Bertrand Bonan

The knowledge of land-surface variables (LSVs) is essential for an accurate description of the carbon cycle and, hence, for deriving the fraction of anthropogenic CO ₂ emissions contributing to global warming. For a better representation of LSVs like the leaf-area index (LAI), we assimilate the observations of the satellite microwave sensors SMAP and AMSR2 with our land-data assimilation system (LDAS) on global cropland. Dedicated observation operators using a neural network (NN) are developed to enable a direct assimilation of the measured brightness temperatures. Direct assimilation is not yet an established method due to the incomplete representation of physical emission processes and the associated computational constraints of physical observation operators. We derive an optimal set of predictors for both instruments resulting in a good match between model equivalent and the observations in the testing period, with correlations up to 0.87/0.93 and a total RMSE of 8.2/3.6 K for SMAP/AMSR2, respectively. The implementation of the derived weights into the LDAS is straightforward and is found to lead to a reasonably good performance of the assimilation system, with a stronger improvement of the departures for SMAP than for AMSR2, which can be attributed to the characteristics of the microwave bands of the observations. For the best assimilation experiments, the verification against LAI observations shows an improvement compared to the open loop on a global scale. In agricultural regions of the world that are particularly susceptible to drought, AMSR2 experiments outperform SMAP experiments in terms of LAI in most cases. We also investigated the impact on the Gross Primary Productivity (GPP) and soil-water index (SWI) anomalies and found relatively little impact with mixed results. Nevertheless, the seasonal and sub-seasonal variability is still not well represented due to systematic errors in our land surface model.

This work integrates new satellite observations into the Interactions between Soil, Biosphere and Atmosphere (ISBA) land surface model via data assimilation. This improves our understanding of the impact of droughts on forests and agricultural land at global and regional scales. First, innovative artificial intelligence methods are employed to generate observation operators, enabling ISBA to simulate satellite microwave radiances ranging from the L to X bands, as well as sun‐induced fluorescence (SIF) observations. Second, these observations are assimilated into ISBA sequentially. Examples of sequential data assimilation are presented for SIF and C‐band backscatter observations derived from the tropospheric monitoring instrument and the advanced scatterometer on board the Sentinel‐5P and MetOp satellites respectively. We also demonstrate how machine learning can be used to derive new variables from those calculated by the land surface model in order to monitor natural disasters related to drought, as well as to characterise the regional exposure to drought‐related hazards. We present a prototype regional application for monitoring and forecasting clay shrink–swell and wildfire danger in France.

Live Fuel Moisture Content (LFMC) is one of the most critical variables for understanding fire dynamics, particularly within forest environments. This study intends to develop a daily LFMC indicator to support operational fire danger management services in France. The product is based on in situ observations from the French National Forest Office and is generated using a lightweight expressive neural network model. The network has been designed to generalise well over time and space. It can be integrated directly into land surface models to enable real-time monitoring of vegetation’s hydric status. The modelling framework combines outputs from a physically based land surface model and satellite-derived leaf area index (LAI) observations, providing high-resolution, spatially consistent estimates of land surface over France. To evaluate the model’s generalisation capacity, we implemented complementary cross-validation strategies to test interannual robustness, spatial transferability, and to simulate an operational deployment scenario. Additionally, we performed a robustness analysis to quantify the sensitivity of predictions to training variability. The results demonstrate a strong ability to estimate the range and dynamics of LFMC across most of France. They also identify regions where additional in situ sampling or improved representation could reduce epistemic uncertainty and enhance the reliability of the model.

