Search

Fire is a crucial factor in terrestrial ecosystems playing a role in disturbance for vegetation dynamics. Process‐based fire models quantify fire disturbance effects in stand‐alone dynamic global vegetation models (DGVMs) and their advances have incorporated both descriptions of natural processes and anthropogenic drivers. Nevertheless, these models show limited skill in modeling fire events at the global scale, due to stochastic characteristics of fire occurrence and behavior as well as the limits in empirical parameterizations in process‐based models. As an alternative, machine learning has shown the capability of providing robust diagnostics of fire regimes. Here, we develop a deep‐learning‐based fire model (DL‐fire) to estimate daily burnt area fraction at the global scale and couple it within JSBACH4, the land surface model used in the ICON‐ESM. The stand‐alone DL‐fire model forced with meteorological, terrestrial and socio‐economic variables is able to simulate global total burnt area, showing 0.8 of monthly correlation (rm) with GFED4 during the evaluation period (2011–2015). The performance remains similar with the hybrid modeling approach JSB4‐DL‐fire (rm = 0.79) outperforming the currently used uncalibrated standard fire model in JSBACH4 (rm = −0.07). We further quantify the importance of each predictor by applying layer‐wise relevance propagation (LRP). Overall, land properties, such as fuel amount and water content in soil layers, stand out as the major factors determining burnt fraction in DL‐fire, paralleled by meteorological conditions over tropical and high latitude regions. Our study demonstrates the potential of hybrid modeling in advancing fire prediction in ESMs by integrating deep learning approaches in physics‐based dynamical models. Plain Language Summary: We develop a fire‐vegetation model based on a hybrid approach integrating artificial intelligence (AI) techniques into physics‐based models. Given the weather conditions, vegetation states, and human factors, our model estimates daily burned area fraction. The spatiotemporal variations in burned area are closely reproduced, especially over fire‐prone regions, such as Africa, South America, and Australia. Our model is able to represent regional variations in the drivers of fire occurrence, showing different importance of input predictors for different regions. This approach shows the possibilities of using deep learning (DL) models to provide in‐depth fire predictions in Earth system models. Key Points: Deep neural networks (DNN) can accurately predict global burnt area fraction on a daily scaleIntegration of the DNN into a physics‐based land model significantly improves the estimation of biomass burnt damage in vegetation dynamicsThe DNN accounts for regional fire variations by assigning varying degrees of importance to each predictor [ABSTRACT FROM AUTHOR]

The evaluation of decadal climate predictions against observations is crucial for their benefit to stakeholders. While the skill of such forecasts has been verified for several atmospheric variables, land hydrological states such as terrestrial water storage (TWS) have not been extensively investigated yet due to a lack of long observational records. Anomalies ofTWSare globally observed with the satellite missionsGRACE(2002-2017) andGRACE-FO(since 2018). By means of a GRACE-like reconstruction of TWS available over 41 years, we demonstrate that this data type can be used to evaluate the skill of decadal prediction experiments made available from different Earth system models as part of both CMIP5 and CMIP6. Analysis of correlation and root-mean-square deviation (RMSD) reveals that for the global land average the initialized simulations outperform the historical experiments in the first three forecast years. This predominance originates mainly from equatorial regions where we assume a longer influence of initialization due to longer soil memory times. Evaluated for individual grid cells, the initialization has a largely positive effect on the forecast year 1 TWS states; however, a general grid-scale prediction skill for TWS of more than 2 years could not be identified in this study for CMIP5. First results from decadal hindcasts of three CMIP6 models indicate a predictive skill comparable to CMIP5 for the multimodel mean in general, and a distinct positive influence of the improved soil-hydrology scheme implemented in the MPIESM for CMIP6 in particular. [ABSTRACT FROM AUTHOR]

Abstract: We evaluate the impact of a new five‐layer soil‐hydrology scheme on seasonal hindcast skill of 2 m temperatures over Europe obtained with the Max Planck Institute Earth System Model (MPI‐ESM). Assimilation experiments from 1981 to 2010 and 10‐member seasonal hindcasts initialized on 1 May each year are performed with MPI‐ESM in two soil configurations, one using a bucket scheme and one a new five‐layer soil‐hydrology scheme. We find the seasonal hindcast skill for European summer temperatures to improve with the five‐layer scheme compared to the bucket scheme and investigate possible causes for these improvements. First, improved indirect soil moisture assimilation allows for enhanced soil moisture‐temperature feedbacks in the hindcasts. Additionally, this leads to improved prediction of anomalies in the 500 hPa geopotential height surface, reflecting more realistic atmospheric circulation patterns over Europe. [ABSTRACT FROM AUTHOR]

Soil moisture-atmosphere feedback effects play an important role in several regions of the globe. For some of these regions, soil moisture memory may contribute significantly to the state and temporal variation of the regional climate. Identifying those regions can help to improve predictability in seasonal to decadal climate forecasts. In order to accurately simulate soil moisture memory and associated soil moisture-atmosphere interactions, an adequate representation of soil hydrology is required. The present study investigates how different setups of a soil hydrology scheme affect soil moisture memory simulated by the global climate model of the Max Planck Institute for Meteorology, ECHAM6/JSBACH. First, the standard setup is applied in which soil water is represented by a single soil moisture reservoir corresponding to the root zone. Second, a new five layer soil hydrology scheme is introduced where not only the root zone is differentiated into several layers but also layers below are added. Here, three variants of the new scheme are utilized to analyse how different characteristics of the soil hydrology and the associated fluxes influence soil moisture memory. Soil moisture memory of the different setups is analysed from global ECHAM6/JSBACH simulations forced by observed SST. Areas are highlighted where the regional climate seems to be sensitive to the improved representation of soil hydrology in the new setup and its variants. Results indicate that soil moisture memory is generally enlarged in regions during the dry season where a soil moisture buffer is present below the root zone due to the 5-layer scheme. This effect is usually enhanced when this buffer is increased. Memory tends to be weakened (strengthened) where bare soil evaporation is increased (decreased), especially in semi-arid regions and wet seasons. For some areas, this effect is compensated by a decreased (increased) transpiration. [ABSTRACT FROM AUTHOR]