Explore the vast range of titles available.

Machine learning approaches are increasingly used to extract patterns and insights from the ever-increasing stream of geospatial data, but current approaches may not be optimal when system behaviour is dominated by spatial or temporal context. Here, rather than amending classical machine learning, we argue that these contextual cues should be used as part of deep learning (an approach that is able to extract spatio-temporal features automatically) to gain further process understanding of Earth system science problems, improving the predictive ability of seasonal forecasting and modelling of long-range spatial connections across multiple timescales, for example. The next step will be a hybrid modelling approach, coupling physical process models with the versatility of data-driven machine learning. Complex Earth system challenges can be addressed by incorporating spatial and temporal context into machine learning, especially via deep learning, and further by combining with physical models into hybrid models.

Extreme weather disasters (EWDs) can jeopardize domestic food supply and disrupt commodity markets. However, historical impacts on European crop production associated with droughts, heatwaves, floods, and cold waves are not well understood—especially in view of potential adverse trends in the severity of impacts due to climate change. Here, we combine observational agricultural data (FAOSTAT) with an extreme weather disaster database (EM-DAT) between 1961 and 2018 to evaluate European crop production responses to EWD. Using a compositing approach (superposed epoch analysis), we show that historical droughts and heatwaves reduced European cereal yields on average by 9% and 7.3%, respectively, associated with a wide range of responses (inter-quartile range +2% to −23%; +2% to −17%). Non-cereal yields declined by 3.8% and 3.1% during the same set of events. Cold waves led to cereal and non-cereal yield declines by 1.3% and 2.6%, while flood impacts were marginal and not statistically significant. Production losses are largely driven by yield declines, with no significant changes in harvested area. While all four event frequencies significantly increased over time, the severity of heatwave and drought impacts on crop production roughly tripled over the last 50 years, from −2.2% (1964–1990) to −7.3% (1991–2015). Drought-related cereal production losses are shown to intensify by more than 3% yr−1. Both the trend in frequency and severity can possibly be explained by changes in the vulnerability of the exposed system and underlying climate change impacts.

Forests are a major and diverse land cover occupying a third of the terrestrial vegetated surface; they store 50 to 65% of terrestrial organic carbon (including the soil) and contribute half to terrestrial productivity. Forest biomass stores close to 80% of all the biomass on Earth. As noted earlier, forests play an important role in the Earth system as carbon stocks, carbon sinks, mediator of the water cycle and as modifier of land surface roughness and albedo. Moreover, forests play a role as habitat for many species, are an economic source of timber and firewood and have recreational value for local populations and touristic visitors. Here, we appraise how ecosystem functions are influenced in particular by biomass and its vertical and horizontal distribution and hypothesize that almost all functions are directly or indirectly related to biomass, in addition to other factors. At landscape or regional scale, heterogeneity of biomass presumably has an important influence on a variety of processes, but there are gaps both in quantifying the heterogeneity of forests globally and in quantifying the effect of this heterogeneity. Similarly, while the role of forests for the global carbon cycle is important, large uncertainties exist regarding stocks, turnover times and the carbon sink function in forest, as an analysis of state-of-the-art carbon cycle and vegetation models shows. Upcoming global satellite missions such as GEDI, NISAR and BIOMASS will be able to address the above uncertainties and lack of understanding in combination with modeling approaches, in particular by exploiting information on vertical and horizontal heterogeneity.

