Soil moisture-atmosphere feedback dominates land carbon uptake variability.

Year-to-year changes in carbon uptake by terrestrial ecosystems have an essential role in determining atmospheric carbon dioxide concentrations1. It remains uncertain to what extent temperature and water availability can explain these variations at the global scale2-5. Here we use factorial climate model simulations6 and show that variability in soil moisture drives 90 per cent of the inter-annual variability in global land carbon uptake, mainly through its impact on photosynthesis. We find that most of this ecosystem response occurs indirectly as soil moisture-atmosphere feedback amplifies temperature and humidity anomalies and enhances the direct effects of soil water stress. The strength of this feedback mechanism explains why coupled climate models indicate that soil moisture has a dominant role4, which is not readily apparent from land surface model simulations and observational analyses2,5. These findings highlight the need to account for feedback between soil and atmospheric dryness when estimating the response of the carbon cycle to climatic change globally5,7, as well as when conducting field-scale investigations of the response of the ecosystem to droughts8,9. Our results show that most of the global variability in modelled land carbon uptake is driven by temperature and vapour pressure deficit effects that are controlled by soil moisture.

Vulnerability of Antarctica's ice shelves to meltwater-driven fracture.

Atmospheric warming threatens to accelerate the retreat of the Antarctic Ice Sheet by increasing surface melting and facilitating 'hydrofracturing'1-7, where meltwater flows into and enlarges fractures, potentially triggering ice-shelf collapse3-5,8-10. The collapse of ice shelves that buttress11-13 the ice sheet accelerates ice flow and sea-level rise14-16. However, we do not know if and how much of the buttressing regions of Antarctica's ice shelves are vulnerable to hydrofracture if inundated with water. Here we provide two lines of evidence suggesting that many buttressing regions are vulnerable. First, we trained a deep convolutional neural network (DCNN) to map the surface expressions of fractures in satellite imagery across all Antarctic ice shelves. Second, we developed a stability diagram of fractures based on linear elastic fracture mechanics to predict where basal and dry surface fractures form under current stress conditions. We find close agreement between the theoretical prediction and the DCNN-mapped fractures, despite limitations associated with detecting fractures in satellite imagery. Finally, we used linear elastic fracture mechanics theory to predict where surface fractures would become unstable if filled with water. Many regions regularly inundated with meltwater today are resilient to hydrofracture-stresses are low enough that all water-filled fractures are stable. Conversely, 60 ± 10 per cent of ice shelves (by area) both buttress upstream ice and are vulnerable to hydrofracture if inundated with water. The DCNN map confirms the presence of fractures in these buttressing regions. Increased surface melting17 could trigger hydrofracturing if it leads to water inundating the widespread vulnerable regions we identify. These regions are where atmospheric warming may have the largest impact on ice-sheet mass balance.

Implicit learning of convective organization explains precipitation stochasticity.

Accurate prediction of precipitation intensity is crucial for both human and natural systems, especially in a warming climate more prone to extreme precipitation. Yet, climate models fail to accurately predict precipitation intensity, particularly extremes. One missing piece of information in traditional climate model parameterizations is subgrid-scale cloud structure and organization, which affects precipitation intensity and stochasticity at coarse resolution. Here, using global storm-resolving simulations and machine learning, we show that, by implicitly learning subgrid organization, we can accurately predict precipitation variability and stochasticity with a low-dimensional set of latent variables. Using a neural network to parameterize coarse-grained precipitation, we find that the overall behavior of precipitation is reasonably predictable using large-scale quantities only; however, the neural network cannot predict the variability of precipitation (R2 ∼ 0.45) and underestimates precipitation extremes. The performance is significantly improved when the network is informed by our organization metric, correctly predicting precipitation extremes and spatial variability (R2 ∼ 0.9). The organization metric is implicitly learned by training the algorithm on a high-resolution precipitable water field, encoding the degree of subgrid organization. The organization metric shows large hysteresis, emphasizing the role of memory created by subgrid-scale structures. We demonstrate that this organization metric can be predicted as a simple memory process from information available at the previous time steps. These findings stress the role of organization and memory in accurate prediction of precipitation intensity and extremes and the necessity of parameterizing subgrid-scale convective organization in climate models to better project future changes of water cycle and extremes.

