1. Introduction
The restoration of degraded ecosystems and converted lands into healthy ecosystems, commonly referred to as ecological restoration, is a crucial social solution for addressing land degradation, preserving biodiversity, stabilizing the climate, enhancing ecosystem functioning, and providing ecosystem services to communities [
1], [
2], [
3]. Numerous international and national political initiatives have recognized the importance of ecosystem restoration and have proposed a range of restorative agenda. Notably, the United Nations (UN) has officially designated the time spanning from 2021 to 2030 as the “UN Decade on Ecosystem Restoration”, while Africa’s Great Green Wall project aims to reforest a 7000 km stretch south of the Sahara. Meanwhile, massive land-sustainability programs have been implemented in China to combat land degradation [
4]. However, in regions with limited water resources, blind ecological restoration may exacerbate the tradeoffs between carbon and water, restricting restoration efficiency, effectiveness, and social well-being, which has sparked much debate [
5], [
6].
Adequate water supply is a prerequisite for sustainable vegetation growth, and the interdependence between vegetation and water significantly impacts the exchange of water, carbon, and energy between the terrestrial and atmospheric systems [
7], [
8], [
9]. The projected effects of global warming include increasingly severe water constraints on vegetation growth, attributable to heightened vapor pressure deficit and potential reductions in soil water content. Furthermore, satellite observations have detected a global greening trend, which indicates heightened water requirements for vegetation [
10], [
11]. These constraints are particularly pronounced in water-limited regions, where water input relies heavily on precipitation and can significantly impact the efficacy of vegetation growth and restoration efforts [
12], [
13], [
14].
Recent studies have predominantly concentrated on the correlation between vegetation and water, specifically examining the sensitivity of vegetation to water availability [
15], [
16], the impacts of drought on the restoration of vegetation [
17], [
18], and the resilience of vegetation in the face of drought [
19], [
20], utilizing field experiments, remote sensing, and process-based models. Nevertheless, utilizing spatially explicit recognition of the vegetation−water relationship has yet to serve well for ecological restoration practice. Nonetheless, it can serve as a crucial point of reference for forest management, such as plant species selection, and facilitate informed decisions regarding the adaptability of vegetation to drought. To bolster ecologically significant restoration strategies, it is imperative to consider both rainfall use efficiency (RUE), defined as the ratio of vegetation productivity to rainfall input, and the sensitivity of vegetation to rainfall measured by sequential linear regression [
14]. Furthermore, the extensively inconsistent change in vegetation greenness and productivity raises concerns about whether ecological restoration can improve ecosystem function to match the increase in vegetation greenness. MODIS-based remote sensing data reveals that 45.6% of vegetated regions across the globe exhibited inconsistent trends in vegetation greenness and productivity [
21], while the increase in the latter is lower than the rate of the former [
22]. By increasing the amount of photosynthetically active radiation absorbed, vegetation greening can enhance productivity [
8], [
23]. However, as reported in previous studies, this greening increase does not signify an improvement in ecosystem functioning, partly constrained by light-use and carbon-use efficiency or excessively dense leaf area [
24], [
25]. As a result, targeted restoration and adaptation efforts need to involve vegetation greenness, vegetation productivity, and vegetation-water relation to guide sustainable restoration.
To combat extensive and dramatic land degradation, which has involved substantial economic costs, the Chinese government established several large-scale ecological projects [
4], [
13], [
26]. Consequently, a noticeable greening trend was observed in the Yellow River Basin (YRB) [
13], [
27], [
28]. It is a densely populated area with numerous ecological barriers and is also on the land route of the “One Belt One Road” initiative. The central government proposed to promote environmental protection and high-quality development of the YRB in 2019. It seeks to increase environmental protection and social advancement. However, scarce water resources and the fragile environment may compromise the government’s greening strategy in the future [
5]. Therefore, incorporating the understanding of vegetation-rainfall relationships into sustainable regional revegetation is imperative to comprehending the biophysical processes involved in the restoration and ensuring the longevity of ecological management [
29], [
30].