Accurate representations of the surface and vegetation are critical for simulating the terrestrial CO ₂ cycle in response to climate and meteorological conditions. To meet this challenge, an increasing number of satellite missions are being launched which can monitor vegetation conditions and biomass. One is the Copernicus Sentinel-5P mission, which carries the TROPOMI instrument and retrieves Sun-Induced Chlorophyll Fluorescence (SIF). As an indicator of plant photosynthetic activity, SIF can provide critical information for evaluating and parameterising gross carbon flux dynamics in surface models. This study aims to assimilate TROPOMI SIF data into the ISBA (Interactions between Soil, Biosphere and Atmosphere) land surface model developed by Meteo-France, with the objective of directly correcting the representation of Leaf Area Index (LAI) and Gross Primary Production (GPP). To achieve this, we have developed a dedicated observation operator that links the modelled LAI to the TROPOMI SIF daily product. This neural network (NN) operator was developed using deep learning and was trained using observations over Europe. This operator achieved good accuracy, was implemented in a land data assimilation system (LDAS), and was used to assimilate TROPOMI SIF in ISBA using a sequential simplified extended Kalman filter. Specific experiments were conducted to study the assimilation process over the Ebro basin in Spain. This area is known for its irrigated croplands, which are not well represented by ISBA. Some experiments assimilate TROPOMI SIF, some assimilate a Copernicus Land Monitoring Service (CLMS) LAI 10 d product, and some assimilate both. This provided us with a useful point of reference for improving the vegetation simulation. Following SIF assimilation, LAI representation improved across the domain, highlighting heavily irrigated croplands. The gross primary production (GPP) derived from the analysis is closing the gap between the simulated and observed values, though a marked difference remains. When compared with other assimilation experiments, assimilating SIF alone provides a similar benefit to standard assimilation of a CLMS LAI product on LAI and GPP. The best improvements to the LAI and GPP results come from co-assimilating TROPOMI SIF with the CLMS LAI product, which combines the advantages of high-frequency SIF observations and robust 10 d LAI assimilation.

A recent study (Shan et al., 2024, https://doi.org/10.1016/j.rse.2024.114167 ) showed that using a Deep‐Neural‐Network (DNN)‐based observation operator in land data assimilation (DA) does not guarantee improvements in surface soil moisture (WG2). Here, we conduct a synthetic experiment to explain this outcome by testing whether DNN‐based operators reproduce physically consistent Jacobians (sensitivities) required by DA. The ISBA‐A‐gs land surface model is perturbed to generate “synthetic true” WG2 and leaf area index (LAI), and a Water Cloud Model (WCM) is used to generate synthetic backscatter. Two DNNs are trained taking simulated states from the open loop run as input and the synthetic backscatter as output. The first is trained on WG2 and LAI while the second uses LAI and soil moisture from multiple layers, that is including redundant inputs. The synthetic observations are then assimilated using the two DNNs and the WCM as observation operators. Results suggest that the assimilation using a DNN‐based observation operator improves the estimates of WG2. However, DNN Jacobians are no longer physically plausible when the model simulations used as training inputs contain errors relative to the “true” data, or when redundant input variables are included. This confirms that a strong predictive performance does not automatically reflect accurate representation of underlying physical sensitivities which prove problematic in subsequent DA. This study has important implications for deep‐learning‐based DA and Earth system models (ESM): DA or ESM trained on reanalysis or simulations may learn inaccurate Jacobians if input contains errors and redundant variables. This potential limitation could remain hidden when evaluation focuses primarily on predictive performance.

