Water resources from the Indus Basin sustain over 270 million people. However, water security in this region is threatened by climate change. This is especially the case for the upper Indus Basin, where most frozen water reserves are expected to decrease significantly by the end of the century, leaving rainfall as the main driver of river flow. However, future precipitation estimates from global climate models differ greatly for this region. To address this uncertainty, this paper explores the feasibility of using probabilistic machine learning to map large-scale circulation fields, better represented by global climate models, to local precipitation over the upper Indus Basin. More specifically, Gaussian processes are trained to predict monthly ERA5 precipitation data over a 15-year horizon. This paper also explores different Gaussian process model designs, including a non-stationary covariance function to learn complex spatial relationships in the data. Going forward, this approach could be used to make more accurate predictions from global climate model outputs and better assess the probability of future precipitation extremes.

Accurate prediction of nondispatchable renewable energy sources is essential for grid stability and price prediction. Regional power supply forecasts are usually indirect through a bottom-up approach of plant-level forecasts, incorporate lagged power values, and do not use the potential of spatially resolved data. This study presented a comprehensive methodology for predicting solar and wind power production at a country scale in France using machine learning models trained with spatially explicit weather data combined with spatial information about production sites’ capacity. A dataset is built spanning from 2012 to 2023, using daily power production data from Réseau de Transport d’Electricité (the national grid operator) as the target variable, with daily weather data from ECMWF Re-Analysis v5, production sites capacity and location, and electricity prices as input features. Three modeling approaches are explored to handle spatially resolved weather data: spatial averaging over the country, dimension reduction through principal component analysis, and a computer vision architecture to exploit complex spatial relationships. The study benchmarks state-of-the-art machine learning models as well as hyperparameter tuning approaches based on cross-validation methods on daily power production data. Results indicate that cross-validation tailored to time series is best suited to reach low error. We found that neural networks tend to outperform traditional tree-based models, which face challenges in extrapolation due to the increasing renewable capacity over time. Model performance ranges from 4% to 10% in normalized root-mean-squared error for midterm horizon, achieving similar error metrics to local models established at a single-plant level, highlighting the potential of these methods for regional power supply forecasting.

Good air quality is a critical determinant of public health, influencing life expectancy, respiratory health, work productivity, and the prevention of chronic diseases. This study presents a novel approach to classifying the Air Quality Index (AQI) using deep learning techniques, specifically convolutional neural networks (CNNs). We collected and curated a dataset comprising 11,000 digital images from three distinct regions in Indonesia—Jakarta, Malang, and Semarang—ensuring uniformity through standardized acquisition settings. The images were categorized into four air quality classes: good, moderate, unhealthy for sensitive groups, and unhealthy. We designed and implemented a CNN architecture optimized for AQI classification. The model achieved an impressive accuracy of 99.81% using K-fold cross-validation. In addition, the model’s interpretative capabilities were examined using techniques such as Grad-CAM, providing valuable insights into how the CNN identifies and classifies air quality conditions based on image features. These findings underscore the effectiveness of CNNs for AQI classification and highlight the potential for future work to incorporate a more diverse set of digital images captured from various perspectives to enhance dataset complexity and model robustness. The dataset is publicly accessible at https://doi.org/10.5281/zenodo.15727522.