Climate variables carry signatures of variability at multiple timescales. How these modes of variability are reflected in the state of the terrestrial biosphere is still not quantified or discussed at the global scale. Here, we set out to gain a global understanding of the relevance of different modes of variability in vegetation greenness and its covariability with climate. We used > 30 years of remote sensing records of the normalized difference vegetation index (NDVI) to characterize biosphere variability across timescales from submonthly oscillations to decadal trends using discrete Fourier decomposition. Climate data of air temperature (Tair) and precipitation (Prec) were used to characterize atmosphere-biosphere covariability at each timescale. Our results show that short-term (intra-annual) and longerterm (interannual and longer) modes of variability make regionally highly important contributions to NDVI variability: short-term oscillations focus in the tropics where they shape 27% of NDVI variability. Longer-term oscillations shape 9% of NDVI variability, dominantly in semiarid shrublands. Assessing dominant timescales of vegetation-climate covariation, a natural surface classification emerges which captures patterns not represented by conventional classifications, especially in the tropics. Finally, we find that correlations between variables can differ and even invert signs across timescales. For southern Africa for example, correlation between NDVI and Tair is positive for the seasonal signal but negative for short-term and longer-term oscillations, indicating that both short- and long-term temperature anomalies can induce stress on vegetation dynamics. Such contrasting correlations between timescales exist for 15% of vegetated areas for NDVI with Tair and 27% with Prec, indicating global relevance of scale-specific climate sensitivities. Our analysis provides a detailed picture of vegetation-climate covariability globally, characterizing ecosystems by their intrinsic modes of temporal variability. We find that (i) correlations of NDVI with climate can differ between scales, (ii) nondominant subsignals in climate variables may dominate the biospheric response, and (iii) possible links may exist between short-term and longer-term scales. These heterogeneous ecosystem responses on different timescales may depend on climate zone and vegetation type, and they are to date not well understood and do not always correspond to transitions in dominant vegetation types. These scale dependencies can be a benchmark for vegetation model evaluation and for comparing remote sensing products.

Funding Information: We thank H. Yang, M. Schrumpf, T. Wutzler, R. Zheng and H. Ma for their comments and suggestions on this study. This work was supported by the National Natural Science Foundation of China (42125503) and the National Key Research and Development Program of China (2020YFA0608000, 2020YFA0607900 and 2021YFC3101600). F.T. was financially supported by China Scholarship Council during his visit at Food and Agricultural Organization of the United Nations (201906210489) and the Max-Planck Institute for Biogeochemistry (202006210289). The contributions of Y.L. were supported through US National Science Foundation DEB 1655499 and 2242034, subcontract CW39470 from Oak Ridge National Laboratory (ORNL) to Cornell University, DOE De-SC0023514, and the USDA National Institute of Food and Agriculture. S.M. has received funding from the ERC under the European Union’s H2020 Research and Innovation Programme (101001608). The contributions of U.M. were supported through a US Department of Energy grant to the Sandia National Laboratories, which is a multi-mission laboratory managed and operated by National Technology and Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International, Inc., for the US Department of Energy’s National Nuclear Security Administration under contract DE-NA-0003525. We thank the WoSIS database ( https://www.isric.org/explore/wosis ) for providing the publicly available global-scale SOC database used in this study. Publisher Copyright: © 2023, The Author(s).

Analyses of potential and actual biomass stocks indicate that trade-offs exist between conserving carbon stocks on managed land and raising the contribution of biomass to raw material and energy supply for the mitigation of climate change. Land management impact on biomass Land use by humans leads to loss of carbon from the Earth's biomass. Karl-Heinz Erb and colleagues estimate the amount of carbon that has been lost from terrestrial vegetation as a result of land conversion and management by compiling global maps of current terrestrial carbon stocks and potential carbon stocks that would exist without human disturbance. They find that land use has halved terrestrial carbon stocks. The effects of land management (forest management and grazing) seem to be similar to those of land conversion: land conversion accounts for 53–58% of the carbon stock losses and land management accounts for 42–47%. The findings imply that avoiding deforestation is necessary but not sufficient for climate-change mitigation. Carbon stocks in vegetation have a key role in the climate system 1 , 2 , 3 , 4 . However, the magnitude, patterns and uncertainties of carbon stocks and the effect of land use on the stocks remain poorly quantified. Here we show, using state-of-the-art datasets, that vegetation currently stores around 450 petagrams of carbon. In the hypothetical absence of land use, potential vegetation would store around 916 petagrams of carbon, under current climate conditions. This difference highlights the massive effect of land use on biomass stocks. Deforestation and other land-cover changes are responsible for 53–58% of the difference between current and potential biomass stocks. Land management effects (the biomass stock changes induced by land use within the same land cover) contribute 42–47%, but have been underestimated in the literature. Therefore, avoiding deforestation is necessary but not sufficient for mitigation of climate change. Our results imply that trade-offs exist between conserving carbon stocks on managed land and raising the contribution of biomass to raw material and energy supply for the mitigation of climate change. Efforts to raise biomass stocks are currently verifiable only in temperate forests, where their potential is limited. By contrast, large uncertainties hinder verification in the tropical forest, where the largest potential is located, pointing to challenges for the upcoming stocktaking exercises under the Paris agreement.