To better understand the relationship between greening and stabilization in the YRB, this study analyzed the vegetation functioning and rainfall availability based on leaf area index (LAI), net primary production (NPP), and rainfall data between 2000-2018. The trend of NPP sensitivity to rainfall was assessed as well. The main aims of our study are: ① to detect the vegetation greenness and spatial productivity distribution in different parts of the YRB (upper, middle, and lower reaches); ② to compare the RUE in YRB upper, middle, and lower reaches; and ③ to identify the key areas for ecological restoration according to the trends of vegetation greenness and productivity and vegetation-rainfall relationship in the YRB.
2. Materials and methods
2.1. Study area
The Yellow River spans a total area of 7.95 × 10
5 km
2, flowing through nine provinces and regions, including Qinghai, Sichuan, Gansu, Ningxia Hui Autonomous Region, the Inner Mongolia Autonomous Region, Shaanxi, Shanxi, Henan, and Shandong. Kenli County of Shandong Province merges into the Bohai Sea (
Fig. 1). The elevation is high in the west and low in the east. The Yellow River links the Qinghai-Xizang Plateau, the Loess Plateau, and the northern China plains. Water scarcity is vital in ecological security barriers, combating desertification and providing water supply. There is a continental climate in the area of the Yellow River. The region exhibits a semi-humid climate in the southeastern area, a semiarid climate in the central region, and an arid climate in the northwestern part [
31]. As a result of the landforms and habitats of the YRB, a variety of plant species can flourish there. In the late 1990s, the Chinese government carried out the “Grain for Green Program.” 16 000 km
2 of rain-fed cropland were successfully converted to non-crop vegetation, resulting in a discernible trend of “greening” ([
32], [
33], Fig. S1 in Appendix A). Over the past 20 years, toward restoring degraded landscapes, the “Grain for Green Program” implemented state-of-the-art management strategies and advanced technical innovations [
34].
2.2. Datasets
LAI, NPP, rainfall, and administrative division data were used in this study. We used yearly NPP data between 2000 and 2018 obtained from the MOD17A3 product with a 1 km spatial resolution (unit: kg C·m
−2) from the Numerical Terra Dynamic Simulation Group (NTSG) and National Aeronautics and Space Administration (NASA), given data availability [
35], [
36]. The MOD17 algorithm has been modified to correct the problem of cloud-contaminated MODIS LAI-fraction of photosynthetically active radiation inputs produced by the NTSG of University of Montana, USA. LAI data between 2000 and 2018 was obtained from the MOD15A2H V6 product (US Geological Survey Earth data
†) with a 500 m spatial resolution [
37]. To mitigate the impact of atmospheric noise and cloud contamination, the LAI data was consolidated into yearly values using the maximum value composite technique [
38]. The annual rainfall data between 2000 and 2018 from climate hazards group infrared precipitation with station data were resampled from the original resolution of 0.05 degrees to match the 1 km NPP data resolution [
39].
2.3. RUE
The RUE is calculated by dividing NPP by rainfall as a proxy for the plant production response to rainfall (Eq. (1)).
where RUE (kg C·m−2·mm−1) denotes the rainfall use efficiency, and NPP and R are yearly data of NPP (kg C·m−2) and rainfall (mm), respectively.
Rainfall is the primary climatic constraint on vegetation growth in arid and semiarid regions. Therefore, RUE can reflect ecosystem changes by describing the ecosystem-specific relationship between NPP and rainfall. Thus, RUE has been frequently used to evaluate ecosystem function and land degradation [
40], [
41], [
42].