Large influence of soil moisture on long-term terrestrial carbon uptake.

Although the terrestrial biosphere absorbs about 25 per cent of anthropogenic carbon dioxide (CO2) emissions, the rate of land carbon uptake remains highly uncertain, leading to uncertainties in climate projections1,2. Understanding the factors that limit or drive land carbon storage is therefore important for improving climate predictions. One potential limiting factor for land carbon uptake is soil moisture, which can reduce gross primary production through ecosystem water stress3,4, cause vegetation mortality5 and further exacerbate climate extremes due to land-atmosphere feedbacks6. Previous work has explored the impact of soil-moisture availability on past carbon-flux variability3,7,8. However, the influence of soil-moisture variability and trends on the long-term carbon sink and the mechanisms responsible for associated carbon losses remain uncertain. Here we use the data output from four Earth system models9 from a series of experiments to analyse the responses of terrestrial net biome productivity to soil-moisture changes, and find that soil-moisture variability and trends induce large CO2 fluxes (about two to three gigatons of carbon per year; comparable with the land carbon sink itself1) throughout the twenty-first century. Subseasonal and interannual soil-moisture variability generate CO2 as a result of the nonlinear response of photosynthesis and net ecosystem exchange to soil-water availability and of the increased temperature and vapour pressure deficit caused by land-atmosphere interactions. Soil-moisture variability reduces the present land carbon sink, and its increase and drying trends in several regions are expected to reduce it further. Our results emphasize that the capacity of continents to act as a future carbon sink critically depends on the nonlinear response of carbon fluxes to soil moisture and on land-atmosphere interactions. This suggests that the increasing trend in carbon uptake rate may not be sustained past the middle of the century and could result in accelerated atmospheric CO2 growth.

The impacts of rising vapour pressure deficit in natural and managed ecosystems.

An exponential rise in the atmospheric vapour pressure deficit (VPD) is among the most consequential impacts of climate change in terrestrial ecosystems. Rising VPD has negative and cascading effects on nearly all aspects of plant function including photosynthesis, water status, growth and survival. These responses are exacerbated by land-atmosphere interactions that couple VPD to soil water and govern the evolution of drought, affecting a range of ecosystem services including carbon uptake, biodiversity, the provisioning of water resources and crop yields. However, despite the global nature of this phenomenon, research on how to incorporate these impacts into resilient management regimes is largely in its infancy, due in part to the entanglement of VPD trends with those of other co-evolving climate drivers. Here, we review the mechanistic bases of VPD impacts at a range of spatial scales, paying particular attention to the independent and interactive influence of VPD in the context of other environmental changes. We then evaluate the consequences of these impacts within key management contexts, including water resources, croplands, wildfire risk mitigation and management of natural grasslands and forests. We conclude with recommendations describing how management regimes could be altered to mitigate the otherwise highly deleterious consequences of rising VPD.

Diagnosing evapotranspiration responses to water deficit across biomes using deep learning.

Accounting for water limitation is key to determining vegetation sensitivity to drought. Quantifying water limitation effects on evapotranspiration (ET) is challenged by the heterogeneity of vegetation types, climate zones and vertically along the rooting zone. Here, we train deep neural networks using flux measurements to study ET responses to progressing drought conditions. We determine a water stress factor (fET) that isolates ET reductions from effects of atmospheric aridity and other covarying drivers. We regress fET against the cumulative water deficit, which reveals the control of whole-column moisture availability. We find a variety of ET responses to water stress. Responses range from rapid declines of fET to 10% of its water-unlimited rate at several savannah and grassland sites, to mild fET reductions in most forests, despite substantial water deficits. Most sensitive responses are found at the most arid and warm sites. A combination of regulation of stomatal and hydraulic conductance and access to belowground water reservoirs, whether in groundwater or deep soil moisture, could explain the different behaviors observed across sites. This variety of responses is not captured by a standard land surface model, likely reflecting simplifications in its representation of belowground water storage.

A multi-satellite framework to rapidly evaluate extreme biosphere cascades: The Western US 2021 drought and heatwave.

