This study presents a global map of the critical soil moisture threshold (θcrit) for plant water stress, derived from satellite observations of land surface temperature (LST) diurnal amplitude (dLST) and soil moisture (SM) during dry-down periods. The average global θcrit is 0.19 m³/m³, varying from 0.12 m³/m³ in arid ecosystems to 0.26 m³/m³ in humid ecosystems. Earth System Models (ESMs) overestimate θcrit in dry areas and underestimate it in wet areas. The observed θcrit reflects plant adaptation to soil water availability and atmospheric demand. Using explainable machine learning, the study identifies aridity index, leaf area, and soil texture as the most influential drivers of θcrit variation. The annual fraction of days with water stress, when SM is below θcrit, has increased over the past four decades, highlighting the importance of understanding θcrit for predicting drought impacts and ecosystem vulnerability. The study combines satellite observations of dLST and SM with in-situ data from flux towers to generate a global θcrit map. The results show that θcrit is influenced by climatic, biotic, and edaphic factors, with aridity index being the most significant. The study also finds that θcrit varies across biomes, with higher values in humid ecosystems and lower values in dryland ecosystems. The study further shows that θcrit is affected by crop type and management practices, with rice having a higher θcrit than maize, wheat, and potato. The study also finds that θcrit is positively correlated with precipitation frequency and negatively correlated with shortwave radiation. The study compares θcrit estimates from Earth System Models with observation-based maps and finds that models significantly underestimate θcrit in wet regions and overestimate it in dry regions. This suggests that models may not accurately represent the soil moisture point of inception of plant water stress in wet regions. The study also finds that θcrit is a key variable in understanding the coupling between soil-plant systems and the atmosphere, and that its spatial variation is important for understanding ecosystem functioning and carbon cycle dynamics. The study provides a global map of θcrit and its drivers, which can be used to improve land-surface model representations of soil moisture constraints on water and carbon cycles. The study also highlights the importance of understanding θcrit for predicting the impacts of climate change on ecosystems and the need for improved model simulations of soil moisture and related processes.This study presents a global map of the critical soil moisture threshold (θcrit) for plant water stress, derived from satellite observations of land surface temperature (LST) diurnal amplitude (dLST) and soil moisture (SM) during dry-down periods. The average global θcrit is 0.19 m³/m³, varying from 0.12 m³/m³ in arid ecosystems to 0.26 m³/m³ in humid ecosystems. Earth System Models (ESMs) overestimate θcrit in dry areas and underestimate it in wet areas. The observed θcrit reflects plant adaptation to soil water availability and atmospheric demand. Using explainable machine learning, the study identifies aridity index, leaf area, and soil texture as the most influential drivers of θcrit variation. The annual fraction of days with water stress, when SM is below θcrit, has increased over the past four decades, highlighting the importance of understanding θcrit for predicting drought impacts and ecosystem vulnerability. The study combines satellite observations of dLST and SM with in-situ data from flux towers to generate a global θcrit map. The results show that θcrit is influenced by climatic, biotic, and edaphic factors, with aridity index being the most significant. The study also finds that θcrit varies across biomes, with higher values in humid ecosystems and lower values in dryland ecosystems. The study further shows that θcrit is affected by crop type and management practices, with rice having a higher θcrit than maize, wheat, and potato. The study also finds that θcrit is positively correlated with precipitation frequency and negatively correlated with shortwave radiation. The study compares θcrit estimates from Earth System Models with observation-based maps and finds that models significantly underestimate θcrit in wet regions and overestimate it in dry regions. This suggests that models may not accurately represent the soil moisture point of inception of plant water stress in wet regions. The study also finds that θcrit is a key variable in understanding the coupling between soil-plant systems and the atmosphere, and that its spatial variation is important for understanding ecosystem functioning and carbon cycle dynamics. The study provides a global map of θcrit and its drivers, which can be used to improve land-surface model representations of soil moisture constraints on water and carbon cycles. The study also highlights the importance of understanding θcrit for predicting the impacts of climate change on ecosystems and the need for improved model simulations of soil moisture and related processes.