2.4. Sensitivity of NPP to rainfall
We described the vegetation-rainfall relation (the sensitivity of NPP to rainfall) by using a sequential regression method between rainfall and vegetation productivity (using NPP as a proxy) within a spatial-temporal window (
Fig. 2) [
14], [
43]. According to Ref. [
43], the optimal moving window size is 7 × 7 pixels, and the optimal time step is four years (
Fig. 2(b)). NPP and rainfall were regressed linearly for all pixels in the moving window (7 × 7) and over the selected period (four years) (
Fig. 2(c)). In order to determine the sensitivity of the NPP to rainfall, the regression slope was applied to the moving window’s center pixel (
Fig. 2(d)). Thus, we can obtain the annual sensitivity of NPP to rainfall for each pixel (
Fig. 2(e)). When more than two-thirds of the pixels had no data, the regression was not fitting, and the regression slope was not assigned to the center pixel. We were able to identify the pixel scale sensitivity of the whole study region as the window moved. NPP is highly sensitive to rainfall in the center pixel if the regression slope has a high absolute value. Normally, slopes greater than zero indicate positive and negative sensitivity if the slope is lower than zero. A positive sensitivity indicates that NPP will change direction with rainfall and vice versa. Alterations in the NPP-rainfall relationship may be indicated by changes in vegetation biophysical processes and ecosystem function concerning water availability [
43].
2.5. Statistical analysis
Vegetation greenness was measured using LAI, and NPP was used to measure vegetation productivity. Some derivative indexes were applied based on vegetation indexes: RUE and sensitivity of NPP to rainfall. Those two derivative indexes’ significantly positive/negative trends can be interpreted as an increase or decrease in unit productivity per unit rainfall, indicative of vegetation functioning changes. In order to simulate the trend of each pixel in these four indicators, linear regression analysis was used based on least-squares methods. Significant trends were considered at
p < 0.05. Vegetation indicators’ trends were categorized into three groups according to significance level: significantly increased (+), significantly decreased (−), or no significant change (0). This procedure defined a significant trend as a significant pattern of vegetation characteristics (e.g., greenness, productivity, or efficiency) in any index (e.g., LAI, NPP, and RUE). Following the rules outlined above, the trend consistency was judged based on the three vegetation characteristics for each pixel. An ideal state would be one where LAI, NPP, and RUE all show a significantly positive trend, illustrating that the greenness, vegetation productivity, and RUE of regional vegetation are improving. A significant trend of sensitivity of NPP to rainfall is an auxiliary indicator. Changes in vegetation water usage and functioning, as illustrated by the sensitivity of NPP to rainfall, may indicate the changing sensitivity (that is, reduced or increased vegetation resistance) of vegetation to environmental stressors [
44]. This section’s state definition and consistency are described by Ref. [
21].
3. Results
3.1. Spatial variation of vegetation greenness and productivity
Fig. 3 shows a combination of changes in the LAI and NPP of the YRB. Based on satellite data, the “greening” percentage in the YRB showed a significant increase since 2000. 34.53% of the YRB exhibited significant greening. Our analysis revealed that approximately 20.54% of the upper reaches of the YRB showed a significant positive trend, and 53.37% and 16.73% in the middle and lower reaches of the YRB, respectively. The middle reaches of the YRB, which accounted for 42.95% of the whole YRB, including northern Shaanxi and east of Ordos in Inner Mongolia, showed a significant greening trend. Meanwhile, the variation trend of NPP in the upper, middle, and lower reaches of the YRB is consistent with LAI. In particularly, our results detected that nearly 65% of middle reaches show significant trends in NPP since 2000. This translates into 63.72% positive and 0.65% negative trends. It is important to note that the significant increase in the proportion of NPP in the upstream (33.75%) and middle reaches (63.72%) is higher than that of LAI (20.54% and 53.37%, respectively) but lower than that of LAI in the downstream (9.68% vs 16.73%).