The increasing frequency and intensity of climate extremes and complex ecosystem responses motivate the need for integrated observational studies at low latency to determine biosphere responses and carbon-climate feedbacks. Here, we develop a satellite-based rapid attribution workflow and demonstrate its use at a 1-2-month latency to attribute drivers of the carbon cycle feedbacks during the 2020-2021 Western US drought and heatwave. In the first half of 2021, concurrent negative photosynthesis anomalies and large positive column CO2 anomalies were detected with satellites. Using a simple atmospheric mass balance approach, we estimate a surface carbon efflux anomaly of 132 TgC in June 2021, a magnitude corroborated independently with a dynamic global vegetation model. Integrated satellite observations of hydrologic processes, representing the soil-plant-atmosphere continuum (SPAC), show that these surface carbon flux anomalies are largely due to substantial reductions in photosynthesis because of a spatially widespread moisture-deficit propagation through the SPAC between 2020 and 2021. A causal model indicates deep soil moisture stores partially drove photosynthesis, maintaining its values in 2020 and driving its declines throughout 2021. The causal model also suggests legacy effects may have amplified photosynthesis deficits in 2021 beyond the direct effects of environmental forcing. The integrated, observation framework presented here provides a valuable first assessment of a biosphere extreme response and an independent testbed for improving drought propagation and mechanisms in models. The rapid identification of extreme carbon anomalies and hotspots can also aid mitigation and adaptation decisions.

Satellite solar-induced chlorophyll fluorescence tracks physiological drought stress development during 2020 southwest US drought.

Monitoring and estimating drought impact on plant physiological processes over large regions remains a major challenge for remote sensing and land surface modeling, with important implications for understanding plant mortality mechanisms and predicting the climate change impact on terrestrial carbon and water cycles. The Orbiting Carbon Observatory 3 (OCO-3), with its unique diurnal observing capability, offers a new opportunity to track drought stress on plant physiology. Using radiative transfer and machine learning modeling, we derive a metric of afternoon photosynthetic depression from OCO-3 solar-induced chlorophyll fluorescence (SIF) as an indicator of plant physiological drought stress. This unique diurnal signal enables a spatially explicit mapping of plants' physiological response to drought. Using OCO-3 observations, we detect a widespread increasing drought stress during the 2020 southwest US drought. Although the physiological drought stress is largely related to the vapor pressure deficit (VPD), our results suggest that plants' sensitivity to VPD increases as the drought intensifies and VPD sensitivity develops differently for shrublands and grasslands. Our findings highlight the potential of using diurnal satellite SIF observations to advance the mechanistic understanding of drought impact on terrestrial ecosystems and to improve land surface modeling.

Large and projected strengthening moisture limitation on end-of-season photosynthesis.

Terrestrial photosynthesis is regulated by plant phenology and environmental conditions, both of which experienced substantial changes in recent decades. Unlike early-season photosynthesis, which is mostly driven by temperature or wet-season onset, late-season photosynthesis can be limited by several factors and the underlying mechanisms are less understood. Here, we analyze the temperature and water limitations on the ending date of photosynthesis (EOP), using data from both remote-sensing and flux tower-based measurements. We find a contrasting spatial pattern of temperature and water limitations on EOP. The threshold separating these is determined by the balance between energy availability and soil water supply. This coordinated temperature and moisture regulation can be explained by "law of minimum," i.e., as temperature limitation diminishes, higher soil water is needed to support increased vegetation activity, especially during the late growing season. Models project future warming and drying, especially during late season, both of which should further expand the water-limited regions, causing large variations and potential decreases in photosynthesis.

Memory-based parameterization with differentiable solver: Application to Lorenz '96.

