC/m2 in annual average NPP. (b) The conversion of low‐yield cropland to forest and grassland increased NPP. Although the widespread conversion of unused land and grassland to cropland boosted both LUI and NPP, it was not conducive to ecosystem sustainability and stability due to huge water consumption and human‐appropriated NPP. Urban sprawl occupied cropland, forest, and grassland and reduced NPP. (c) Increase in temperature and precipitation generally improved NPP. The temperature decreasing <1.2°C also promoted the NPP of hardy vegetation due to the simultaneous precipitation increase. However, warming‐induced water stress compromised the NPP in arid sparse grassland and deserts. Cropland had greater NPP and NPP increase than natural vegetation due to the irrigation, fertilizers, and other artificial inputs it received. The decrease in both temperature and precipitation generally reduced NPP, but the NPP in the well‐protection or less‐disturbance areas still increased slightly.

Net primary production is mainly estimated by ground observations and model simulations. Field surveys and continuous observations at fixed points or FLUXNET can accurately assess local NPP and environmental factors, but it is difficult to reflect large-scale spatial characteristics. Therefore, various models are generally used to simulate NPP at the macroscale (Cramer & Field, 1999;Kicklighter et al., 1999). Climate statistics models (e.g., Miami, Thornthwaite Memorial, and Chikugo models) estimate NPP distribution based on the empirical regression relationships between NPP and climate factors (Gang et al., 2015;Lieth, 1972;Uchijima & Seino, 1985). Although these models are easy-to-use, they are less accurate as empirical relationships vary with time and space. Biological process models (e.g., CENTURY, TEM, BEPS, InTEC, CARAIB, DLEM, and ORCHIDEE models) estimate NPP based on the processes and mechanisms of biomass accumulation (Liu, Chen, Cihlar, & Park, 1997;McGuire et al., 1997;Parton et al., 1993;Piao et al., 2012;Tian et al., 2010;Warnant, Francois, Strivay, & Gerard, 1994). Different process parameters used in these models cause less comparability in simulation results. Moreover, the difficulty of obtaining spatialized parameters increases the uncertainty of model performance on large scales. Since the convenient data accessibility, light use efficiency models (e.g., C-Fix, CASA, GLO-PEM, SDBM, and TURC models) based on remote sensing have been widely used for large-scale and long-term NPP estimation (Bonan, 1995;Nayak, Patel, & Dadhwal, 2010;Potter et al., 1993;Ruimy, Saugier, & Dedieu, 1994;Zhu, Pan, He, Yu, & Hu, 2006). The MOD17 NPP product of Moderate Resolution Imaging Spectroradiometer (MODIS) considered multiple physiological processes (e.g., photosynthesis, and leaf and root respiration) and the relationship between light use efficiency and environmental factors. Thus, MOD17 improved the accuracy of NPP estimation and was commonly used to analyze the spatiotemporal patterns of NPP on regional and global scales (Running et al., 1999(Running et al., ,2004Turner et al., 2006). However, MOD17 did not address the issue of cloud contamination in the fraction of photosynthetically active radiation (FPAR)/LAI products it used (MOD15A2) (Zhao, Heinsch, Nemani, & Running, 2005;Zhao & Running, 2010). Therefore, in order to promote the sustainable development of ecosystems, we should comprehensively study the coupling response mechanism of NPP to land use and climate changes on the basis of a more accurate estimation of NPP.
Future climate change will lead to a wetter climate in humid regions and a drier climate in arid and semi-arid regions (MA, 2005).
Compared with the humid regions, the eco-environment in arid and semi-arid regions is more fragile, with a faster population growth.
Environmental change and human interference will further aggravate the ecological imbalance in arid and semi-arid regions, resulting in more serious or even irreversible negative impacts (Liu et al., 2019;MA, 2005). Thus, it is urgently needed to study the response of NPP to land use and climate changes in these regions. To this end, using the middle-reaches of the Heihe River Basin (MHRB) in northwestern China as an example, the objectives of this study are as follows: (a) to analyze the spatiotemporal pattern changes of the NPP in the MHRB from 2000 to 2014; (b) to reveal the response mechanism of NPP to land use and climate changes; (c) to discuss the reasons for the spatial heterogeneity of the response mechanism of NPP; and to provide strategies for the sustainable development of ecosystem.