Climate change poses an existential threat, necessitating effective climate policies to enact impactful change. Decisions in this domain are incredibly complex, involving conflicting entities and evidence. In the last decades, policymakers increasingly use simulations and computational methods to guide some of their decisions. Integrated Assessment Models (IAMs) are one of such methods, which combine social, economic, and environmental simulations to forecast potential policy effects. For example, the UN uses outputs of IAMs for their recent Intergovernmental Panel on Climate Change (IPCC) reports. Traditionally these have been solved using recursive equation solvers, but have several shortcomings, e.g. struggling at decision making under uncertainty. Recent preliminary work using Reinforcement Learning (RL) as an alternative to traditional solvers shows promising results in decision making in uncertain and noisy scenarios. We extend on this work by introducing multiple interacting RL agents as a preliminary analysis on modelling the complex interplay of socio-interactions between various stakeholders or nations that drives much of the current climate crisis. Our findings show that cooperative agents in this framework can consistently chart pathways towards more desirable futures in terms of reduced carbon emissions and improved economy. However, upon introducing competition between agents, for instance by using opposing reward functions, desirable climate futures are rarely reached. Modelling competition is key to increased realism in these simulations, as such we employ policy interpretation by visualizing what states lead to more uncertain behavior, to understand algorithm failure. Finally, we highlight the current limitations and avenues for further work to ensure future technology uptake for policy derivation.

We developed a cloud microphysics parameterization for the icosahedral nonhydrostatic modeling framework (ICON) model based on physics-informed machine learning (ML). By training our ML model on high-resolution simulation data, we enhance the representation of cloud microphysics in Earth system models (ESMs) compared to traditional parameterization schemes, in particular by considering the influence of high-resolution dynamics that are not resolved in coarse ESMs. We run a global, kilometer-scale ICON simulation with a one-moment cloud microphysics scheme, the complex graupel scheme, to generate 12 days of training data. Our ML approach combines a microphysics trigger classifier and a regression model. The microphysics trigger classifier identifies the grid cells where changes due to the cloud microphysical parameterization are expected. In those, the workflow continues by calling the regression model and additionally includes physical constraints for mass positivity and water mass conservation to ensure physical consistency. The microphysics trigger classifier achieves an F1 score of 0.93 on classifying unseen grid cells. The regression model reaches an $ {R}^2 $ score of 0.72 averaged over all seven microphysical tendencies on simulated days used for validation only. This results in a combined offline performance of 0.78. Using explainability techniques, we explored the correlations between input and output features, finding a strong alignment with the graupel scheme and, hence, physical understanding of cloud microphysical processes. This parameterization provides the foundation to advance the representation of cloud microphysical processes in climate models with ML, leading to more accurate climate projections and improved comprehension of the Earth’s climate system.

Visible satellite imagery (VIS) is essential for monitoring weather patterns and tracking ground surface changes associated with climate change. However, its availability is limited during nighttime. To address this limitation, we present a discrete variational autoencoder (VQVAE) method for translating infrared satellite imagery to VIS. This method departs from previous efforts that utilize a U-Net architecture. By removing the connections between corresponding layers of the encoder and decoder, the model learns a discrete and rich codebook of latent priors for the translation task. We train and test our model on mesoscale data from the Geostationary Operational Environmental Satellite (GOES) West Advanced Baseline Imager (ABI) sensor, spanning 4 years (2019 to 2022) using the Conditional Generative Adversarial Nets (CGAN) framework. This work demonstrates the practical use of a VQVAE for meteorological satellite image translation. Our approach provides a modular framework for data compression and reconstruction, with a latent representation space specifically designed for handling meteorological satellite imagery.

Climate change will impact wind and, therefore, wind power generation with largely unknown effects and magnitude. Climate models can provide insight and should be used for long-term power planning. In this work, we use Gaussian processes to predict power output given wind speeds from a global climate model. We validate the aggregated predictions from past climate model data with actual power generation, which supports using CMIP6 climate model data for multi-decadal wind power predictions and highlights the importance of being location-aware. We find that wind power projections for the two in-between climate scenarios, SSP2–4.5 and SSP3–7.0, closely align with actual wind power generation between 2015 and 2023. Our location-aware future predictions up to 2050 reveal only minor changes in yearly wind power generation. Our analysis also reveals larger uncertainty associated with Germany’s coastal areas in the North than Germany’s South, motivating wind power expansion in regions where the future wind is likely more reliable. Overall, our results indicate that wind energy will likely remain a reliable energy source.

