Grasslands contribute greatly to increasing soil organic carbon (SOC) storage in cropland. Models can be used to predict effects of grasslands under different soil, weather, and management conditions. We calibrated the STICS soil-crop model to simulate the grassland yield and SOC dynamics of productive-grassland systems, taking into account the contribution of roots to soil carbon (C) inputs. We used observations from three contrasted long-term (13-27 years) French experiments that included different grassland durations and management practices (i.e. temporary/permanent, mown/grazed, fertilised/unfertilised). We optimised some root parameters using a subset of treatments from one of the three sites and then evaluated STICS using the remaining treatments. STICS accurately predicted the observed dynamics of grassland root carbon and nitrogen at one site, and the predicted ranges of root variables (e.g. root:shoot ratio and root C:N ratio) at the three sites were consistent with the literature. STICS satisfactorily predicted SOC dynamics for all three sites and treatments, with low relative error (nRMSE) and bias, ranging from 2.5 to 5.9% and -2.5 to 0.6 t C ha-1, respectively. It also reproduced the observed positive effect of grasslands on SOC stocks. Although STICS slightly underpredicted effects of the sites and treatments on grassland yield and nitrogen content, these predictions were considered satisfactory. Therefore, STICS could be used to predict net C footprints of cattle farms based on productive grasslands or to develop predictive metamodels of SOM dynamics, which would support implementation of better practices in livestock farming.

This work is devoted to the problem of clustering a set of samples according to the effect they have as one of the many varying inputs of a model. An example is the problem of clustering weather series according to the effect they have on the yield simulated with a crop model when also other inputs such as soil or plant parameters vary. We introduce both simple solutions based on the K-Means algorithms and the derivation of two possible formulations of the clustering problem in a sensitivity analysis framework. We show that these formulations coincide in the sense that their criteria are closely related to each other, leading to a possibility to use the K-Means algorithm with an improved expression of clustering performance through the use of Sobol' indices.

Near-real-time (NRT) daily crop monitoring at the field scale is crucial for precision agriculture, yet remains challenging due to limitations in the spatial or temporal resolution of existing remote sensing methods. While Sentinel-2 provides adequate spatial resolution for field-level applications, its temporal resolution is insufficient for capturing rapid crop dynamics, especially in cloudy regions. Existing spatiotemporal fusion techniques require multiple clear-sky images and lack true NRT capability, while ground-based sensors offer continuous monitoring but with limited spatial coverage. To address these limitations, this study develops the Near-RealTime Ground-Satellite Fusion (NRT-GSF) algorithm, a novel approach based on a Bayesian dynamic linear model and Kalman filtering. The algorithm uniquely integrates Sentinel-2 imagery with continuous measurements from Internet of Things for Agriculture (IoTA) systems to generate daily 10-m Green Area Index (GAI) products. Its recursive framework supports both forward prediction in NRT mode following satellite overpasses and backward updating to refine historical profiles. Implemented over French wheat fields using 34 IoTA systems and Sentinel-2 time series from 2019, the algorithm effectively enhanced spatiotemporal completeness and accuracy (R = 0.75-0.98, RMSE = 0.1-0.49). A comprehensive leave-one-out Sentinel-2 evaluation demonstrated its superiority over the current Consistent Adjustment of the Climatology to Actual Observations (CACAO) algorithm. Ground validation using handheld RGB cameras further confirmed the accuracy of the GAI products from the new algorithm (RMSE = 0.5). The NRT-GSF framework offers a robust and operationally solution for daily, high-resolution crop GAI mapping in NRT mode, and it can be extended to other traits or applications in the near-real-time context.

Crop simulation models are important tools in agronomy. Typically, they need to be calibrated before being used for new environments or cultivars. However, there is a large variability in calibration approaches, which contributes to uncertainty in simulated values, so it is important to develop improved calibration procedures that are widely applicable. The AgMIP calibration group recently proposed a comprehensive, generic calibration protocol that is directly based on standard statistical parameter estimation in regression models. Weighted least squares (WLS) is used to handle multiple response variables and forward regression using the corrected Akaike Information Criterion (AICc) is used to select the parameters to be calibrated. The protocol includes two adaptations, which are specific to each model and data set. First, initial approximations to the WLS parameters are obtained by fitting variables one group at a time. Secondly, “major” parameters are identified that are intended to reduce bias, analogously to the constant in linear regression. In this study, new diagnostic tools to be included in the protocol are proposed and tested in a case study. The diagnostics test whether the protocol does indeed lead to good initial approximations to the WLS parameters, and whether the protocol does indeed substantially reduce bias. These diagnostics provide in-depth understanding of the calibration process, reveal problems and help suggest solutions. The diagnostics should increase confidence in the results of the protocol. Having a reliable, generic calibration approach, like the augmented AgMIP protocol, is essential to using crop models more effectively.