So far, various studies have aimed at decomposing the integrated terrestrial water storage variations observed by satellite gravimetry (GRACE, GRACE-FO) with the help of large-scale hydrological models. While the results of the storage decomposition depend on model structure, little attention has been given to the impact of the way that vegetation is represented in these models. Although vegetation structure and activity represent the crucial link between water, carbon, and energy cycles, their representation in large-scale hydrological models remains a major source of uncertainty. At the same time, the increasing availability and quality of Earth-observation-based vegetation data provide valuable information with good prospects for improving model simulations and gaining better insights into the role of vegetation within the global water cycle. In this study, we use observation-based vegetation information such as vegetation indices and rooting depths for spatializing the parameters of a simple global hydrological model to define infiltration, root water uptake, and transpiration processes. The parameters are further constrained by considering observations of terrestrial water storage anomalies (TWS), soil moisture, evapotranspiration (ET) and gridded runoff (Q) estimates in a multi-criteria calibration approach. We assess the implications of including varying vegetation characteristics on the simulation results, with a particular focus on the partitioning between water storage components. To isolate the effect of vegetation, we compare a model experiment in which vegetation parameters vary in space and time to a baseline experiment in which all parameters are calibrated as static, globally uniform values. Both experiments show good overall performance, but explicitly including varying vegetation data leads to even better performance and more physically plausible parameter values. The largest improvements regarding TWS and ET are seen in supply-limited (semi-arid) regions and in the tropics, whereas Q simulations improve mainly in northern latitudes. While the total fluxes and storages are similar, accounting for vegetation substantially changes the contributions of different soil water storage components to the TWS variations. This suggests an important role of the representation of vegetation in hydrological models for interpreting TWS variations. Our simulations further indicate a major effect of deeper moisture storages and groundwater–soil moisture–vegetation interactions as a key to understanding TWS variations. We highlight the need for further observations to identify the adequate model structure rather than only model parameters for a reasonable representation and interpretation of vegetation–water interactions.

We are in debt to FLUXNET principal investigators and researchers for the fundamental measurements and synthesis datasets used to build the upscaled and in situ flux datasets used in this study. The work used eddy covariance data from La Thuile Synthesis Dataset, which were provided by the FLUXNET community. In particular, we thank A. Altaf, J. Beringer, P. Blanken, C. Brümmer, S. Burns, J. Cleverly, E. Cremonese, T. Grünwald, P. Kolari, W. Jans, M. Leonardo, T. Manise, M. Mund, A. Noormets, E. Pendall, C. Pio, S. Prober, L. Šigut, A. Varlagin and W. Woodgate, who provided us with site-level measurements of soil carbon and vegetation biomass, and B. Amiro, J. Ardö, S. Arndt, D. Baldocchi, L. Belelli, F. Bosveld, D. Bowling, N. Buchmann, A. Christen, M. Cuntz, A. Desai, B. Drake, I. Goded, A. Goldstein, C. Gough, S. Ivan, L. Hutley, I. Janssens, M. Karan, H. Kobayashi, M. Korkiakoski, B. Kruijt, S. Linder, B. Loubet, I. Mammarella, S. Minerbi, W. Munger, Z. Nagy, D. Papale, A. Richardson, B. Ruiz, E.P. Sanchez-Canete, FCE. Silva, E. Veenendaal, S. Wharton, G. Wohlfahrt, J. Wood, D. Yakir and D. Zona, who provided contacts and/or references for us to find site-level measurements of soil carbon and vegetation biomass. We are thankful to S. Bao and S. Besnard for helping with collected and processed site-level FLUXNET and vegetation biomass data. We thank M. Migliavacca and M. Schrumpf for providing reference and useful resources for data collection. N.F. acknowledges support from the International Max Planck Research School for Global Biogeochemical Cycles. Publisher Copyright: © 2022, The Author(s).