Earth’s forests play an important role in the fight against climate change and are in turn negatively affected by it. Effective monitoring of different tree species is essential to understanding and improving the health and biodiversity of forests. In this work, we address the challenge of tree species identification by performing tree crown semantic segmentation using an aerial image dataset spanning over a year. We compare models trained on single images versus those trained on time series to assess the impact of tree phenology on segmentation performance. We also introduce a simple convolutional block for extracting spatio-temporal features from image time series, enabling the use of popular pretrained backbones and methods. We leverage the hierarchical structure of tree species taxonomy by incorporating a custom loss function that refines predictions at three levels: species, genus, and higher-level taxa. Our best model achieves a mean Intersection over Union (mIoU) of 55.97%, outperforming single-image approaches particularly for deciduous trees where phenological changes are most noticeable. Our findings highlight the benefit of exploiting the time series modality via our Processor module. Furthermore, leveraging taxonomic information through our hierarchical loss function often, and in key cases significantly, improves semantic segmentation performance.

We present a deep learning architecture that reconstructs a source of data at given spatio-temporal coordinates using other sources. The model can be applied to multiple sources in a broad sense: the number of sources may vary between samples, the sources can differ in dimensionality and sizes, and cover distinct geographical areas at irregular time intervals. The network takes as input a set of sources that each include values (e.g., the pixels for two-dimensional sources), spatio-temporal coordinates, and source characteristics. The model is based on the Vision Transformer, but separately embeds the values and coordinates and uses the embedded coordinates as relative positional embedding in the computation of the attention. To limit the cost of computing the attention between many sources, we employ a multi-source factorized attention mechanism, introducing an anchor-points-based cross-source attention block. We name the architecture MoTiF (multi-source transformer via factorized attention). We present a self-supervised setting to train the network, in which one source chosen randomly is masked and the model is tasked to reconstruct it from the other sources. We test this self-supervised task on tropical cyclone (TC) remote-sensing images, ERA5 states, and best-track data. We show that the model is able to perform TC ERA5 fields and wind intensity forecasting from multiple sources, and that using more sources leads to an improvement in forecasting accuracy.

Adaptation to climate change requires robust climate projections, yet the uncertainty in these projections performed by ensembles of Earth system models (ESMs) remains large. This is mainly due to uncertainties in the representation of subgrid-scale processes such as turbulence or convection that are partly alleviated at higher resolution. New developments in machine learning-based hybrid ESMs demonstrate great potential for systematically reduced errors compared to traditional ESMs. Building on the work of hybrid (physics + AI) ESMs, we here discuss the additional potential of further improving and accelerating climate models with quantum computing. We discuss how quantum computers could accelerate climate models by solving the underlying differential equations faster, how quantum machine learning could better represent subgrid-scale phenomena in ESMs even with currently available noisy intermediate-scale quantum devices, how quantum algorithms aimed at solving optimization problems could assist in tuning the many parameters in ESMs, a currently time-consuming and challenging process, and how quantum computers could aid in the analysis of climate models. We also discuss hurdles and obstacles facing current quantum computing paradigms. Strong interdisciplinary collaboration between climate scientists and quantum computing experts could help overcome these hurdles and harness the potential of quantum computing for this urgent topic.

The global food system puts enormous pressure on the environment. Managing these pressures requires understanding not only where they occur (i.e., where food is produced), but also who drives them (i.e., where food is consumed). However, the size and complexity of global supply chains make it difficult to trace food production to consumption. Here, we provide the most comprehensive dataset of bilateral trade flows of environmental pressures stemming from food production from producing to consuming nations. The dataset provides environmental pressures for greenhouse gas emissions, water use, nitrogen and phosphorus pollution, and the area of land/water occupancy of food production for crops and animals from land, freshwater, and ocean systems. To produce these data, we improved upon reported food trade and production data to identify producing and consuming nations for each food item, allowing us to match food flows with appropriate environmental pressure data. These data provide a resource for research on sustainable global food consumption and the drivers of environmental impact.