Physical parameterizations (or closures) are used as representations of unresolved subgrid processes within weather and global climate models or coarse-scale turbulent models, whose resolutions are too coarse to resolve small-scale processes. These parameterizations are typically grounded on physically based, yet empirical, representations of the underlying small-scale processes. Machine learning-based parameterizations have recently been proposed as an alternative solution and have shown great promise to reduce uncertainties associated with the parameterization of small-scale processes. Yet, those approaches still show some important mismatches that are often attributed to the stochasticity of the considered process. This stochasticity can be due to coarse temporal resolution, unresolved variables, or simply to the inherent chaotic nature of the process. To address these issues, we propose a new type of parameterization (closure), which is built using memory-based neural networks, to account for the non-instantaneous response of the closure and to enhance its stability and prediction accuracy. We apply the proposed memory-based parameterization, with differentiable solver, to the Lorenz '96 model in the presence of a coarse temporal resolution and show its capacity to predict skillful forecasts over a long time horizon of the resolved variables compared to instantaneous parameterizations. This approach paves the way for the use of memory-based parameterizations for closure problems.

Surface temperatures reveal the patterns of vegetation water stress and their environmental drivers across the tropical Americas.

Vegetation is a key component in the global carbon cycle as it stores ~450 GtC as biomass, and removes about a third of anthropogenic CO2 emissions. However, in some regions, the rate of plant carbon uptake is beginning to slow, largely because of water stress. Here, we develop a new observation-based methodology to diagnose vegetation water stress and link it to environmental drivers. We used the ratio of remotely sensed land surface to near surface atmospheric temperatures (LST/Tair ) to represent vegetation water stress, and built regression tree models (random forests) to assess the relationship between LST/Tair and the main environmental drivers of surface energy fluxes in the tropical Americas. We further determined ecosystem traits associated with water stress and surface energy partitioning, pinpointed critical thresholds for water stress, and quantified changes in ecosystem carbon uptake associated with crossing these critical thresholds. We found that the top drivers of LST/Tair , explaining over a quarter of its local variability in the study region, are (1) radiation, in 58% of the study region; (2) water supply from precipitation, in 30% of the study region; and (3) atmospheric water demand from vapor pressure deficits (VPD), in 22% of the study region. Regions in which LST/Tair variation is driven by radiation are located in regions of high aboveground biomass or at high elevations, while regions in which LST/Tair is driven by water supply from precipitation or atmospheric demand tend to have low species richness. Carbon uptake by photosynthesis can be reduced by up to 80% in water-limited regions when critical thresholds for precipitation and air dryness are exceeded simultaneously, that is, as compound events. Our results demonstrate that vegetation structure and diversity can be important for regulating surface energy and carbon fluxes over tropical regions.

Tropical tall forests are more sensitive and vulnerable to drought than short forests.

Our limited understanding of the impacts of drought on tropical forests significantly impedes our ability in accurately predicting the impacts of climate change on this biome. Here, we investigated the impact of drought on the dynamics of forest canopies with different heights using time-series records of remotely sensed Ku-band vegetation optical depth (Ku-VOD), a proxy of top-canopy foliar mass and water content, and separated the signal of Ku-VOD changes into drought-induced reductions and subsequent non-drought gains. Both drought-induced reductions and non-drought increases in Ku-VOD varied significantly with canopy height. Taller tropical forests experienced greater relative Ku-VOD reductions during drought and larger non-drought increases than shorter forests, but the net effect of drought was more negative in the taller forests. Meta-analysis of in situ hydraulic traits supports the hypothesis that taller tropical forests are more vulnerable to drought stress due to smaller xylem-transport safety margins. Additionally, Ku-VOD of taller forests showed larger reductions due to increased atmospheric dryness, as assessed by vapor pressure deficit, and showed larger gains in response to enhanced water supply than shorter forests. Including the height-dependent variation of hydraulic transport in ecosystem models will improve the simulated response of tropical forests to drought.

Land-atmosphere feedbacks exacerbate concurrent soil drought and atmospheric aridity.

Compound extremes such as cooccurring soil drought (low soil moisture) and atmospheric aridity (high vapor pressure deficit) can be disastrous for natural and societal systems. Soil drought and atmospheric aridity are 2 main physiological stressors driving widespread vegetation mortality and reduced terrestrial carbon uptake. Here, we empirically demonstrate that strong negative coupling between soil moisture and vapor pressure deficit occurs globally, indicating high probability of cooccurring soil drought and atmospheric aridity. Using the Global Land Atmosphere Coupling Experiment (GLACE)-CMIP5 experiment, we further show that concurrent soil drought and atmospheric aridity are greatly exacerbated by land-atmosphere feedbacks. The feedback of soil drought on the atmosphere is largely responsible for enabling atmospheric aridity extremes. In addition, the soil moisture-precipitation feedback acts to amplify precipitation and soil moisture deficits in most regions. CMIP5 models further show that the frequency of concurrent soil drought and atmospheric aridity enhanced by land-atmosphere feedbacks is projected to increase in the 21st century. Importantly, land-atmosphere feedbacks will greatly increase the intensity of both soil drought and atmospheric aridity beyond that expected from changes in mean climate alone.