Crop models need to be regularly upgraded with parametrization for new cultivars but this requires calibration, which is a major challenge. With winter wheat cultivar Rubisko as a case study, we propose to apply a calibration protocol to estimate the parameters of this new cultivar with multi-trials experimental data. We tested the calibration protocol in different conditions including or not LAI and/or biomass experimental data and we found that the resulting LAI and biomass dynamics strongly diverge. Several key findings emerge from this study: (1) RUE parameters should be excluded from the calibration process, as their critical role in biomass dynamics causes the optimization algorithm to treat them as adjustment parameters, resulting in unrealistic values for multiple parameters; (2) either LAI or biomass variables alone are sufficient for calibration, enabling experimental efforts to focus on one variable rather than both; and (3) the use of a synthetic dataset has facilitated the identification of the optimal type and timing of data collection needed to parameterize a new variety in the model. Moreover, the proposed methodology offers extrapolatable solutions applicable to other contexts (e.g., different models or datasets) and provides guidance on acquiring the most effective dataset for optimal calibration. The unbalanced structure of our dataset also highlighted the need to mobilize other calibration criteria (weighted RMSE) and alternative solutions to bridge the gap between quantitative metrics and empirical visual assessments. ### Competing Interest Statement The authors have declared no competing interest.

This chapter offers a general presentation of the STICS soil-crop model. Through numerous illustrations dealing notably with genetic x environment x management interactions, it presents an overview of the wide domain of validity of the model, along with its performances, demonstrating its potential for a wide range of agronomic and environmental applications. Since the beginning of its story in the early 90’s, STICS has been able to remain a generic and robust model that allows to simulate the functioning of agro-ecosystems, both in temperate and tropical environments. As highlighted in this chapter, and particularly in the last two sections, many efforts were, and are still, put to keep the model up-to-date and in a permanent state of evolution towards better representation of cropping systems, allowing to analyze novel research and applied questions.

Crop models need to be regularly updated with parameterizations for new cultivars, but this requires calibration, which is a major challenge. Using the winter wheat cultivar Rubisko as a case study, we applied for the first time on experimental data a new calibration protocol to estimate the parameters of the STICS crop model for this new cultivar with multi-trial experimental data. We tested the calibration protocol in different conditions, with or without LAI and/or biomass experimental data, and we found that the resulting LAI and biomass dynamics strongly diverged. This study contributes to provide guidance to modelers for the calibration of a new cultivar in a crop model by focusing on the selection of variables and parameters to estimate as well as criteria for evaluating calibration strategies. With an application to winter wheat for the STICS crop model, this study has shown that the choice of calibration steps has a major impact on simulated outputs, but with a strong dependence on the structure of the experimental dataset. Firstly, this paper provides a methodology for the selection of calibration variables and associated parameters based on three criteria: 1) the relevance of the values of the estimated parameters, 2) the bias part of the mean square error, and 3) the analysis of the residuals. Secondly, by applying this methodology, we have shown that calibration based on LAI measurements is the most robust in the case of sparse observed data at the end of the cycle. Based on these results, we recommend caution when including parameters related to radiation-use efficiency; in particular, they should not be calibrated together with parameters related to leaf growth on biomass data alone. This study has enabled an appropriate calibration strategy to be defined, which will allow more modern French wheat cultivars to be parameterized in the STICS crop model.