In time-dependent systems, autoregressive models are frequently employed to investigate the interactions between variables of interest in fields such as climate science, macroeconomics, and neuroscience. Typically, these variables are aggregated from smaller-scale variables into large-scale variables, for instance, representing modes of climate variability in climate science. A key aspect of these models is estimating the long-term effects of external perturbations, once the system stabilizes. Our primary contribution is an explicit formula for quantifying these long-term effects on small-scale variables, which is directly estimable from the model’s linear coefficients and aggregation weights. This improves traditional autoregressive models by providing a localized understanding of the system behavior. We conduct a series of numerical experiments to evaluate the performance of various methods to estimate perturbation effects from data. Our second contribution is the derivation of the asymptotic properties of these estimators under suitable assumptions. These asymptotic properties can be leveraged for uncertainty quantification. In a numerical experiment, we compare the uncertainty ranges of the proposed asymptotic-based approach with four bootstrap-based methods. Finally, we apply our methods to investigate the effects of economic activities on air pollution in Northern Italy, demonstrating their ability to reveal local effects. Our novel approach provides a comprehensive framework for analyzing the impacts of perturbations on both large- and small-scale variables, thereby enhancing our understanding of complex systems. Our research has implications for various disciplines where the study of perturbation effects is crucial for understanding and predicting systems’ behavior.

Sea surface salinity and temperature are essential climate variables in monitoring and modeling ocean health. Multispectral ocean color satellites allow the estimation of these properties at a resolution of 10 to 300 m, which is required to correctly represent their spatial variability in coastal waters. This paper investigates the effect of pre-applying an unsupervised classification in the performance of both temperature and salinity inversion. Two methodologies were explored: clustering based solely on spectral radiances, and clustering applied directly to satellite images. The former improved model generalization by identifying similar water clusters across different locations, reducing location dependency. It also demonstrated results correlating cluster type with salinity and temperature distributions thereby enhancing regression model performance and improving a global ocean color sea surface temperature regression model RMSE error by 10%. The latter approach, applying clustering directly to satellite images, incorporated spatial information into the models and enabled the identification of front boundaries and gradient information, improving global sea surface temperature models RMSE by 20% and sea surface salinity models by 30%, compared to the initial ocean color model. Beyond improving algorithm performance, optical water classification can be used to monitor and interpret changes to water optics, including algal blooms, sediment disturbance or other climate change or antropogenic disturbances. For example, the clusters have been used to show the impact of a category 4 hurricane landfall on the Mississippi estuarine region.

The rapid advancement of large language models (LLMs) has enabled their integration into a wide range of scientific disciplines. This article introduces a comprehensive benchmark dataset specifically designed for testing recent LLMs in the hydrology domain. Leveraging a collection of research articles and hydrology textbooks, we generated a wide array of hydrology-specific questions in various formats, including true/false, multiple-choice, open-ended, and fill-in-the-blank. These questions serve as a robust foundation for evaluating the performance of state-of-the-art LLMs, including GPT-4o-mini, Llama3:8B, and Llama3.1:70B, in addressing domain-specific queries. Our evaluation framework employs accuracy metrics for objective question types and cosine similarity measures for subjective responses, ensuring a thorough assessment of the models’ proficiency in understanding and responding to hydrological content. The results underscore both the capabilities and limitations of artificial intelligence (AI)-driven tools within this specialized field, providing valuable insights for future research and the development of educational resources. By introducing HydroLLM-Benchmark, this study contributes a vital resource to the growing body of work on domain-specific AI applications, demonstrating the potential of LLMs to support complex, field-specific tasks in hydrology.

