Search

Data assimilation (DA) is the overarching term for an ensemble of techniques to combine all possible information (models, observations, a priori data and statistics) to obtain the best possible estimate of the state of a system (Zhang & Moore, 2015). Data assimilation has its origins in meteorology and found its way into operational weather forecasting, oceanography and hydrology, but it is also a valuable technique for estimating variables related to crop growth (soil moisture, LAI (leaf area index), biomass, etc.) by combining models and observations of crop variables.

Machine learning models for crop yield forecasting often rely on expert-designed features or predictors. The effectiveness and interpretability of these handcrafted features depends on the expertise of the people designing them. Neural networks have the ability to learn features directly from input data and train the feature learning and prediction steps simultaneously. In this paper, we evaluate the performance and interpretability of neural network models for crop yield forecasting using data from the MARS Crop Yield Forecasting System of the European Commission's Joint Research Centre. The selected neural networks can handle sequential or time series data and include long short-term memory (LSTM) recurrent neural network and 1-dimensional convolutional neural network (1DCNN). Performance was compared with a linear trend model and a Gradient-Boosted Decision Trees (GBDT) model, trained using hand-designed features. Feature importance scores of input variables were computed using feature attribution methods and were analyzed by crop yield modeling and agronomy experts. Results showed that LSTM models perform statistically better than GBDT models for soft wheat in Germany and similar to GBDT models for all other case studies. In addition, LSTM models captured the effect of yield trend, static features (e.g. elevation, soil water holding capacity) and biomass features on crop yield well, but struggled to capture the impact of extreme temperature and moisture conditions. Our work shows the potential of deep learning to automatically learn features and produce reliable crop yield forecasts, and highlights the importance and challenges of involving human stakeholders in assessing model interpretability.

Crop yield forecasting at national level relies on predictors aggregated from smaller spatial units to larger ones according to harvested crop areas. Such crop areas come from land cover maps or reported statistics, both of which can have errors and uncertainties. Sub-national or regional crop yield forecasting minimizes the propagation of these errors to some extent. In addition, regional forecasts provide added value and insights to stakeholders on regional differences within a country, which would otherwise compensate each other at national level. We propose a crop yield forecasting approach for multiple spatial levels based on regional crop yield forecasts from machine learning. Machine learning, with its data-driven approach, can leverage larger data sizes and capture nonlinear relationships between predictors and yield at regional level. We designed a generic machine learning workflow to demonstrate the benefits of regional crop yield forecasting in Europe. To evaluate the quality and usefulness of regional forecasts, we predicted crop yields for 35 case studies, including nine countries that are major producers of six crops (soft wheat, spring barley, sunflower, grain maize, sugar beets and potatoes). Machine learning models at regional level had lower normalized root mean squared errors (NRMSE) and uncertainty than a linear trend model, with Wilcoxon p-values of 3e-7 and 2e-7 for 60 days before harvest and end of season respectively. Similarly, regional machine learning forecasts aggregated to national level had lower NRMSEs than forecasts from an operational system in 18 out of 35 cases 60 days before harvest, with a Wilcoxon p-value of 0.95 indicating similar performance. Our models have room for improvement, especially during extreme years. Nevertheless, regional crop yield forecasts from machine learning and aggregated national forecasts provide a consistent forecasting method across spatial levels and insights from regional differences to support important policy decisions.

Digital twins provide a new paradigm for the integrated use of sensor data, process-based and data-driven modelling, and user interaction, to explore the behaviour of individual objects and processes. Digital twins originate from an engineering context and were developed for machines and mainly physical and chemical processes. In this paper, we further develop an understanding of digital twins for the green life sciences, which also include biological and social processes. We report on three use cases, in precision farming, greenhouse control and personalized dietary advice, focusing on practical benefits and challenges of digital twins compared with other research methods. This research extends earlier more conceptual research on digital twins in this domain. We find benefits in increased accuracy and impact because of the real-time data connection of digital twins to their real-life counterparts. Specification, availability and accuracy of relevant data sources are still major challenges. Specifically, when using digital twins for personalized advice, further research is needed on nontechnical aspects so that users will comply with the advice from the digital twins. We have outlined four directions of future research and expect that further research will include data-driven modelling to simulate the complex character of living objects and processes and at the same time develop approaches to limit the amount of required input data.