Detecting forest response to droughts with global observations of vegetation water content.

Droughts in a warming climate have become more common and more extreme, making understanding forest responses to water stress increasingly pressing. Analysis of water stress in trees has long focused on water potential in xylem and leaves, which influences stomatal closure and water flow through the soil-plant-atmosphere continuum. At the same time, changes of vegetation water content (VWC) are linked to a range of tree responses, including fluxes of water and carbon, mortality, flammability, and more. Unlike water potential, which requires demanding in situ measurements, VWC can be retrieved from remote sensing measurements, particularly at microwave frequencies using radar and radiometry. Here, we highlight key frontiers through which VWC has the potential to significantly increase our understanding of forest responses to water stress. To validate remote sensing observations of VWC at landscape scale and to better relate them to data assimilation model parameters, we introduce an ecosystem-scale analog of the pressure-volume curve, the non-linear relationship between average leaf or branch water potential and water content commonly used in plant hydraulics. The sources of variability in these ecosystem-scale pressure-volume curves and their relationship to forest response to water stress are discussed. We further show to what extent diel, seasonal, and decadal dynamics of VWC reflect variations in different processes relating the tree response to water stress. VWC can also be used for inferring belowground conditions-which are difficult to impossible to observe directly. Lastly, we discuss how a dedicated geostationary spaceborne observational system for VWC, when combined with existing datasets, can capture diel and seasonal water dynamics to advance the science and applications of global forest vulnerability to future droughts.

Enforcing Analytic Constraints in Neural Networks Emulating Physical Systems.

Neural networks can emulate nonlinear physical systems with high accuracy, yet they may produce physically inconsistent results when violating fundamental constraints. Here, we introduce a systematic way of enforcing nonlinear analytic constraints in neural networks via constraints in the architecture or the loss function. Applied to convective processes for climate modeling, architectural constraints enforce conservation laws to within machine precision without degrading performance. Enforcing constraints also reduces errors in the subsets of the outputs most impacted by the constraints.

Deep learning to represent subgrid processes in climate models.

The representation of nonlinear subgrid processes, especially clouds, has been a major source of uncertainty in climate models for decades. Cloud-resolving models better represent many of these processes and can now be run globally but only for short-term simulations of at most a few years because of computational limitations. Here we demonstrate that deep learning can be used to capture many advantages of cloud-resolving modeling at a fraction of the computational cost. We train a deep neural network to represent all atmospheric subgrid processes in a climate model by learning from a multiscale model in which convection is treated explicitly. The trained neural network then replaces the traditional subgrid parameterizations in a global general circulation model in which it freely interacts with the resolved dynamics and the surface-flux scheme. The prognostic multiyear simulations are stable and closely reproduce not only the mean climate of the cloud-resolving simulation but also key aspects of variability, including precipitation extremes and the equatorial wave spectrum. Furthermore, the neural network approximately conserves energy despite not being explicitly instructed to. Finally, we show that the neural network parameterization generalizes to new surface forcing patterns but struggles to cope with temperatures far outside its training manifold. Our results show the feasibility of using deep learning for climate model parameterization. In a broader context, we anticipate that data-driven Earth system model development could play a key role in reducing climate prediction uncertainty in the coming decade.

Critical impact of vegetation physiology on the continental hydrologic cycle in response to increasing CO2.

