Many different emission pathways exist that are compatible with the Paris climate agreement, and many more are possible that miss that target. While some of the most complex Earth System Models have simulated a small selection of Shared Socioeconomic Pathways, it is impractical to use these expensive models to fully explore the space of possibilities. Such explorations therefore mostly rely on one‐dimensional impulse response models, or simple pattern scaling approaches to approximate the physical climate response to a given scenario. Here we present ClimateBench—the first benchmarking framework based on a suite of Coupled Model Intercomparison Project, AerChemMIP and Detection‐Attribution Model Intercomparison Project simulations performed by a full complexity Earth System Model, and a set of baseline machine learning models that emulate its response to a variety of forcers. These emulators can predict annual mean global distributions of temperature, diurnal temperature range and precipitation (including extreme precipitation) given a wide range of emissions and concentrations of carbon dioxide, methane and aerosols, allowing them to efficiently probe previously unexplored scenarios. We discuss the accuracy and interpretability of these emulators and consider their robustness to physical constraints such as total energy conservation. Future opportunities incorporating such physical constraints directly in the machine learning models and using the emulators for detection and attribution studies are also discussed. This opens a wide range of opportunities to improve prediction, robustness and mathematical tractability. We hope that by laying out the principles of climate model emulation with clear examples and metrics we encourage engagement from statisticians and machine learning specialists keen to tackle this important and demanding challenge.

Due to technical limitations, it is impossible to have high resolution in both spatial and temporal dimensions for current NDVI datasets. Therefore, several methods are developed to produce high resolution (spatial and temporal) NDVI time-series datasets, which face some limitations including high computation loads and unreasonable assumptions. In this study, an unmixing-based method, NDVI Linear Mixing Growth Model (NDVI-LMGM), is proposed to achieve the goal of accurately and efficiently blending MODIS NDVI time-series data and multi-temporal Landsat TM/ETM+ images. This method firstly unmixes the NDVI temporal changes in MODIS time-series to different land cover types and then uses unmixed NDVI temporal changes to predict Landsat-like NDVI dataset. The test over a forest site shows high accuracy (average difference: −0.0070; average absolute difference: 0.0228; and average absolute relative difference: 4.02%) and computation efficiency of NDVI-LMGM (31 seconds using a personal computer). Experiments over more complex landscape and long-term time-series demonstrated that NDVI-LMGM performs well in each stage of vegetation growing season and is robust in regions with contrasting spatial and spatial variations. Comparisons between NDVI-LMGM and current methods (i.e., Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM), Enhanced STARFM (ESTARFM) and Weighted Linear Model (WLM)) show that NDVI-LMGM is more accurate and efficient than current methods. The proposed method will benefit land surface process research, which requires a dense NDVI time-series dataset with high spatial resolution.

The observed decline of spring dust storms in Northeast Asia since the 1950s has been attributed to surface wind stilling. However, spring vegetation growth could also restrain dust storms through accumulating aboveground biomass and increasing surface roughness. To investigate the impacts of vegetation spring growth on dust storms, we examine the relationships between recorded spring dust storm outbreaks and satellite-derived vegetation green-up date in Inner Mongolia, Northern China from 1982 to 2008. We find a significant dampening effect of advanced vegetation growth on spring dust storms (r = 0.49, p = 0.01), with a one-day earlier green-up date corresponding to a decrease in annual spring dust storm outbreaks by 3%. Moreover, the higher correlation (r = 0.55, p < 0.01) between green-up date and dust storm outbreak ratio (the ratio of dust storm outbreaks to times of strong wind events) indicates that such effect is independent of changes in surface wind. Spatially, a negative correlation is detected between areas with advanced green-up dates and regional annual spring dust storms (r = −0.49, p = 0.01). This new insight is valuable for understanding dust storms dynamics under the changing climate. Our findings suggest that dust storms in Inner Mongolia will be further mitigated by the projected earlier vegetation green-up in the warming world.