An accurate representation of the land surface is essential for simulating the exchange of energy, water and carbon between the land and the atmosphere. This study evaluates the impact of land cover (LC) representation on snow simulations in the Interactions Between Soil, Biosphere and Atmosphere (ISBA) land surface model in Europe between 2010 and 2022. The study employs the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 atmospheric forcing dataset. Offline simulation experiments were conducted using two different versions of the model to prescribe LC. The most recent version uses the latest LC data from the European Space Agency's (ESA) Climate Change Initiative (CCI). The model's ability to reproduce snow dynamics was evaluated through a comparison of the simulations with ESA CCI satellite snow water equivalent (SWE) and land surface temperature (LST) retrievals and ERA5 snow analyses. The ERA5 analysis shows the highest level of agreement with satellite-derived SWE at the domain scale. On average, both the ERA5 and ISBA simulations tend to overestimate SWE compared to the CCI SWE. However, it is also possible that the CCI SWE product underestimates the actual SWE. This bias is particularly large during the warm winter of 2020, while the scaled SWE anomalies are comparable to those observed by ESA CCI and ERA5. Using ESA CCI LC data reduces the ISBA SWE bias by around 23 %, with this reduction being observed over most of the domain. Further analysis of LC transitions shows that the reduction in SWE bias is driven primarily by a few vegetation changes, particularly the transition from grasslands to forests. Changes in SWE are also found in areas where the dominant LC remains unchanged. This indicates that modifications in sub-grid vegetation fractions can affect snow-vegetation interactions and SWE bias. The updated LC has a very limited impact on the model's performance for land surface temperature, indicating that the impact of LC updates is more noticeable for SWE than for LST. These findings emphasise the importance of accurate land cover data for improving snow representation in land surface models and highlight the need for updated vegetation information in future snow-related applications.

Live Fuel Moisture Content (LFMC) is one of the most critical variables for understanding fire dynamics, particularly within forest environments. This study intends to develop a daily LFMC indicator to support operational fire danger management services in France. The product is based on in situ observations from the French National Forest Office and is generated using a lightweight expressive neural network model. The network has been designed to generalise well over time and space. It can be integrated directly into land surface models to enable real-time monitoring of vegetation’s hydric status. The modelling framework combines outputs from a physically based land surface model and satellite-derived leaf area index (LAI) observations, providing high-resolution, spatially consistent estimates of land surface over France. To evaluate the model’s generalisation capacity, we implemented complementary cross-validation strategies to test interannual robustness, spatial transferability, and to simulate an operational deployment scenario. Additionally, we performed a robustness analysis to quantify the sensitivity of predictions to training variability. The results demonstrate a strong ability to estimate the range and dynamics of LFMC across most of France. They also identify regions where additional in situ sampling or improved representation could reduce epistemic uncertainty and enhance the reliability of the model.

The knowledge of land-surface variables (LSVs) is essential for an accurate description of the carbon cycle and, hence, for deriving the fraction of anthropogenic CO₂ emissions contributing to global warming. For a better representation of LSVs like the leaf-area index (LAI), we assimilate the observations of the satellite microwave sensors SMAP and AMSR2 with our land-data assimilation system (LDAS). Dedicated observation operators using a neural network (NN) are developed to enable a direct assimilation of the measured brightness temperatures. Direct assimilation is not yet an established method due to the incomplete representation of physical emission processes and the associated computational constraints of physical observation operators. We derive an optimal set of predictors for both instruments resulting in a good match between model equivalent and the observations in the testing period, with correlations up to 0.87/0.93 and a total RMSE of 8.2/3.6 K for SMAP/AMSR2, respectively. The implementation of the derived weights into the LDAS is straightforward and is found to lead to a reasonably good performance of the assimilation system, with a stronger improvement of the departures for SMAP than for AMSR2, which can be attributed to the characteristics of the microwave bands of the observations. For the best assimilation experiments, the verification against LAI observations shows an improvement compared to the open loop on a global scale. For different global cropland regions especially prone to droughts, AMSR2 outperforms SMAP in most cases. Nevertheless, the seasonal and subseasonal variability is still not well represented though due to unsolved issues in our model.