Accurate prediction of phenotypes across genotypes and environments is crucial for accelerating crop improvement. Process-based crop growth models (CGMs) can capture complex genotype-by-environment interactions, but their use is limited by labor-intensive genotypic parameter measurements. Here, we developed a faster data assimilation pipeline integrating high-throughput phenotyping (HTP) observations with the SiriusQuality wheat model to efficiently estimate key genotypic parameters and predict genotype performance. Using time-series RGB imagery from a ground-based Phenomobile, we assimilated intercepted photosynthetically active radiation (fIPAR), heading date, and final grain yield to jointly assimilated to calibrate twelve genotypic parameters governing phenology, canopy development, light interception, biomass accumulation, and grain filling. Two data assimilation strategies—a Bayesian DREAM(zs) algorithm and a lookup table (LUT) inversion—were compared through both in silico experiment and eight years of multi-environment field trials of nine durum wheat cultivars. The LUT method demonstrated superior computational efficiency, with prediction accuracy comparable to Bayesian inference on real field data. Multi-year field trials showed that two environments (year / site) were sufficient to reliably characterize genotypic parameters and predict performance across environments. By combining time-series HTP data with ecophysiological modeling, our data assimilation pipeline offers breeders a powerful tool for genotype characterization. It streamlines the process of capturing environmental variance and phenotypic stability, reducing time and effort in crop improvement.

It has become common to compare crop model results in multi-model simulation experiments. In general, one observes a large variability in such studies, which reduces the confidence one can have in such models. It is important to understand the causes of this variability as a first step toward reducing it. For a given data set, the variability in a multi-model study can arise from uncertainty in model structure or in parameter values for a given structure. Previous studies have made assumptions about the origin of parameter uncertainty, and then quantified its contribution, generally finding that parameter uncertainty is less important than structure uncertainty. However, those studies do not take account of the full parameter variability in multi-model studies. Here we propose estimating parameter uncertainty based on open-call multi-model ensembles where the same structure is used by more than one modeling group. The variability in such a case is due to the full variability of parameters among modeling groups. Then structure and parameter contributions can be estimated using random effects analysis of variance. Based on three multi-model studies for simulating wheat phenology, it is found that the contribution of parameter uncertainty to total uncertainty is, on average, more than twice as large as the uncertainty from structure. A second estimate, based on a comparison of two different calibration approaches for multiple models leads to a very similar result. We conclude that improvement of crop models requires as much attention to parameters as to model structure.

Evapotranspiration (ET) is a fundamental element of the hydrological cycle which plays a major role on surface water balance and surface energy balance. At local scale, ET can be estimated from detailed ground observations, for example using flux towers, but these measurements are only representative of very limited homogeneous area. When regional information is required, e.g. for monitoring ground water resources, ET can be mapped using thermal infrared and spectral reflectance data. Various ET models have been developed but there was no competitive evaluation of them over a large range of situations, so that it is not possible to evaluate the intrinsic performance of one model compared to another. In such situation, ensemble model averaging may provide a coherent estimation of ET with an increased overall accuracy. In this work the ensemble modelling approach is extended to a multi-model – multi-data framework that provides ET estimations together with an uncertainty of estimation.We developed the EVASPA framework for estimating ET through ensemble averaging with the objective of providing estimates of ET together with an estimation uncertainty. In this presentation we present a full analysis of the uncertainties of ET estimation in relation to uncertainties in input variables and models. Airborne remote sensing data were acquired over the Grosseto area in Italy in the frame of the ESA SurfSense experiment (high spatio-temporal Resolution Land Surface Temperature Experiment) in support of the LSTM mission project (Copernicus Land Surface Temperature Monitoring). Evapotranspiration was computed using two different types of models considering: -1) the evaporative fraction (EF) computed from the variability of surface temperature versus vegetation amount (fraction cover) or albedo over the investigated areas ('triangle' approach) and -2) the residual aerodynamic equation. Two types of uncertainties were computed: the ‘novice user’ uncertainty and the ‘expert user’ uncertainty which differed by the previous knowledge on the accuracy of input data and on the performances of models that was available to users. Evapotranspiration uncertainties ranged between 0.8 mm.d-1 (EF model, expert case) and 2.7 mm.d-1 (aerodynamic model, novice case). The analysis showed that the main uncertainty sources were related to model formulations (evaporative fraction calculation and ground heat flux calculation for both types of models) and to solar radiation (both types of models), wind speed and air temperature (aerodynamic model).The EVASPA framework is presently used for the definition of the ET product in the frame of the TRISHNA thermal infrared space mission (CNES/ISRO).

