A deep learning-based hybrid model of global terrestrial evaporation

Artificial intelligence and Eddy covariance: A review

The Eddy Covariance (EC) method allows for monitoring carbon, water, and energy fluxes between Earth's surface and atmosphere. Due to its varying interdependent data streams and abundance of data as a whole, EC is naturally suited to Artificial Intelligence (AI) approaches. The integration of AI and EC will likely play a crucial role in the climate change mitigation and adaptation goals defined in the Sustainable Development Goals (SDGs) of the Agenda 2030. To aid this, we present a scoping review in which the novelty of various AI techniques in monitoring fluxes through the EC method from the past two decades has been collected. Overall, we find a clear positive trend in the quantity of research in this area, particularly in the last five years. We also find a lack of uniformity in available techniques, due to the diverse technologies and variables employed across environmental conditions and ecosystems. We highlight the most applied Machine Learning (ML) models, over the 71 algorithms identified in the scoping review, such as Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN), Support Vector Regression (SVR), and K-Nearest Neigbor (KNN). We suggest that future progress in this field requires an international, collaborative effort involving computer scientists and ecologists. Modern Deep Learning (DL) techniques such as Transformers and generative AI must be investigated to find how they may benefit our field. A forward-looking strategy must be formed for the optimal utilization of AI combined with EC to define future actions in flux monitoring in the face of climate change.

Dryland self-expansion enabled by land-atmosphere feedbacks

Dryland expansion causes widespread water scarcity and biodiversity loss. Although the drying influence of global warming is well established, the role of existing drylands in their own expansion is relatively unknown. In this work, by tracking the air flowing over drylands, we show that the warming and drying of that air contributes to dryland expansion in the downwind direction. As they dry, drylands contribute less moisture and more heat to downwind humid regions, reducing precipitation and increasing atmospheric water demand, which ultimately causes their aridification. In ~40% of the land area that recently transitioned from a humid region into a dryland, self-expansion accounted for >50% of the observed aridification. Our results corroborate the urgent need for climate change mitigation measures in drylands to decelerate their own expansion.

Cost, market, and policy constraints on mitigating climate change through afforestation in China

Afforestation is a promising nature-based climate solution for mitigating climate change, but it is subject to a complex web of biophysical, cost-benefit, market, and policy processes. Although its biophysical feasibility has been established, the cost, market, and policy constraints that affect climate change mitigation through afforestation are still unclear. Here, we estimate such cost, market, and policy constraints on the basis of biophysical feasibility. Our findings reveal that implementation costs are a more relevant constraint than opportunity costs on mitigating climate change through afforestation. The China Certified Emission Reduction market currently provides only a 0.308 % incentive for climate change mitigation through afforestation, due to market access constraints. The current market prices of China Certified Emission Reduction, China Carbon Emissions Trading Exchange, and Nature Based Carbon Offset in Voluntary Carbon Market constrain 88.15 %, 87.95 %, and 85.75 % of CO2 removal actions through afforestation, compared to the carbon price scenario (US$62.97 tCO2-1) of the EU Emissions Trading System. Moreover, land policy under the scenarios of prohibiting conversion of cultivated land to forest and forest restoration in degraded areas exhibit 8.87-29.59 % and 65.16-74.10 % constraints, respectively, on mitigating climate change through afforestation compared to land-use freedom conversion scenarios from 2020 to 2060. Thus, enhancing the incentive price of CO2 removal, addressing the market access barrier, strengthening cooperation between global carbon markets, and exploring carbon-neutral and food multi-oriented land policies can be valuable sources of mitigation efforts over the next 40 years.

Diminishing carryover benefits of earlier spring vegetation growth

Spring vegetation growth can benefit summer growth by increasing foliage area and carbon sequestration potential, or impair it by consuming additional resources needed for sustaining subsequent growth. However, the prevalent driving mechanism and its temporal changes remain unknown. Using satellite observations and long-term atmospheric CO2 records, here we show a weakening trend of the linkage between spring and summer vegetation growth/productivity in the Northern Hemisphere during 1982-2021. This weakening is driven by warmer and more extreme hot weather that becomes unfavourable for peak-season growth, shifting peak plant functioning away from earlier periods. This is further exacerbated by seasonally growing ecosystem water stress due to reduced water supply and enhanced water demand. Our finding suggests that beneficial carryover effects of spring growth on summer growth are diminishing or even reversing, acting as an early warning sign of the ongoing shift of climatic effects from stimulating to suppressing plant photosynthesis during the early to peak seasons.

Vegetation as the catalyst for water circulation on global terrestrial ecosystem

Global climate change is expected to further intensify the global water cycle, leading to more rapid evaporation and more intense precipitation. At the same time, the growth and expansion of natural vegetation caused by climate change and human activities create potential conflicts between ecosystems and humans over available water resources. Clarifying how terrestrial ecosystem evapotranspiration responds to global precipitation and vegetation facilitates a better understanding of and prediction for the responses of global ecosystem energy, water, and carbon budgets under climate change. Relying on the spatial and temporal distribution of evapotranspiration, precipitation, and solar-induced chlorophyll fluorescence (SIF) from remote sensing platforms, we decouple the interaction mechanism of evapotranspiration, precipitation, and vegetation in linear and nonlinear scenarios using correlation and partial correlation analysis, multiple linear regression analysis, and binning. Major conclusions are as follows: (1) As a natural catalyst of the global water cycle, vegetation plays a crucial role in regulating the relationship between climate change and the water‑carbon-energy cycle. (2) Vegetation, a key parameter affecting the water cycle, participates in the entire water cycle process. (3) The increase in vegetation productivity and photosynthesis plays a dominant role in promoting evapotranspiration in vegetated areas, while the increase in precipitation dominates the promotion of evapotranspiration in non-vegetated areas.

