Search

An accurate assessment of the physical states of the Earth system is an essential component of many scientific, societal and economical considerations. These assessments are becoming an increasingly challenging computational task since we aim to resolve models with high resolutions in space and time, to consider complex coupled partial differential equations, and to estimate uncertainties, which often requires many realizations. Machine learning methods are becoming a very popular method for the construction of surrogate models to address these computational issues. However, they also face major challenges in producing explainable, scalable, interpretable and robust models. In this manuscript, we evaluate the perspectives of geoscience applications of physics-based machine learning, which combines physics-based and data-driven methods to overcome the limitations of each approach taken alone. Through three designated examples (from the fields of geothermal energy, geodynamics, and hydrology), we show that the non-intrusive reduced basis method as a physics-based machine learning approach is able to produce highly precise surrogate models that are explainable, scalable, interpretable, and robust. [ABSTRACT FROM AUTHOR]

Land surface models (LSM) are an important tool for advancing our knowledge of the Earth system. LSM are constantly improved to represent the various terrestrial processes in more detail. High quality data, freely available from various observation networks, are providing being used to improve the prediction of terrestrial states and fluxes of water and energy. To optimize LSM with observations, data assimilation methods and tools have been developed in the past decades. We apply the coupled Community Land Model version 5 (CLM5) and Parallel Data Assimilation Framework (PDAF) system (CLM5-PDAF) for thirteen forest field sites throughout Europe covering different climate zones. The goal of this study is to assimilate in-situ soil moisture measurements into CLM5 to improve the modeled evapotranspiration fluxes. The modeled fluxes will be evaluated using the predicted evapotranspiration fluxes with eddy covariance (EC) systems. Most of the sites use point scale measurements from, however for three of the forest sites we use soil water content data from cosmic-ray neutron sensors, which have a measurement scale closer to the typical land surface model grid scale and EC footprint. Our results show that while data assimilation reduced the root-mean-square error for soil water content on average by 56 to 64 %, the root-mean-square error for the evapotranspiration estimation is increased by 4 %. This finding indicates that state-of-the-art LSM such as CLM5 still suffer from uncertainties in the representation of soil hydrological processes in forests, e.g. deep root water uptake, or highly uncertain vegetation parameters. [ABSTRACT FROM AUTHOR]

Long-range weather forecasts provide predictions of atmospheric, ocean and land surface conditions that can potentially be used in land surface and hydrological models to predict the water and energy status of the land surface or in crop growth models to predict yield for water resources or agricultural planning. However, the coarse spatial and temporal resolutions of available forecast products have hindered their widespread use in such modelling applications that usually require high resolution input data. In this study, we applied sub-seasonal (up to 4 months) and seasonal (7 months) weather forecasts from the latest European Centre for Medium-Range Weather Forecasts (ECMWF) seasonal forecasting system (SEAS5) in a land surface modelling approach using the Community Land Model version 5.0 (CLM5). Simulations were conducted for 2017-2020 forced with subseasonal and seasonal weather forecasts over two different domains with contrasting climate and cropping conditions: the German state of North Rhine-Westphalia and the Australian state of Victoria. We found that, after pre-processing of the forecast products (temporal downscaling of precipitation and incoming shortwave radiation), the simulations forced with seasonal and sub-seasonal forecasts were able to generate a model system response very close to reference simulation results forced by reanalysis data. Differences between seasonal and sub-seasonal experiments were insignificant. The forecast experiments were able to satisfactorily capture recorded inter-annual variations of crop yield. In addition, they also reproduced the generally higher inter-annual variability in crop yield across the Australian domain (approximately 50 % inter-annual variability in recorded yields and up to 17 % in simulated yields) compared to the German domain (approximately 15 % inter-annual variability in recorded yields and up to 5 % in simulated yields). The high and low yield seasons (2020 and 2018) among the four simulated years were clearly reproduced in forecast simulation results. Furthermore, sub-seasonal and seasonal simulations reflected the early harvest in the drought year of 2018 in the German domain. However, the simulated inter-annual yield variability was lower in all simulations compared to the official statistics. While general soil moisture trends, such as the European drought in 2018, were captured by the seasonal experiments, we found systematic over- and underestimations in both the forecast and the reference simulations compared to the Soil Moisture Active Passive Level-3 soil moisture product (SMAP L3) and the Soil Moisture Climate Change Initiative Combined dataset from the European Space Agency's (ESA CCI). These observed biases of soil moisture as well as the low inter-annual variability of simulated crop yield indicate the need to improve the representation of these variables in CLM5 to increase the model sensitivity to drought stress and other crop stressors. [ABSTRACT FROM AUTHOR]