Methods for accurately estimating within-field yield are essential to improve site-specific crop management and resource use efficiencies, which would be a major step toward sustainable intensification of agricultural systems. We set out to assess the accuracy of within-field soybean yields predicted by two data assimilation methods and to assess these methods’ assimilation efficiency (AE). Yields were estimated by assimilating remotely sensed leaf area index (LAI) data from Sentinel-2 into a soybean crop growth model on a pixel basis. The LAI data was integrated into the model by Ensemble Kalman Filtering (EnKF) or by recalibrating with the Subplex algorithm (recalibration-based). An open-loop setting which only integrates information on the soil layers was used as a baseline scenario for quantifying the AE. We assessed both data assimilation techniques on eight fields (3067 pixels) in the Corn Belt region (Nebraska, Kansas and Kentucky) in the United States. The data set encompassed substantial variation in crop growth conditions: three growing seasons (2018, 2019 and 2020), rainfed and irrigated fields, and early and late planting dates. Ground truth yield acquired from combine monitors was used to validate the yield estimations. Agreement between predicted and observed yield at pixel level was two times higher for both data assimilation methods compared to the open-loop. The root mean square error (RMSE) was 476 kg.ha−1 (RRMSE of 10 %) in the recalibration-based method and 573 kg.ha−1 (RRMSE of 12 %) in the EnKF-based method. For both data assimilation methods, assimilating the LAI improved predictions for 68 % of the pixels. For a further 12 % of pixels, there was no accuracy improvement. For the remaining 20 %, AE was positive for one of the two assimilation methods. The high proportion of pixels with positive AE indicates the potential for overcoming the limitations in applying crop models at high spatial resolution by integrating a crop growth indicator. Assimilating an in-season indicator of crop growth (LAI) into a soybean model made it possible to adjust the simulation pathway, thereby greatly improving the accuracy of the yield estimations at the pixel level. This study elucidates the practical applications of data assimilation strategies for fine-scale within-field crop yield mapping.

International audience; Calibration, the estimation of model parameters based on fitting the model to experimental data, is among the first steps in many applications of process-based models and has an important impact on simulated values. We propose a novel method of developing guidelines for calibration of process-based models, based on development of recommendations for calibration of the phenology component of crop models. The approach was based on a multi-model study, where all teams were provided with the same data and asked to return simulations for the same conditions. All teams were asked to document in detail their calibration approach, including choices with respect to criteria for best parameters, choice of parameters to estimate and software. Based on an analysis of the advantages and disadvantages of the various choices, we propose calibration recommendations that cover a comprehensive list of decisions and that are based on actual practices.