Networking the forest infrastructure towards near real-time monitoring - A white paper

Forests account for nearly 90 % of the world's terrestrial biomass in the form of carbon and they support 80 % of the global biodiversity. To understand the underlying forest dynamics, we need a long-term but also relatively high-frequency, networked monitoring system, as traditionally used in meteorology or hydrology. While there are numerous existing forest monitoring sites, particularly in temperate regions, the resulting data streams are rarely connected and do not provide information promptly, which hampers real-time assessments of forest responses to extreme climate events. The technology to build a better global forest monitoring network now exists. This white paper addresses the key structural components needed to achieve a novel meta-network. We propose to complement - rather than replace or unify - the existing heterogeneous infrastructure with standardized, quality-assured linking methods and interacting data processing centers to create an integrated forest monitoring network. These automated (research topic-dependent) linking methods in atmosphere, biosphere, and pedosphere play a key role in scaling site-specific results and processing them in a timely manner. To ensure broad participation from existing monitoring sites and to establish new sites, these linking methods must be as informative, reliable, affordable, and maintainable as possible, and should be supplemented by near real-time remote sensing data. The proposed novel meta-network will enable the detection of emergent patterns that would not be visible from isolated analyses of individual sites. In addition, the near real-time availability of data will facilitate predictions of current forest conditions (nowcasts), which are urgently needed for research and decision making in the face of rapid climate change. We call for international and interdisciplinary efforts in this direction.

Recent global decline in rainfall interception loss due to altered rainfall regimes

Evaporative loss of interception (Ei) is the first process occurring during rainfall, yet its role in large-scale surface water balance has been largely underexplored. Here we show that Ei can be inferred from flux tower evapotranspiration measurements using physics-informed hybrid machine learning models built under wet versus dry conditions. Forced by satellite and reanalysis data, this framework provides an observationally constrained estimate of Ei, which is on average 84.1 ± 1.8 mm per year and accounts for 8.6 ± 0.2% of total rainfall globally during 2000-2020. Rainfall frequency regulates long-term average Ei changes, and rainfall intensity, rather than vegetation attributes, determines the fraction of Ei in gross precipitation (Ei/P). Rain events have become less frequent and more intense since 2000, driving a global decline in Ei (and Ei/P) by 4.9% (6.7%). This suggests that ongoing rainfall changes favor a partitioning towards more soil moisture and runoff, benefiting ecosystem functions but simultaneously increasing flood risks.

An Outlook for Deep Learning in Ecosystem Science

Rapid advances in hardware and software, accompanied by public- and private-sector investment, have led to a new generation of data-driven computational tools. Recently, there has been a particular focus on deep learning—a class of machine learning algorithms that uses deep neural networks to identify patterns in large and heterogeneous datasets. These developments have been accompanied by both hype and scepticism by ecologists and others. This review describes the context in which deep learning methods have emerged, the deep learning methods most relevant to ecosystem ecologists, and some of the problem domains they have been applied to. Deep learning methods have high predictive performance in a range of ecological contexts, leveraging the large data resources now available. Furthermore, deep learning tools offer ecosystem ecologists new ways to learn about ecosystem dynamics. In particular, recent advances in interpretable machine learning and in developing hybrid approaches combining deep learning and mechanistic models provide a bridge between pure prediction and causal explanation. We conclude by looking at the opportunities that deep learning tools offer ecosystem ecologists and assess the challenges in interpretability that deep learning applications pose.

Spatiotemporal Changes and Driver Analysis of Ecosystem Respiration in the Tibetan and Inner Mongolian Grasslands

Ecosystem respiration (RE) plays a critical role in terrestrial carbon cycles, and quantification of RE is important for understanding the interaction between climate change and carbon dynamics. We used a multi-level attention network, Geoman, to identify the relative importance of environmental factors and to simulate spatiotemporal changes in RE in northern China’s grasslands during 2001–2015, based on 18 flux sites and multi-source spatial data. Results indicate that Geoman performed well (R2 = 0.87, RMSE = 0.39 g C m−2 d−1, MAE = 0.28 g C m−2 d−1), and that grassland type and soil texture are the two most important environmental variables for RE estimation. RE in alpine grasslands showed a decreasing gradient from southeast to northwest, and that of temperate grasslands showed a decreasing gradient from northeast to southwest. This can be explained by the enhanced vegetation index (EVI), and soil factors including soil organic carbon density and soil texture. RE in northern China’s grasslands showed a significant increase (1.81 g C m−2 yr−1) during 2001–2015. The increase rate of RE in alpine grassland (2.36 g C m−2 yr−1) was greater than that in temperate grassland (1.28 g C m−2 yr−1). Temperature and EVI contributed to the interannual change of RE in alpine grassland, and precipitation and EVI were the main contributors in temperate grassland. This study provides a key reference for the application of advanced deep learning models in carbon cycle simulation, to reduce uncertainties and improve understanding of the effects of biotic and climatic factors on spatiotemporal changes in RE.