3.2. RUE trend
Fig. 4 shows the trends of rainfall and RUE across the YRB. Except for a few areas, such as the upper reaches, western Ordos, and northern Lanzhou, the majority of the YRB shows an increasing trend in rainfall (
Fig. 4(a)). On average, the rainfall trend in the upper, middle, and lower reaches of the YRB is 2.38 mm·a
−1, 4.60 mm·a
−1, and 0.31 mm·a
−1, respectively. Our results also found that about 16% of middle reaches exhibit significant trends, translating into 15.09% positive and 1.29% negative trends. Compared with the middle reaches of the YRB, a smaller area of significant upstream and downstream changes was observed. For example, in the YRB’s lower reaches, only about 2% exhibit significant trends, translating into 1.21% positive and 0.70% negative trends. In the upper reaches of YRB, approximately 12% exhibit significant trends, translating into 9.98% positive and 2.57% negative trends. Significant variations in the distribution and direction of trends were evident in the upper, middle, and lower reaches, with the midstream showing the largest positive trends (15.09%), followed by the upstream (9.98%) and downstream (1.21%). On the other hand, a greater fraction of negative trends was found in the northwest Sichuan Province of the upper reaches (
Fig. 4(b-i)), followed by western Inner Mongolia (
Fig. 4(b-ii)) and parts of Shanxi Province (
Fig. 4(b-iii)) of the middle reaches of the YRB.
3.3. Upper-middle-lower reaches differences in vegetation-rainfall sensitivity
We observed an important upper-middle-lower difference in the distribution and direction of trends of vegetation-rainfall sensitivity (
Fig. 5). Our results showed that in YRB’s upper reaches, 19.66% show significant trends in vegetation-rainfall sensitivity, 16.90% of the trends are positive, and 2.76% are negative, while in the lower reaches of the YRB, only 4.98% show significant negative trends. However, our results show that the trends in vegetation-rainfall sensitivity of the middle reaches differ from those of the upstream and downstream. 16.86% of the middle reaches of the YRB show significant trends in vegetation-rainfall sensitivity, translating into 14.08% negative and 2.78% positive trends. In particular, the significant negative trends in the upper and lower reaches of the YRB were lower than in the middle reaches. Specifically, negative trends in vegetation-rainfall sensitivity of the upper reaches were mainly in the west of Qinghai Lake (
Fig. 5(a-i)). In contrast, the middle reaches are mainly concentrated in northern Inner Mongolia (
Fig. 5(a-ii)) and southeastern Shanxi, including Taiyuan, Linfen, and Yuncheng Cities (
Figs. 5(a-iii)-(a-v)).
3.4. Key area identification in the combined vegetation-rainfall relationship
Comparing the changing directions of vegetation greenness, productivity, and RUE, we found pronounced regional differences in the distribution and direction across the YRB (
Fig. 6). The areas with a consistently significant increase in vegetation greenness, productivity, and RUE accounted for 8.41% of the YRB, mainly distributed in the north of Yan’an and the south of Yulin City, Shaanxi Province of the middle reaches of YBR (
Figs. 6(a-vii) and
(a-viii)). Furthermore, the areas with increased greenness and productivity but unchanged RUE accounted for 16.55% of the YRB and were mainly concentrated in the eastern Gansu Province (Pingliang and Qingyang Cities, respectively) and southern Ningxia Hui Autonomous Region. Moreover, areas with an increase in vegetation greenness and no changes in productivity or RUE accounted for 9.23% of the YRB, indicating enhanced vegetation growth. Within the YRB’s middle reaches, these areas were widely distributed, especially in the northwest of Shaanxi Province, and also located near areas with enhanced greenness and NPP. However, about 2.26% of the YRB consistently decreased in these three indicators. Such regions are mainly distributed in the northern Inner Mongolia and northern Ningxia Hui Autonomous Region (
Figs. 6(a-i)-(a-iii)), the northwest Sichuan Province of the upper reaches and the Henan section in the YRB’s lower reaches (
Figs. 6(a-v) and
(a-viii)). Overall, the regions exhibiting variations in vegetation greenness and productivity were predominantly located in the middle reaches of the YRB and were of considerable magnitude.