Predicting how increasing atmospheric CO2 will affect the hydrologic cycle is of utmost importance for a range of applications ranging from ecological services to human life and activities. A typical perspective is that hydrologic change is driven by precipitation and radiation changes due to climate change, and that the land surface will adjust. Using Earth system models with decoupled surface (vegetation physiology) and atmospheric (radiative) CO2 responses, we here show that the CO2 physiological response has a dominant role in evapotranspiration and evaporative fraction changes and has a major effect on long-term runoff compared with radiative or precipitation changes due to increased atmospheric CO2 This major effect is true for most hydrological stress variables over the largest fraction of the globe, except for soil moisture, which exhibits a more nonlinear response. This highlights the key role of vegetation in controlling future terrestrial hydrologic response and emphasizes that the carbon and water cycles are intimately coupled over land.

Dry Deposition of Ozone over Land: Processes, Measurement, and Modeling.

Dry deposition of ozone is an important sink of ozone in near surface air. When dry deposition occurs through plant stomata, ozone can injure the plant, altering water and carbon cycling and reducing crop yields. Quantifying both stomatal and nonstomatal uptake accurately is relevant for understanding ozone's impact on human health as an air pollutant and on climate as a potent short-lived greenhouse gas and primary control on the removal of several reactive greenhouse gases and air pollutants. Robust ozone dry deposition estimates require knowledge of the relative importance of individual deposition pathways, but spatiotemporal variability in nonstomatal deposition is poorly understood. Here we integrate understanding of ozone deposition processes by synthesizing research from fields such as atmospheric chemistry, ecology, and meteorology. We critically review methods for measurements and modeling, highlighting the empiricism that underpins modeling and thus the interpretation of observations. Our unprecedented synthesis of knowledge on deposition pathways, particularly soil and leaf cuticles, reveals process understanding not yet included in widely-used models. If coordinated with short-term field intensives, laboratory studies, and mechanistic modeling, measurements from a few long-term sites would bridge the molecular to ecosystem scales necessary to establish the relative importance of individual deposition pathways and the extent to which they vary in space and time. Our recommended approaches seek to close knowledge gaps that currently limit quantifying the impact of ozone dry deposition on air quality, ecosystems, and climate.

Hydraulic traits explain differential responses of Amazonian forests to the 2015 El Niño-induced drought.

Reducing uncertainties in the response of tropical forests to global change requires understanding how intra- and interannual climatic variability selects for different species, community functional composition and ecosystem functioning, so that the response to climatic events of differing frequency and severity can be predicted. Here we present an extensive dataset of hydraulic traits of dominant species in two tropical Amazon forests with contrasting precipitation regimes - low seasonality forest (LSF) and high seasonality forest (HSF) - and relate them to community and ecosystem response to the El Niño-Southern Oscillation (ENSO) of 2015. Hydraulic traits indicated higher drought tolerance in the HSF than in the LSF. Despite more intense drought and lower plant water potentials in HSF during the 2015-ENSO, greater xylem embolism resistance maintained similar hydraulic safety margin as in LSF. This likely explains how ecosystem-scale whole-forest canopy conductance at HSF maintained a similar response to atmospheric drought as at LSF, despite their water transport systems operating at different water potentials. Our results indicate that contrasting precipitation regimes (at seasonal and interannual time scales) select for assemblies of hydraulic traits and taxa at the community level, which may have a significant role in modulating forest drought response at ecosystem scales.

Beyond soil water potential: An expanded view on isohydricity including land-atmosphere interactions and phenology.

Over the past decade, the concept of isohydry or anisohydry, which describes the link between soil water potential (ΨS ), leaf water potential (ΨL ), and stomatal conductance (gs ), has soared in popularity. However, its utility has recently been questioned, and a surprising lack of coordination between the dynamics of ΨL and gs across biomes has been reported. Here, we offer a more expanded view of the isohydricity concept that considers effects of vapour pressure deficit (VPD) and leaf area index (AL ) on the apparent sensitivities of ΨL and gs to drought. After validating the model with tree- and ecosystem-scale data, we find that within a site, isohydricity is a strong predictor of limitations to stomatal function, though variation in VPD and leaf area, among other factors, can challenge its diagnosis. Across sites, the theory predicts that the degree of isohydricity is a good predictor of the sensitivity of gs to declining soil water in the absence of confounding effects from other drivers. However, if VPD effects are significant, they alone are sufficient to decouple the dynamics of ΨL and gs entirely. We conclude with a set of practical recommendations for future applications of the isohydricity framework within and across sites.