How areas of land are allocated for different uses, such as forests, urban areas, and agriculture, has a large effect on the terrestrial carbon balance and, therefore, climate change. Based on available historical data on land-use changes and a simulation of the associated carbon emissions and removals, a surrogate model can be learned that makes it possible to evaluate the different options available to decision-makers efficiently. An evolutionary search process can then be used to discover effective land-use policies for specific locations. Such a system was built on the Project Resilience platform and evaluated with the Land-Use Harmonization dataset LUH2 and the bookkeeping model BLUE. It generates Pareto fronts that trade off carbon impact and amount of land-use change customized to different locations, thus providing a proof-of-concept tool that is potentially useful for land-use planning.

Artificial neural networks are increasingly used for geophysical modeling to extract complex nonlinear patterns from geospatial data. However, it is difficult to understand how networks make predictions, limiting trust in the model, debugging capacity, and physical insights. EXplainable Artificial Intelligence (XAI) techniques expose how models make predictions, but XAI results may be influenced by correlated features. Geospatial data typically exhibit substantial autocorrelation. With correlated input features, learning methods can produce many networks that achieve very similar performance (e.g., arising from different initializations). Since the networks capture different relationships, their attributions can vary. Correlated features may also cause inaccurate attributions because XAI methods typically evaluate isolated features, whereas networks learn multifeature patterns. Few studies have quantitatively analyzed the influence of correlated features on XAI attributions. We use a benchmark framework of synthetic data with increasingly strong correlation, for which the ground truth attribution is known. For each dataset, we train multiple networks and compare XAI-derived attributions to the ground truth. We show that correlation may dramatically increase the variance of the derived attributions, and investigate the cause of the high variance: is it because different trained networks learn highly different functions or because XAI methods become less faithful in the presence of correlation? Finally, we show XAI applied to superpixels, instead of single grid cells, substantially decreases attribution variance. Our study is the first to quantify the effects of strong correlation on XAI, to investigate the reasons that underlie these effects, and to offer a promising way to address them.

Emerging wildlife pathogens often display geographic variability due to landscape heterogeneity. Modeling approaches capable of learning complex, non-linear spatial dynamics of diseases are needed to rigorously assess and mitigate the effects of pathogens on wildlife health and biodiversity. We propose a novel machine learning (ML)-guided approach that leverages prior physical knowledge of ecological systems, using partial differential equations. We present our approach, taking advantage of the universal function approximation property of neural networks for flexible representation of the underlying dynamics of the geographic spread and growth of wildlife diseases. We demonstrate the benefits of our approach by comparing its forecasting power with commonly used methods and highlighting the obtained insights on disease dynamics. Additionally, we show the theoretical guarantees for the approximation error of our model. We illustrate the implementation of our ML-guided approach using data from white-nose syndrome (WNS) outbreaks in bat populations across the US. WNS is an infectious fungal disease responsible for significant declines in bat populations. Our results on WNS are useful for disease surveillance and bat conservation efforts. Our methods can be broadly used to assess the effects of environmental and anthropogenic drivers impacting wildlife health and biodiversity.

Aerosol-cloud interactions contribute significant uncertainty to modern climate model predictions. Analysis of complex observed aerosol-cloud parameter relationships is a crucial piece of reducing this uncertainty. Here, we apply two machine learning methods to explore variability in in-situ observations from the NASA ACTIVATE mission. These observations consist of flights over the Western North Atlantic Ocean, providing a large repository of data including aerosol, meteorological, and microphysical conditions in and out of clouds. We investigate this dataset using principal component analysis (PCA), a linear dimensionality reduction technique, and an autoencoder, a deep learning non-linear dimensionality reduction technique. We find that we can reduce the dimensionality of the parameter space by more than a factor of 2 and verify that the deep learning method outperforms a PCA baseline by two orders of magnitude. Analysis in the low dimensional space of both these techniques reveals two consistent physically interpretable regimes—a low pollution regime and an in-cloud regime. Through this work, we show that unsupervised machine learning techniques can learn useful information from in-situ atmospheric observations and provide interpretable results of low-dimensional variability.