| Study area
The Heihe River Basin (HRB) is the second largest inland river basin in the arid and semi-arid regions of northwestern China, located in the middle of the Silk Road Economic Belt (Figure 1). The upper, middle, and lower reaches of the HRB are water conservation areas, agricultural oasis areas, and ecological conservation areas, respectively.
The MHRB is one of the top 10 commodity grain bases in China by virtue of its fertile soil, abundant sunshine, and convenient irrigation.
The climate in MHRB is a typical temperate continental climate, with an average annual rainfall of 100-250 mm (70% of the precipitation are concentrated in June-August), an average annual temperature of 6-8°C, an annual sunshine duration of over 3,000 hr, and an annual potential evapotranspiration of 1,600-2,400 mm (Liu, Song, & Deng, 2016;Song, Liu, Deng, Zhang, & Han, 2018). Water is the key factor constraining the development of the ecosystem and human society in the MHRB, in which agriculture consumes more than 90% of the water resources (Liu, Song, & Deng, 2017). Land use and climate changes affect the supply and distribution of water resources, resulting in a change in the ecosystem state of MHRB (Song, Liu, Arowolo, Zhang, & Xu, 2018a;Tan & Zheng, 2017). Changes in temperature and precipitation have enhanced snowmelt, which have led to a slight increase in available water resources of MHRB. However, F I G U R E 1 Location of the middlereaches of the Heihe River Basin the increase in agricultural and urban water consumption has led to ecological water reduction and ecosystem degradation (Liu, Song, & Deng, 2017;Song, Liu, Arowolo et al., 2018a).

| Data sources
The NPP data used in this study were the MODIS MOD17A3 product, a level 4 version-55 product at an annual interval, with a spatial resolution of 1,000 m (https://modis.gsfc.nasa.gov/). MOD17A3 developed by NTSG/UMT has effectively eliminated the cloud pollution in the version-4 NPP product of NASS and has passed the stage-3 accuracy evaluation, which can be applied systematically and firmly (Running et al., 1999;Zhao et al., 2005). We collected two images of the MHRB each for 2000 and 2014, and their row/column numbers were 25/4 and 25/5, respectively.
The LUC data used in this study were from the Climate Change Initiative Land Cover dataset (CCI_LC) (https://www.esa-landcover-cci. org/), which has an annual interval and a spatial resolution of 300 m.

| NTSG MOD17A3 algorithm
Compared with the previous NASA MOD17 algorithm (Running et al., 1999(Running et al., ,2004, the NTSG MOD17A3 algorithm has made numerous improvements such as spatial interpolation, cloud pollution removal, and biome parameter recalibration (Zhao et al., 2005;Zhao & Running, 2010). Thus, the version-55 NPP product developed by NTSG is more accurate and reliable than version-4 product. The NTSG MOD17A3 algorithm incorporates multisource data into the light use efficiency-based model to estimate NPP (Zhao et al., 2005;Zhao & Running, 2010).
where NPP is equal to the gross primary production (GPP) subtracts the maintenance (R m ) and growth respiration costs (R g ). GPP depends on the active radiation use efficiency (ε), the FPAR, and the photosynthetically active radiation (PAR). FPAR is from the satellitederived MOD15 product, and PAR is from the independent estimates of the Global Modeling and Assimilation Office in NASA (GMAO/NASA).
As two stress factors, minimum temperature (TMIN s ) and vapor pressure deficit (VPD s ) attenuate the maximum radiation conversion efficiency (ε max ) to produce the final ε.
where R m consists of the maintenance respiration costs for fine roots (FR m ) and leaves (LR m ). R g is empirically parameterized as 25% of NPP. F w and L w refer to the weights of fine roots and leaves, respectively. FRB m and LRB m refer to the maintenance respiration per unit fine root and leaf carbon at 20°C, respectively. T avg is the average temperature.