Biogenic volatile organic compound (BVOC) emissions from vegetation represent a major source of volatile compounds globally and play an important role as precursors for tropospheric ozone. Understanding their emissions is therefore crucial for quantifying the impact of ozone on air quality. We present two datasets of biogenic volatile organic compound emissions that cover the European modelling domain of the Copernicus Atmospheric Monitoring Service at a resolution of 0.1° × 0.1° to support the study of European scale air quality. The compounds included in the dataset follow the VOCs included in the regional atmospheric chemistry model mechanism (RACM). The datasets were produced within the framework of the EU's SEEDS project. We produced each dataset by coupling modelling output variables from the SURFEX land surface model with the MEGAN3.0 BVOC emission model. In one instance, the SURFEX model was run in free-running mode, which we term the open-loop (OL) and in the other case we assimilated satellite observations of leaf area index (LAI), which we term the analysis. The OL and analysis land surface model outputs form the basis for each emission dataset that are called SURFEX-MEGAN3.0 OL (https://doi.org/10.7910/DVN/LAUVTU, Hamer et al., 2025a) and SURFEX-MEGAN3.0 analysis (https://doi.org/10.7910/DVN/69G1FX, Hamer et al., 2025b), respectively. The OL dataset is available over a five-year period from 2018–2022 and the analysis dataset is available over the three-year period 2018–2020. SURFEX was run for both the OL and analysis simulations in a configuration that allowed simulated vegetation to respond to variations in meteorology over time to more realistically track vegetation phenology. Evaluation of the land surface model output LAI and root-zone soil moisture (RZSM) showed that the OL and analysis simulations had good skill at tracking temporal changes in both variables, with the analysis performing better in each instance. We perform a variety of evaluations on the isoprene emissions specifically given the importance of this compound for atmospheric chemistry. We evaluated the temporal variability of isoprene emissions in both datasets and found that the majority of the interannual and monthly variability was linked to variability in LAI that in specific cases, like the summer of 2019, could be linked to drought impacts on vegetation growth simulated by SURFEX. We evaluated the daily temporal variability of the OL and analysis isoprene emission datasets against in-situ online observations of isoprene concentrations at 8 sites in western Europe and found moderate to strong correlation between the emissions and observations in almost all location-year pairings. We also evaluated the OL and analysis emission datasets against other published bottom-up isoprene emission datasets over the same European domain used in this study. We found that the SURFEX-MEGAN3.0 OL and analysis isoprene emission datasets lie between the minimum (CAMS-GLOB-BIOv3.1) and maximum (MEGAN-MACC) published emission datasets based on bottom-up approaches. Furthermore, we were able to attribute differences in seasonality between SURFEX-MEGAN3.0 and other emission inventories to differences in the temporal variability of the underlying LAI dataset used to compile them. Overall, our findings show the importance of variability in LAI in controlling isoprene emissions on monthly to annual timescales. Combining this with the demonstrated skill of the emissions in evaluation with independent data, this points towards the value of an Earth-system approach to BVOC emission modelling.

An accurate representation of the land surface is essential for simulating the exchange of energy, water and carbon between the land and the atmosphere. This study evaluates the impact of land cover representation on snow simulations in the Interactions Between Soil, Biosphere and Atmosphere (ISBA) land surface model in Europe between 2010 and 2022. The study employs the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 atmospheric forcing dataset. Offline simulation experiments were conducted using two different versions of the model to prescribe land cover. The most recent version uses the latest land cover data from the European Space Agency's (ESA) Climate Change Initiative (CCI). The model's ability to reproduce snow dynamics was evaluated through a comparison of the simulations with ESA CCI satellite snow water equivalent (SWE) retrievals and ERA5 snow analyses. The ERA5 analysis shows the highest level of agreement with satellite observations of SWE at the domain scale. On average, both the ERA5 and ISBA simulations tend to overestimate SWE compared to the CCI SWE. However, it is also possible that the CCI SWE product underestimates the actual SWE. This bias is particularly large during the warm winter of 2020, while the scaled SWE anomalies are comparable to those observed by ESA CCI and ERA5. Using ESA CCI land cover data reduces the ISBA SWE bias by around 33 %, with this reduction being observed over most of the domain. These findings emphasise the importance of accurate land cover data for improving snow representation in land surface models and highlight the need for updated vegetation information in future snow-related applications.