We address the question of sensitivity analysis for model outputs of any dimension using regional sensitivity analysis (RSA). Classical RSA computes sensitivity indices related to the impact of model input variations on the occurrence of a target region of the model output space. In this work, we take this perspective one step further by proposing to find, for a given model input, the region whose occurrence is best explained by the variations of this input. When it exists, this region can be seen as a model behavior whose occurrence is particularly sensitive to the variations of the model input under study. We name this method mRSA (for maximized RSA). mRSA is formalized as an optimization problem using region-based sensitivity indices. Two formulations are studied, one theoretically and one numerically, using a dedicated algorithm. Using a 2D test model and an environmental model producing time series, we show that mRSA, as a new model exploration tool, can provide interpretable insights on the sensitivity of model outputs of various dimensions.

Process-based soil-crop models are widely used in agronomic research. They are major tools for evaluating climate change impact on crop production. Multi-model simulation studies show a wide diversity of results among models, implying that simulation results are very uncertain. A major path to improving simulation results is to propose improved calibration practices that are widely applicable. This study proposes an innovative generic calibration protocol. The two major innovations concern the treatment of multiple output variables and the choice of parameters to estimate, both of which are based on standard statistical procedure adapted to the particularities of soil-crop models. The protocol performed well in a challenging artificial-data test. The protocol is formulated so as to be applicable to a wide range of models and data sets. If widely adopted, it could substantially reduce model error and inter-model variability, and thus increase confidence in soil-crop model simulations.

A major effect of environment on crops is through crop phenology, and therefore, the capacity to predict phenology for new environments is important. Mechanistic crop models are a major tool for such predictions, but calibration of crop phenology models is difficult and there is no consensus on the best approach. We propose an original, detailed approach for calibration of such models, which we refer to as a calibration protocol. The protocol covers all the steps in the calibration workflow, namely choice of default parameter values, choice of objective function, choice of parameters to estimate from the data, calculation of optimal parameter values, and diagnostics. The major innovation is in the choice of which parameters to estimate from the data, which combines expert knowledge and data-based model selection. First, almost additive parameters are identified and estimated. This should make bias (average difference between observed and simulated values) nearly zero. These are “obligatory” parameters, that will definitely be estimated. Then candidate parameters are identified, which are parameters likely to explain the remaining discrepancies between simulated and observed values. A candidate is only added to the list of parameters to estimate if it leads to a reduction in BIC (Bayesian Information Criterion), which is a model selection criterion. A second original aspect of the protocol is the specification of documentation for each stage of the protocol. The protocol was applied by 19 modeling teams to three data sets for wheat phenology. All teams first calibrated their model using their “usual” calibration approach, so it was possible to compare usual and protocol calibration. Evaluation of prediction error was based on data from sites and years not represented in the training data. Compared to usual calibration, calibration following the new protocol reduced the variability between modeling teams by 22% and reduced prediction error by 11%.

Soil available water capacity (SAWC) is a key factor to be considered when assessing soil capability to provide ecosystem services. The current study deepens the use of remotely sensed data for mapping SAWC and its components from crop model inversion. The inversion was conducted using the STICS (Simulateur mulTI-discplinaire pour les Cultures Standard) crop model along with the GLUE (Generalized Likelihood Uncertainty Estimation) algorithm on a panel of 14 sites within a rainfed vineyard catchment located in Southern France. Several constraint variables derived from Landsat 7 ETM + satellite imagery (leaf area index -LAI -and evapotranspiration -ET) or in-situ measurements (surface soil moisture -SSM), were used in the inversion process alone or in combination. Three main outcomes could be reported when comparing retrievals of both SAWC and its components against field estimates. First, retrievals were significantly correlated with ground estimates for some SAWC components and some scenarios of constraint variables, although overall retrieving performances were quite poor. Second, poor retrieving performances for two scenarios of constraint variables were related to few sites for which specific processes were disregarded by the modelling framework, namely allochthonous water supply and waterlogging during wet autumn and summer. Third, we could identify some promising combinations of constraint variables, after the removal of the aforementioned sites with specific processes. These promising combinations were (LAI, ET) and even more (LAI, ET, SSM) for estimating SAWC and root zone thickness, as well as SSM for estimating soil moistures at field capacity and wilting point of the topsoil layer. Provided we can avoid site-specific processes, our approach may further provide spatial sampling of SAWC and related components, to be used as surrogate input data for DSM models.