| Reclassification of land use
Based on the previous research (Liu et al., 2019), we regrouped the land cover types of the CCI_LC classification system into six land use types, namely, urban land, cropland, forested areas, grassland, water areas, and unused land ( Table 1). The reclassification was performed on the User Tool 3.14, a software developed by the European Space Agency that can be used for coordinate transformation, reclassification, and resampling.

| Mapping of land use intensity
According to the degree of human activity interference, previous research divided the six main land use types into four grades: Namely, the grading index of unused land was 1, that of forested areas, grassland, or water areas was 2, that of cropland was 3, and that of urban land was 4 . In order to match the where LUI x denotes the LUI of grid x, G i denotes the grading index of land use i, and P i,x denotes the area proportion of land use i on the grid x.    ). Along with the increase in LUI, NPP increased when LUI was between 100 and 300, while NPP decreased when LUI was above 300. Generally, the LUI changes (increase or decrease) in the MHRB promoted NPP to a certain extent, and the increase in amplitude of NPP was greater when LUI increases (Figure 5b).

| Response of NPP to temperature changes
The annual average temperature in the MHRB increased from southwest to northeast ( Figure 6). From 2000 to 2014, the low-temperature region (<0°C) has increased slightly, while the high-temperature region (>7°C) has significantly expanded (Figure 6a,b). Consequently, the average temperature in the MHRB increased to 0.58°C (10.78%) ( Table 3). The temperature in high-elevation area was below 0°C and decreased significantly (Figure 6c). The northeast corner of MHRB also showed a decreasing temperature. The increase in temperature above 0.7°C was mainly occurred in the middle MHRB. The two major urban expansion areas (Zhangye and Jiuquan-Jiayuguan) had the largest increase in temperature above 1.4°C.
In the temperature increasing areas, the NPP values of most cropland, forested areas, and grassland increased (positive correlation) (Figure 6d). However, temperature increasing led to a decline NPP (negative correlation) in sparse grassland and desert in the extremely arid areas. In the low-temperature region, the NPP of most forested areas and grassland increased despite decrease in temperature (negative correlation). When temperature was 0-5°C, the NPP of MHRB increased with increasing temperature. However, when temperature was above 5°C or below 0°C, NPP decreased with increasing temperature.
Generally, the NPP of MHRB increased when temperature increased or the decrease in amplitude was <1.2°C (Figure 7b,d,f,h).

| Response of NPP to precipitation changes
The annual precipitation in the MHRB decreased from southwest to northeast (Figure 8). From 2000 to 2014, the areas with precipitation above 150 mm have expanded (Figure 8a,b). Consequently, the annual precipitation of MHRB increased by 15.76 mm (9.79%) ( Table 3).
The changes in annual precipitation showed a decreasing trend from southeast to southwest (Figure 8c). The precipitation of MHRB increased in the low-elevation agricultural oasis, while decreased in the high-elevation mountainous areas. The largest increase in precipitation (>40 mm) was in the southeast, while the largest decline (<−40 mm) was in the southwest.
In the precipitation increasing areas, the NPP values of most cropland, forested areas, and grassland increased (positive correlation), while that of most unused land (arid desert) decreased slightly (negative correlation) (Figure 8d). In the southwest areas with a relatively high precipitation, although both precipitation and temperature decreased, the NPP of most cropland and forested areas still increased slightly (negative correlation).