4. Discussion
4.1. Ecological engineering drives vegetation greenness and productivity improvement
Vegetation restoration is a critical and effective solution to prevent soil erosion and promote ecological management. As a pilot region, several large-scale and ambitious ecological restoration programs, including the “Grain for Green Project,” “Natural Forest Land Resource Conservation Project,” and “Three-North Shelterbelt Project,” have been carried out in the YRB since 1998 to attain the target of conserving and expanding forests and mitigating soil erosion. Consequently, a greening trend has been widely detected, and human-induced ecosystem degradation has been greatly reduced [
13], [
45], [
46]. Our study also generally showed a significant increasing trend of LAI and NPP in each reach of the YRB since 2000 (
Fig. 3). This was consistent with that of Ref. [
46], who reported a policy-driven massive greening in the Loess Plateau of China. The statistical book revealed that in Shanxi, Shaanxi, and Ningxia, a total area of 4.793 × 10
4 km
2 underwent afforestation [
47], [
48]. Undoubtedly, ecological restoration has played a notable role in greening vegetation in the middle reaches of the YRB.
Furthermore, our findings indicate that the increase in the proportion of NPP was significantly greater than that of LAI in the upper and middle reaches of the YRB (
Fig. 3), and
Fig. 6 also shows inconsistent changes in LAI and NPP. The structure and function of an ecosystem will be directly affected when vegetation grows inconsistently in terms of greenness and productivity. By increasing vegetation leaf area, chlorophyll and photosynthetically active radiation are absorbed more efficiently, thus increasing vegetative productivity and respiration [
23]. Nevertheless, photosynthesis slows down when vegetation leaves reach a specific area due to self-shadowing and nutrient limitation [
24]. Differences in the driving factors for LAI and NPP are a possible explanation. LAI changes were primarily influenced by CO
2 fertilization, and climate change contributed less (8%), while the latter can explain 28.6% of productivity changes [
8], [
49]. Previous research has reported a parabolic relationship between NPP and LAI in humid areas and the decoupling of greenness and productivity as aridity decreases [
50], [
51]. It is related to the tradeoffs between ecosystem structure and physiology becoming stronger in more humid climates [
50]. This decoupling of greenness-productivity may be further caused by environmental variables (e.g., solar radiation, water availability, and temperature) affecting light use efficiency [
25], [
52]. Besides, evenness is regarded as a key factor mediating the relationship between forest productivity and richness [
53]. In ecological restoration, a single tree species is often selected and planted uniformly, which limits community productivity. This conclusion also aligns with the differences between LAI and NPP of the YRB. On the other hand, despite plantations such as timber stands and tree crops increasing the vegetation greenness rapidly, the inconsistency also depends on the timeframe. For instance, many apple orchards are on the Loess Plateau, and aggressive expansion is planned. The fruits accounted for a certain percentage of apple trees and showed large seasonal variations.
Overall, the ecological restoration efforts have resulted in a notable augmentation of vegetation greenness and productivity in the YRB, with the middle reach exhibiting the most significant improvements. From 2000 onwards, the carbon sequestration potential in the Loess Plateau has increased by approximately one-third [
48]. The vegetation restoration reduced soil erosion and sediment flow into the Yellow River, accounting for 57% of the total sediment reduction [
54]. In addition to ecological benefits, ecological restoration promoted social benefits, including the advancement of agricultural transformation and the indirect promotion of emerging industries (e.g., ecotourism and folk tourism), as previous studies reported [
48], [
55]. Increased farmer income resulted from implementing policies supporting ecological restoration, ecological compensation, and food subsidies [
56]. According to the UN Sustainable Development Goals (SDGs) framework, ecological restoration in the YRB will promote multiple SDGs, including SDG1 (no poverty), SDG6 (clean water and sanitation), SDG13 (climate action), and SDG15 (life on land), through ecological restoration efforts in the YRB.