The objective of this study is to evaluate the performances of a semi-empirical approach based on the Bayesian theory to retrieve Green Area Index (GAI) from multiple decametric satellites. It is designed to overcome some limitations in existing Radiative Transfer Model (RTM) inversion methods, including the high dimensionality of the inverse problem, the convergence problem due to possible equifinality, and the dependence of some RTM variables on the crop-specific architecture. The PROSAIL model is first inverted in a calibration step using the Hamiltonian Monte Carlo (HMC) algorithm over a global dataset of ground GAI measurements (for maize, wheat, and rice) and the corresponding reflectance observations from Landsat-8, Sentinel-2, and Quickbird to derive crop-specific distributions of PROSAIL input variables. These distributions were then used as prior information to predict GAI over an independent set of reflectance observations. Results show that the full Bayesian approach provides close estimates of GAI to ground truth, with respective Root Mean Square Error (RMSE) of 1.01, 1.33, and 0.97 for maize, wheat, and rice (R2=0.67, 0.76 and 0.63, respectively). The performances are better than those approaches generally reported using radiative transfer models that are non-crop-specific, like the SNAP algorithm for Sentinel-2, but are slightly behind the purely empirical models based on machine learning. However, the proposed approach provides an explicit insight of the joint distribution of PROSAIL variables that are valid for any satellite platform. This constitutes a major advantage against purely empirical models, as it enables to fully exploit large observational datasets from multiple sensors and generalize to other platforms.

Crop models are powerful tools to support breeding because of their capability to explore genotype x environmentxmanagement interactions that can help design promising plant types under climate change. However, relationships between plant traits and model parameters are often model specific and not necessarily direct, depending on how models formulate plant morphological and physiological features. This hinders model application in plant breeding. We developed a novel trait-based multi-model ensemble approach to improve the design of rice plant types for future climate projections. We conducted multi-model simulations targeting enhanced productivity, and aggregated results into model-ensemble sets of phenotypic traits as defined by breeders rather than by model parameters. This allowed to overcome the limitations due to ambiguities in trait-parameter mapping from single modelling approaches. Breeders' knowledge and perspective were integrated to provide clear mapping from designed plant types to breeding traits. Nine crop models from the AgMIP-Rice Project and sensitivity analysis techniques were used to explore trait responses under different climate and management scenarios at four sites. The method demonstrated the potential of yield improvement that ranged from 15.8% to 41.5% compared to the current cultivars under mid-century climate projections. These results highlight the primary role of phenological traits to improve crop adaptation to climate change, as well as traits involved with canopy development and structure. The variability of plant types derived with different models supported model ensembles to handle related uncertainty. Nevertheless, the models agreed in capturing the effect of the heterogeneity in climate conditions across sites on key traits, highlighting the need for context-specific breeding programmes to improve crop adaptation to climate change. Although further improvement is needed for crop models to fully support breeding programmes, a trait-based ensemble approach represents a major step towards the integration of crop modelling and breeding to address climate change challenges and develop adaptation options.

Soil–plant–atmosphere models and certain land surface models usually require information about the ability of soils to store and release water. Thus, a critical soil parameter for such reservoir-like models is the available water capacity (AWC), which is usually recognized as the most influential parameter when modeling water transfer. AWC does not have a single definition despite its wide use by scientists in research models, by regional managers as land-management tools and by farmers as decision-aid tools. Methods used to estimate AWC are also diverse, including laboratory measurements of soil samples, field monitoring, use of pedotransfer functions, and inverse modeling of soil-vegetation models. However, the resulting estimates differ and, depending on the method and scale, may have high uncertainty. Here, we review the many definitions of AWC, as well as soil and soil–plant approaches used to estimate it from local to larger spatial scales. We focus especially on the limits and uncertainties of each method. We demonstrate that in soil science, AWC represents a capacity—the size of the water reservoir that plants can use—whereas in agronomy, it represents an ability—the quantity of water that a plant can withdraw from the soil. We claim that the two approaches should be hybridized to improve the definitions and estimates of AWC. We also recommend future directions: (i) adapt pedotransfer functions to provide information about plants, (ii) integrate newly available information from soil mapping in spatial inverse-modeling applications, and (iii) integrate model-inversion results into methods for digital soil mapping.