| Spatial heterogeneity of NPP response to land use and climate changes
Land use and climate changes jointly determined the spatiotemporal pattern of NPP and its changes. Driven by economic benefits, farmers in the MHRB reclaimed more cropland and adjusted the crop planting structure (Liu et al., 2016). With the help of artificial irrigation and other agricultural inputs, the NPP of cropland was generally On the contrary, affected by economic and population growth, the urban sprawl of MHRB caused a reduction of cropland, forested areas, and grassland around cities (Liu et al., 2019) and an increase in LUI. Consequently, the NPP of newly expanded urban land decreased . However, due to the expansion of urban green space, the NPP in urban centers increased (Yan et al., 2018).
Previous research confirmed that climate change plays the major role in NPP variations in the arid and semi-arid regions Li, Wang et al., 2018). Generally, the climate warming and wetting in the MHRB were conducive to NPP increase. A certain degree of warming facilitated the photosynthesis of thermophilic corn in the central MHRB (Liu et al., 2016) and promoted its NPP. Moreover, warming boosted the snowmelt of MHRB (Song, Liu, Deng et al., 2018), which was beneficial to NPP increase.
Meanwhile, the temperature decreasing <1.2°C also promoted the NPP of hardy forest, grass, and crops (wheat, barley, and rapeseed) in high-altitude mountains, due to the simultaneous precipitation increase. In addition, rural-urban migration has weakened the human activities (e.g., wood cutting and grazing) in mountain areas Xiao, Hu, Tan, Li, & Li, 2018), which further increased the NPP in forested areas and grassland.
However, warming-induced water stress compromised photosynthesis (Zhao & Running, 2010), leading to a decline NPP in arid sparse grassland and desert areas. The decrease in both temperature and precipitation generally reduced NPP, but the NPP in the well-protection areas still increased slightly due to ecological conservation measures such as grazing prohibition, livestock reduction, and the Grain-for-Green project.

| Policy implications for improving ecosystem NPP and sustainability
Although previous studies have confirmed the NPP increase in arid and semi-arid regions due to the greening of cities Yan et al., 2018), urban sprawl-induced cropland and ecological land losses, and other negative impacts should pay more attention (Liu et al., 2019). Therefore, it is crucial to control urban sprawl and foster urban green space (Liu et al., 2019) for the improvement of NPP and the sustainable development of socio-ecological systems. Besides, cropland expansion, which was mainly from unused land and grassland, improved the NPP of MHRB. However, most of the cropland NPP was appropriated by human society and consumed huge water. Thus, cropland expansion is an unsustainable way to food security. We should cultivate water-saving crops (Liu et al., 2016), import high-virtual-water agricultural products from waterrich areas (Liu et al., 2019), improve crop water use efficiency (Song, Liu, Deng et al., 2018), and close crop yield gaps (Lu & Fan, 2013;Xin, Li, Zhu, & Tan, 2009) if we are to achieve food-water-ecological security in arid and semi-arid areas. Furthermore, the conversion of low-yield cropland to forested areas and grassland increased the NPP of the MHRB. However, previous research has demonstrated that it is worth nothing to widely plant trees and grass in the arid and semi-arid climatic conditions, because such man-made NPP growth is generally temporary with limited water resources Yin, Pflugmacher, Li, Li, & Hostert, 2018 Although cropland experienced worse climate change, its NPP increase was still greater than natural vegetation due to the irrigation, fertilizers, and other artificial inputs it received. The decrease in both temperature and precipitation generally reduced NPP, but the NPP in the well-protection or less-disturbance areas still increased slightly. In order to well adapt the future climatic changes, effective ecosystem management should focus on alleviating or avoiding negative human activities (e.g., excessive expansion of cropland and urban land, deforestation) and strengthening positive ones (e.g., grazing prohibition, reforestation). A comprehensive understanding of the NPP response mechanism to land use and climatic changes provides decision-makers with the foundation for improving ecosystem NPP and sustainability. However, the interactions between land use and climate changes complicated this response mechanism. How to eliminate the interrelated influences of the two factors and attribute the coupling response mechanism is the focus and difficulty of future ecosystem management.

ACK N OWLED G M ENTS
This work was supported by the National Natural Science

CO N FLI C T O F I NTE R E S T
None declared.