Funding Information: This work was in part supported by the Collaborative Research Center 1253 CAMPOS (Project 7: Stochastic Modeling Framework), funded by the German Research Foundation ( DFG , Grant Agreement SFB 1253/1 2017 ), the Academy of Finland through projects AICropPro ( 316172 ) and DivCSA ( 316215 ) and Natural Resources Institute Finland (Luke) through a strategic project BoostIA, the BonaRes project ''Soil3'' ( BOMA 03037514 ) of the Federal Ministry of Education and Research ( BMBF ), Germany, the Deutsche Forschungsgemeinschaft ( DFG , German Research Foundation ) under Germany's Excellence Strategy - EXC 2070–390732324 EXC (PhenoRob), the project BiomassWeb of the GlobeE programme (Grant number: FKZ031A258B ) funded by the Federal Ministry of Education and Research ( BMBF , Germany), the INRA ACCAF meta-programme, the German Federal Ministry of Education and Research ( BMBF ) in the framework of the funding measure “Soil as a Sustainable Resource for the Bioeconomy – BonaRes”, project “BonaRes (Module B): BonaRes Centre for Soil Research , subproject B” (grant 031B0511B ), the National Key Research and Development Program of China ( 2017YFD0300205 ), the National Science Foundation for Distinguished Young Scholars ( 31725020 ), the Priority Academic Program Development of Jiangsu Higher Education Institutions ( PAPD ), the 111 Project ( B16026 ), and China Scholarship Council , the Agriculture and Agri-Food Canada's Project 1387 under he Canadian Agricultural Partnership, the DFG Research Unit FOR 1695 ‘Agricultural Landscapes under Global Climate Change – Processes and Feedbacks on a Regional Scale, the U.S. Department of Agriculture National Institute of Food and Agriculture (award no. 2015-68007-23133 ) and USDA / NIFA HATCH grant N. MCL02368 , the National Key Research and Development Program of China ( 2016YFD0300105 ), The Broadacre Agriculture Initiative, a research partnership between University of Southern Queensland and the Queensland Department of Agriculture and Fisheries , the Academy of Finland through project AI-CropPro ( 315896 ), the JPI FACCE MACSUR2 project, funded by the Italian Ministry for Agricultural, Food, and Forestry Policies ( D.M. 24064/7303/15 of 6/Nov/2015), the Ministry of Education, Youth and Sports of Czech Republic through SustES - Adaption strategies for sustainable ecosystem services and food security under adverse environmental conditions (project no. CZ.02.1.01/0.0/0.0/16_019/000797 ). The order in which the donors are listed is arbitrary. Funding Information: This work was in part supported by the Collaborative Research Center 1253 CAMPOS (Project 7: Stochastic Modeling Framework), funded by the German Research Foundation (DFG, Grant Agreement SFB 1253/1 2017), the Academy of Finland through projects AICropPro (316172) and DivCSA (316215) and Natural Resources Institute Finland (Luke) through a strategic project BoostIA, the BonaRes project ''Soil3'' (BOMA 03037514) of the Federal Ministry of Education and Research (BMBF), Germany, the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany's Excellence Strategy - EXC 2070?390732324 EXC (PhenoRob), the project BiomassWeb of the GlobeE programme (Grant number: FKZ031A258B) funded by the Federal Ministry of Education and Research (BMBF, Germany), the INRA ACCAF meta-programme, the German Federal Ministry of Education and Research (BMBF) in the framework of the funding measure ?Soil as a Sustainable Resource for the Bioeconomy ? BonaRes?, project ?BonaRes (Module B): BonaRes Centre for Soil Research, subproject B? (grant 031B0511B), the National Key Research and Development Program of China (2017YFD0300205), the National Science Foundation for Distinguished Young Scholars (31725020), the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD), the 111 Project (B16026), and China Scholarship Council, the Agriculture and Agri-Food Canada's Project 1387 under he Canadian Agricultural Partnership, the DFG Research Unit FOR 1695 ?Agricultural Landscapes under Global Climate Change ? Processes and Feedbacks on a Regional Scale, the U.S. Department of Agriculture National Institute of Food and Agriculture (award no. 2015-68007-23133) and USDA/NIFA HATCH grant N. MCL02368, the National Key Research and Development Program of China (2016YFD0300105), The Broadacre Agriculture Initiative, a research partnership between University of Southern Queensland and the Queensland Department of Agriculture and Fisheries, the Academy of Finland through project AI-CropPro (315896), the JPI FACCE MACSUR2 project, funded by the Italian Ministry for Agricultural, Food, and Forestry Policies (D.M. 24064/7303/15 of 6/Nov/2015), the Ministry of Education, Youth and Sports of Czech Republic through SustES - Adaption strategies for sustainable ecosystem services and food security under adverse environmental conditions (project no. CZ.02.1.01/0.0/0.0/16_019/000797). The order in which the donors are listed is arbitrary. Publisher Copyright: © 2021 Elsevier Ltd Calibration, the estimation of model parameters based on fitting the model to experimental data, is among the first steps in many applications of process-based models and has an important impact on simulated values. We propose a novel method of developing guidelines for calibration of process-based models, based on development of recommendations for calibration of the phenology component of crop models. The approach was based on a multi-model study, where all teams were provided with the same data and asked to return simulations for the same conditions. All teams were asked to document in detail their calibration approach, including choices with respect to criteria for best parameters, choice of parameters to estimate and software. Based on an analysis of the advantages and disadvantages of the various choices, we propose calibration recommendations that cover a comprehensive list of decisions and that are based on actual practices.

