The extensive afforestation efforts on the Loess Plateau, incurring hundreds of billions of CNY, trigger heightened vegetation cover, depleting soil water, and imperiling ecosystem sustainability. Widespread debate persists over the feasibility and optimal locations for afforestation. However, what has been overlooked is the potential presence of alternative stable states within ecosystems, a captivating system equilibrium behavior. This study integrates remote sensing, minimal model, and environmental data to investigate the equilibrium behavior (quantified by tree cover) of forest ecosystems on the Loess Plateau and its implications. The findings suggest a threshold relationship between tree cover and annual precipitation, with a significant increase observed up to 400 mm. Beyond this threshold, alternative stable states emerge, characterized by high tree cover (forest, >35%) and medium tree cover (open woodland, 7%~35%). The equilibrium behavior of the forest ecosystem combines thresholds and alternative stable states. Increasing spatial heterogeneity, especially the positive feedback between vegetation and precipitation, results in advancing transition thresholds with higher annual precipitation. Regime shifts from forest to open woodland increase carbon stock but decrease water yield, revealing a trade-off between carbon sequestration and water resources. This nuanced understanding of equilibrium enhances both theoretical comprehension and practical planning for afforestation on the Loess Plateau, promoting the functions and services of the forest ecosystem.

The current approaches have known limitations to understanding the coupling of terrestrial ecosystem evapotranspiration (ET) and photosynthesis (referred to as gross primary productivity, GPP). To better characterize the relationship between ET and GPP, we developed a novel remote sensing (RS)-driven approach (RCEEP) based on the underlying water use efficiency (uWUE). RCEEP partitions transpiration (T) from ET using a RS vegetation index (VI)-derived ratio of T to ET (VI-fT) and then links T and GPP via RS VI-derived Gc (VI-Gc) rather than leaf-to-air vapor pressure difference. RCEEP and other two uWUE versions (VI-T or VI-G), which only incorporate VI-fT or VI-Gc , were evaluated and compared with the original uWUE model in terms of their performances (Nash-Sutcliffe efficiency, NSE) in estimating GPP from ET over 180 flux sites covering 11 biome types over the globe. Results revealed better performances of VI-T and VI-G compared to the original uWUE, implying remarkable contributions of VI-fT and VI-Gc to a more meaningful relationship between ET and GPP. RCEEP yielded the best performances with a reasonable mean NSE value of 0.70 (0.76) on a daily (monthly) scale and across all biome types. Further comparisons of RCEEP and approaches modified from recent studies revealed consistently better performances of RCEEP and thus, positive implications of introducing VI-fT and VI-Gc in bridging ecosystem ET and GPP. These results are promising in view of improving or developing algorithms on coupled estimates of ecosystem ET and GPP and understanding the GPP dynamics concerning ET on a global scale.