4.2. Inconsistent trends between vegetation-rainfall sensitivity and other vegetation indicators
The changes in greenness, productivity, RUE, and sensitivity showed the diversity of vegetation growth dynamics (
Fig. 4,
Fig. 5). Numerous studies showed that massive greening drives the growth of water use efficiency or RUE, especially in the middle reach of the YRB. A significantly positive trend of RUE was generally found in regions of eastern Gansu and northwestern Shaanxi, respectively, where many plantations have been developed [
57], [
58]. Sustained vegetation greening further enhances vegetation productivity, accelerating the process of vegetation photosynthesis to a large extent and triggering an essential response in the ecosystem through long-term greening, which means that the total transpiration and evaporation in land vegetation correspondingly increased to a certain degree [
59]. The observed trends may be attributed to the implementation of extensive ecological engineering practices, which have developed more productive systems. A notable illustration of this phenomenon is the expansion of Shaanxi Province’s green map by 400 km towards the north. The regional ecological environment has changed from “overall deterioration and local improvement” to “overall improvement and local virtuous circle.” On the other hand, our analysis found that the proportion of a significant increase in RUE (11.93%) is smaller than that of NPP (45.94%) and LAI (34.53%) (
Fig. 3,
Fig. 4). In particularly, more than a quarter of pixels showed vegetation greenness significantly increased, and yet RUE did not increase (
Fig. 6), suggesting that greening did not mean increased RUE.
Moreover, significant positive and negative trends were observed in vegetation-rainfall sensitivity across the YRB. Generally, the vegetation is becoming more responsive to rainfall over time due to positive trends. Negative trends indicate less-responsive vegetation and lower productivity per available unit of rainfall, demonstrating a shift towards vegetation that may be hindered in its functioning. As the primary driver for vegetation productivity, rainfall is the main water source for restoration. Our results further revealed that the greening of vegetation in the YRB’s middle reaches is the most obvious. In contrast, the proportion of a significant decrease in sensitivity is higher than that of a significant increase, indicating that vegetation has a weak response to rainfall. The vegetation productivity per rainfall unit is low, meaning vegetation with lower RUE is concentrated. One possible explanation is that the rainfall input is sufficient for potential carbon sequestration in parts of the middle reaches of the YRB but still in great deficit. On the other hand, under equivalent precipitation, the sensitivity trend is spatially divergent between natural and artificial forests. Our results found that the areas with a significant negative trend of vegetation-rainfall sensitivity in the middle reaches of the Yellow River are mainly in the middle reach of Taihang and Lvliang Mountains of Shanxi, where an area of large-scale plantation construction with higher rainfall is located, indicating that this region is also the main one of vegetation with lower RUE distribution. Previous studies reported a positive trend in vegetation-rainfall sensitivity can be observed with a change in ecosystem composition [
43], especially replacing large woody plants with shrubs. However, an open question is whether such a positive change would also result in sustainable environmental improvements, benefiting the ecosystem and people’s livelihood. By studying the changing trend of sensitivity, we can also provide new thoughts on the transformation of inefficient forests and the optimization of landscape patterns for future vegetation restoration strategies, especially from the perspective of sustainable utilization of water resources, previous study also expressed the concern that the water constraint of ecological restoration in the Loess Plateau might reach the threshold [
5].