The WOFOST cropping systems model has been applied operationally over the last 25 years as part of the MARS crop yield forecasting system. In this paper we provide an updated description of the model and reflect on the lessons learned over the last 25 years. The latter includes issues like system performance, model sensitivity, spatial model setup, parameterization and calibration approaches as well as software implementation and version management. Particularly for spatial model calibrations we provide experience and guidelines on how to execute calibrations and how to evaluate WOFOST model simulation results, particularly under conditions of limited field data availability. As an open source model WOFOST has been a success with at least 10 different implementations of the same concept. An overview is provided for those implementations which are managed by MARS or Wageningen groups. However, the proliferation of WOFOST implementations has also led to questions on the reproducibility of results from different implementations as is demonstrated with an example from MARS. In order to certify that the different WOFOST implementations and versions available can reproduce basic sets of inputs and outputs we make available a large set of test cases as appendix to this publication. Finally, new methodological extensions have been added to WOFOST in simulating the impact of nutrients limitations, extreme events and climate variability. Also, a difference is made in the operational and scientific versions of WOFOST with different licensing models and possible revenue generation. Capitalizing both on academic development as well as model testing in real-world situations will help to enable new applications of the WOFOST model in precision agriculture and smart farming.

Current potential yield simulations of potato crops can be lower than actual yields observed in fields and the potential yield simulations miss important cultivar differences, for instance in terms of cultivar earliness. This causes issues not only for potential yield simulations but also when estimating the limiting effects of water (and nitrogen) and yield gaps in farmers’ fields. To address this issue, we recalibrated potential yield simulations with WOFOST for five commercial Dutch potato cultivars. An experiment conducted during the 2020 season was used for model calibration while the 2019 version of the same experiment was used for evaluation. The focus of the calibration was on phenology, photosynthesis, and biomass allocation. To aid the latter, the WOFOST biomass allocation table functions were replaced with sigmoid functions and biomass reallocation from haulm to tubers was introduced. In total 19 parameters were adjusted and/or introduced. Twelve of these parameters were not cultivar-specific, e.g. the specific leaf area over the plant developmental stage (SLATB), and the maximum leaf CO2 assimilation rate as function of developmental time (AMAXTB). Seven parameters differed between the cultivars, e.g. the thermal time between emergence and tuber initiation (TSUM1), and most of these were correlated to cultivar earliness. For the calibration year, final yields were underestimated by on average 5.4%, while for the evaluation year the final yield was overestimated by 4.4%. In comparison, the final dry matter yields ranged from 16.4 to 18.5 t ha−1 in 2019 and 15.5–17.5 t ha−1 in 2020. The old WOFOST parameters for potato resulted in an underestimation of 23.7% and 28.7% for, respectively, 2019 and 2020. Improving the parameter sets greatly increased accuracy of the simulations and provided insight in which parameters are cultivar specific and related to cultivar-earliness.

Plant phenology, which describes the timing of plant development, is a major aspect of plant response to environment and for crops, a major determinant of yield. Since climate change is projected to alter crop phenology worldwide, there is a large effort to predict phenology as a function of environment. Many studies have focused on comparing model equations for describing how phenology responds to weather but the effect of crop model calibration, also expected to be important, has received much less attention. The objective here was to obtain a rigorous evaluation of prediction capability of wheat crop phenology models, and to analyze the role of calibration. The 27 participants in this multi-model study were provided experimental data for calibration and asked to submit predictions for sites and years not represented in those data. Participants were instructed to use and document their 99usual99 calibration approach. Overall, the models provided quite good predictions of phenology (median of mean absolute error of 6.1 days) and did much better than simply using the average of observed values as predictor. Calibration was found to compensate to some extent for differences between models, specifically for differences in simulated time to emergence and differences in the choice of input variables. Conversely, different calibration approaches led to major differences in prediction error between models with the same structure. Given the large diversity of calibration approaches and the importance of calibration, there is a clear need for guidelines and tools to aid with calibration. Arguably the most important and difficult choice for calibration is the choice of parameters to estimate. Several recommendations for calibration practices are proposed. Model applications, including model studies of climate change impact, should focus more on the data used for calibration and on the calibration methods employed.