Funding Information: This work has received funding from the Youth Innovation Subject 2022 of National Space Science Center, Chinese Academy of Sciences, Sino-German (CSC-DAAD) Postdoc Scholarship Program, and from the European Space Agency (ESA) via the Earth System Data Lab (ESDL). This work used eddy covariance data acquired and shared by the FLUXNET community, including these networks: AmeriFlux, AfriFlux, AsiaFlux, CarboAfrica, CarboEuropeIP, CarboItaly, CarboMont, ChinaFlux, Fluxnet-Canada, GreenGrass, ICOS, KoFlux, LBA, NECC, OzFlux-TERN, TCOS-Siberia, and USCCC. The FLUXNET eddy covariance data processing and harmonization was carried out by the ICOS Ecosystem Thematic Center, AmeriFlux Management Project and Fluxdata project of FLUXNET, with the support of CDIAC, and the OzFlux, ChinaFlux and AsiaFlux offices. Publisher Copyright: © 2023 The Authors. Journal of Advances in Modeling Earth Systems published by Wiley Periodicals LLC on behalf of American Geophysical Union.

Understanding of terrestrial carbon and water cycles is currently hampered by an uncertainty in how to capture the large variety of plant responses to drought. In FLUXNET, the global network of CO2 and H2O flux observations, many sites do not uniformly report the ancillary variables needed to study drought response physiology. To this end, we outline two data-driven indicators based on diurnal energy, water, and carbon flux patterns derived directly from the eddy covariance data and based on theorized physiological responses to hydraulic and non-stomatal limitations. Hydraulic limitations (i.e. intra-plant limitations on water movement) are proxied using the relative diurnal centroid (CET*), which measures the degree to which the flux of evapotranspiration (ET) is shifted toward the morning. Non-stomatal limitations (e.g. inhibitions of biochemical reactions, RuBisCO activity, and/or mesophyll conductance) are characterized by the Diurnal Water–Carbon Index (DWCI), which measures the degree of coupling between ET and gross primary productivity (GPP) within each day. As a proof of concept we show the response of the metrics at six European sites during the 2003 heat wave event, showing a varied response of morning shifts and decoupling. Globally, we found indications of hydraulic limitations in the form of significantly high frequencies of morning-shifted days in dry/Mediterranean climates and savanna/evergreen plant functional types (PFTs), whereas high frequencies of decoupling were dominated by dry climates and grassland/savanna PFTs indicating a prevalence of non-stomatal limitations in these ecosystems. Overall, both the diurnal centroid and DWCI were associated with high net radiation and low latent energy typical of drought. Using three water use efficiency (WUE) models, we found the mean differences between expected and observed WUE to be −0.09 to 0.44 µmol mmol−1 and −0.29 to −0.40 µmol mmol−1 for decoupled and morning-shifted days, respectively, compared to mean differences −1.41 to −1.42 µmol mmol−1 in dry conditions, suggesting that morning shifts/hydraulic responses are associated with an increase in WUE, whereas decoupling/non-stomatal limitations are not.