4.3. Ecological restoration based on the vegetation-rainfall relationship
In practice, the relationship among LAI, NPP, and RUE provides an intuitive map for decision-making and ecological restoration planning for a high-quality development strategy of the YRB. Based on the changes in RUE trends, we can identify the key restoration areas under the background of high-quality development in the YRB. A reinterpretation of these three indicators can also help identify appropriate conservation strategies for given locations in the context of climate change. Conservation strategies may arise from an LAI perspective (monitoring and identifying present greening), an NPP perspective (impacts and improving lower productivity plantations), and/or an RUE perspective (enhancing ecosystems’ capacity to recover from rainfall change). In regions with a consistent increase in LAI, NPP, and RUE, which account for 8.41% of the YRB, there is no need for additional adaptation measures or maintenance of the current status. This trend indicates a positive sign of ecological restoration in terms of habitat quality, carbon sequestration, and vegetation stability. However, nearly all significantly decreasing trends of LAI, NPP, and/or RUE mainly occurred in regions with rapid urbanization, such as Xi’an City, Xining City, and Yinchuan City. Consequently, balancing the tradeoffs between urbanization and ecosystem protection is necessary in areas where RUE and sensitivity showed a significant decreasing trend, for example, in the middle reaches of the Taihang and Lvliang Mountains of Shanxi Province. However, the rainfall is relatively stable, and inefficient forests (where vegetation utilization efficiency of rainfall is low) were widely distributed (
Fig. 6(a-i)). Management practices should prioritize strengthening the forest quality and ecosystem service function, planting trees with particularly traits, and adapting to the land and trees to improve low-efficiency forests. Considering future climate change scenarios, transforming inefficient forests in the YRB, especially in the YRB’s middle reaches, is still necessary.
Our findings have important implications for vegetation protection and ecological restoration in arid and semiarid ecosystems. High-quality development of the YRB is the current national strategy. Under the current vegetation greenness, productivity, and RUE, we distinguished the corresponding water conservation areas, key soil and water conservation areas, and biodiversity conservation areas in the upper, middle, and lower reaches of the YRB, respectively. Four highlighted recommendations, which may benefit vegetation restoration in practice, are stressed here (
Fig. 7). First, based on our results, we recommend that areas with increased greenness and RUE maintain the current conditions. Special attention, therefore, should be paid to natural restoration planning and management strategies. For areas with declining RUE, valid measures should be taken to improve the forests with lower RUE, especially in the face of accelerated climate change. Furthermore, in areas where the greenness decreased but the RUE was maintained, suitable vegetation types should be carefully selected and planted according to sloping conditions for ecological restoration. Moreover, greater attention should be paid to weighing livelihood and eco-environmental improvement in areas where greenness and RUE are declining. Finally, continuous
in-situ monitoring should be carried out for the areas where RUE and greenness are not changing significantly.
Satellite observations are becoming increasingly crucial for detecting vegetation changes, and uncertainty about regional ecosystem change might be exacerbated if studies rely on single indices to explore driving mechanisms. In summary, our study demonstrated that considering the trends of LAI, NPP, and RUE can be used to identify the regions that meet the requirements for ecological protection and high-quality development of the YRB. Based on the three indicators mapped in this study, it is possible to identify more appropriate adaptive conservation strategies to restore the YRB to its ecological health. It has been demonstrated that satellite-based data and statistical methods can be used to inform conservation planning and restoration prioritization based on vegetation-rainfall relationships at the YRB. Nevertheless, it should be noted that sensitivity assessments solely capture the vegetation’s temporary reaction to climate disturbances within brief timeframes and are inadequate for representing the long-term ecosystem attributes in the future.
Our study identifies the biophysical constraints of ecological restoration based on the vegetation-water relationship, while practical restoration necessitates broadening the focus beyond the biophysical to economic constraints [
60], [
61]. More than 45 billion CNY had been invested in the “Grain for Green Program” in the Shanxi Province, Shaanxi Province, and Ningxia Hui Autonomous Region by 2014 [
48]. In addition, several other restoration programs, such as the “Natural Forest Land Resource Conservation Project” and “Three-North Shelterbelt Project” in the YRB, also involve huge amounts of economic investment [
4]. These investments have also led to a large increase in gross ecosystem product and indirect social benefits [
4], [
48]. Also, according to the experience gained from the African Great Green Wall, every USD invested in land restoration generates yields ranging from 1.1 to 4.4 USD across different scenarios. This is an optimistic sign, while it also relies on a well-developed carbon market to ensure sustainable restoration investments [
62]. Although our study does not consider economic constraints, the biophysical constraints based on the vegetation-water relationship can clarify which areas of vegetation with lower RUE need to be further optimized and which areas need to be monitored more intensively. This could provide a significant reference for more economical ecological restoration practices.