Predicting wheat phenology is important for cultivar selection, for effective crop management and provides a baseline for evaluating the effects of global change. Evaluating how well crop phenology can be predicted is therefore of major interest. Twenty-eight wheat modeling groups participated in this evaluation. Model predictions depend not only on model structure but also on the parameter values. This study is thus an evaluation of modeling groups, which choose the structure and fix or estimate the parameters, rather than an evaluation just of model structures. Our target population was wheat fields in the major wheat growing regions of Australia under current climatic conditions and with current local management practices. The environments used for calibration and for evaluation were both sampled from this same target population. The calibration and evaluation environments had neither sites nor years in common, so this is a rigorous evaluation of the ability of modeling groups to predict phenology for new sites and weather conditions. Mean absolute error (MAE) for the evaluation environments, averaged over predictions of three phenological stages and over modeling groups, was 9 days, with a range from 6 to 20 days. Predictions using the multi-modeling group mean and median had prediction errors nearly as small as the best modeling group. For a given modeling group, MAE for the evaluation environments was significantly correlated with MAE for the calibration environments, which suggests that it would be of interest to test ensemble predictors that weight individual modeling groups based on performance for the calibration data. About two thirds of the modeling groups performed better than a simple but relevant benchmark, which predicts phenology by assuming a constant temperature sum for each development stage. The added complexity of crop models beyond just the effect of temperature was thus justified in most cases. Finally, there was substantial variability between modeling groups using the same model structure, which implies that model improvement could be achieved not only by improving model structure, but also by improving parameter values, and in particular by improving calibration techniques.

Many studies have applied machine learning to crop yield prediction with a focus on specific case studies. The data and methods they used may not be transferable to other crops and locations. On the other hand, operational large-scale systems, such as the European Commission's MARS Crop Yield Forecasting System (MCYFS), do not use machine learning. Machine learning is a promising method especially when large amounts of data are being collected and published. We combined agronomic principles of crop modeling with machine learning to build a machine learning baseline for large-scale crop yield forecasting. The baseline is a workflow emphasizing correctness, modularity and reusability. For correctness, we focused on designing explainable predictors or features (in relation to crop growth and development) and applying machine learning without information leakage. We created features using crop simulation outputs and weather, remote sensing and soil data from the MCYFS database. We emphasized a modular and reusable workflow to support different crops and countries with small configuration changes. The workflow can be used to run repeatable experiments (e.g. early season or end of season predictions) using standard input data to obtain reproducible results. The results serve as a starting point for further optimizations. In our case studies, we predicted yield at regional level for five crops (soft wheat, spring barley, sunflower, sugar beet, potatoes) and three countries (the Netherlands (NL), Germany (DE), France (FR)). We compared the performance with a simple method with no prediction skill, which either predicted a linear yield trend or the average of the training set. We also aggregated the predictions to the national level and compared with past MCYFS forecasts. The normalized RMSE (NRMSE) for early season predictions (30 days after planting) were comparable for NL (all crops), DE (all except soft wheat) and FR (soft wheat, spring barley, sunflower). For example, NRMSE was 7.87 for soft wheat (NL) (6.32 for MCYFS) and 8.21 for sugar beet (DE) (8.79 for MCYFS). In contrast, NRMSEs for soft wheat (DE), sugar beet (FR) and potatoes (FR) were twice as much compared to MCYFS. NRMSEs for end of season were still comparable to MCYFS for NL, but worse for DE and FR. The baseline can be improved by adding new data sources, designing more predictive features and evaluating different machine learning algorithms. The baseline will motivate the use of machine learning in large-scale crop yield forecasting.