4.4. Contributions, limitations, and uncertainties
This study used two indicators depicting the vegetation-water relationship, RUE and sensitivity of vegetation to rainfall, combined with LAI and NPP as the proxy of vegetation greenness and productivity to spatially explicitly guide ecological restoration. The present study underscores the significance of examining the correlation between vegetation and precipitation within the context of the high-quality development strategy of the YRB. The outcomes of our analysis and the proposed ecological restoration framework can provide decision-makers with valuable insights for executing rational basin pattern optimization and sustainable management. Our analytical framework is transferable to other regions that have implemented or plan to implement large-scale ecological restoration programs worldwide.
However, there were still some limitations. On the one hand, the mechanisms of the inconsistencies among LAI, NPP, and RUE need to be further monitored or detected, particularly for the disparities among different vegetation types, and supported by field monitoring. On the other hand, the influences of social and economic factors (e.g., population pressure or urbanization development) on the vegetation-rainfall sensitivity need to be considered. Although positive trends have been observed, specific ecosystem properties, such as biodiversity loss or impact on groundwater, have not been considered. Moreover, our proposed restoration strategy is based on the relationship between vegetation and rainfall but needs to consider the needs of agricultural and domestic water. Regarding the datasets used in this research, the LAI datasets’ resolution differs from that of the NPP datasets, which may lead to some uncertainties. Previous studies have identified these problems and limitations [
63]. Consequently, detailed studies of the driving factors need to be conducted to fully understand the mechanisms behind the inconsistencies in vegetation growth to provide references for the ecological protection and high-quality development of the YRB.
5. Conclusions
Given the constraints of water availability on ecosystem functioning in water-limited ecosystems, restoration planning should consider the long-term functional response of vegetation productivity to rainfall. Extensive ecological initiatives (e.g., afforestation) facilitated the substantial proliferation of greenery and enhanced vegetation productivity in most of the YRB, particularly in its middle reaches. However, the proportion of significant positive trends of RUE is far lower, suggesting that massive greening did not equal an overall increase in RUE. Furthermore, a large negative trend in vegetation-rainfall sensitivity was observed, particularly in northern Inner Mongolia and southeastern Shanxi, pointing towards less vegetation productivity per available unit of rainfall. This indicates the possible presence of inefficient forests and functional degradation risks for the ecosystem, highlighting the urgency of strengthening forest management and incorporating vegetation-water relations into conservation/restoration planning. In combination with the above indicators, an intuitive map for vegetation restoration planning and decision-making for the high-quality development strategy of the YRB was provided. Management priorities should be given to the middle reaches of the Taihang and Lvliang Mountains of Shanxi and optimizing the inefficient forests for the observed negative trends of RUE and vegetation-rainfall sensitivity. This study provides comprehensive insights for understanding structural and functional vegetation changes and their functional response to rainfall, which is helpful for optimizing ecological restoration in conditions of limited water availability.
Acknowledgments
This study was supported by the Fundamental Research Funds for the Central Universities (QNTD202303), the National Natural Science Foundation of China (42177310 and 42377331), and the National Key Research and Development Program (2022YFF1300803). Yang Yu received the Outstanding Chinese and Foreign Youth Exchange Program supported by China Association for Science and Technology (2020-2022).
Compliance with ethics guidelines
Yang Yu, Ting Hua, Liding Chen, Zhiqiang Zhang, and Paulo Pereira declare that they have no conflict of interest or financial conflicts to disclose.
Appendix A. Supplementary data
Supplementary data to this article can be found online at
https://doi.org/10.1016/j.eng.2023.07.012.
PDF
(4720KB)