We present results from the Agricultural Model Intercomparison and Improvement Project (AgMIP) Global Gridded Crop Model Intercomparison (GGCMI) Phase I, which aligned 14 global gridded crop models (GGCMs) and 11 climatic forcing datasets (CFDs) in order to understand how the selection of climate data affects simulated historical crop productivity of maize, wheat, rice and soybean. Results show that CFDs demonstrate mean biases and differences in the probability of extreme events, with larger uncertainty around extreme precipitation and in regions where observational data for climate and crop systems are scarce. Countries where simulations correlate highly with reported FAO national production anomalies tend to have high correlations across most CFDs, whose influence we isolate using multi-GGCM ensembles for each CFD. Correlations compare favorably with the climate signal detected in other studies, although production in many countries is not primarily climate-limited (particularly for rice). Bias-adjusted CFDs most often were among the highest model-observation correlations, although all CFDs produced the highest correlation in at least one top-producing country. Analysis of larger multi-CFD-multi-GGCM ensembles (up to 91 members) shows benefits over the use of smaller subset of models in some regions and farming systems, although bigger is not always better. Our analysis suggests that global assessments should prioritize ensembles based on multiple crop models over multiple CFDs as long as a top-performing CFD is utilized for the focus region.

From 2011 to 2013 the MOCCCASIN project (MOnitoring Crops in Continental Climates through Asimillation of Satellite INformation) was carried out financed by the European Commission 7th Framework Programme. During the project, two field campaigns (2011 and 2012) were carried out at two sites (Odoyev and Plavsk) in the Tula region of Russia. During these two campaigns, observations were made at selected winter-wheat fields consisting of phenological stage, biomass samples, hemispherical photographs, spectral properties of the canopy and the soil as well as ancillary information about the field. Meteorological observations from synoptic and agrometeorological stations were collected from the stations in and surrounding the Tula region. Finally, a large trajectory throughout the whole Tula region was surveyed in order to collect fields with different crop types. Data access via doi 10.7910/DVN/CG0UQB.

This paper studies the changes of groundwater, climate and land use in the Pampas of Argentina. These changes offer opportunities and threats. Lowering groundwater without irrigation causes drought and successive crop and yield damage. Rising groundwater may alleviate drought as capillary rise supports root water uptake and crop growth, thus narrowing the difference between potential and actual yields. However, rising groundwater may also limit soil water storage, cause flooding in metropolitan areas and have a negative impact on crop yields. Changing land use from continuous soy bean into crop rotations or natural vegetation may decrease groundwater recharge and thus decrease groundwater levels. In case of crop rotation however, leaching of nutrients like nitrate may increase. We quantified these impacts using integrated dynamic crop growth and soil hydrology modelling. The models were tested at field scale using a local dataset from Argentina. We applied distributed modelling at regional scale to evaluate the impacts on groundwater recharge and crop yields using long term weather data. The experiments showed that threats arise from continuous monotone land use. Opportunities are created when a proper balance is found between supply and demand of soil water using a larger differentiation of land use. Increasing the areas of land use types with higher evapotranspiration, like permanent grassland and trees, will contribute to a more stable hydrologic system with more water storage capacities in the soil system and lower groundwater levels. Modelling tools clearly support the evaluation of the impact of land use and climate change on groundwater levels and crop yields. Fil: Kroes, Joop. Wageningen University And Research Centre; Países Bajos Fil: van Dam, Jos. Wageningen University And Research Centre; Países Bajos Fil: Supit, Iwan. Wageningen University And Research Centre; Países Bajos Fil: de Abelleyra, Diego. Instituto Nacional de Tecnología Agropecuaria. Centro Regional Buenos Aires; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentina Fil: Verón, Santiago Ramón. Instituto Nacional de Tecnología Agropecuaria. Centro Regional Buenos Aires; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentina Fil: de Wit, Allard. Wageningen University And Research Centre; Países Bajos Fil: Boogaard, Hendrik. Wageningen University And Research Centre; Países Bajos Fil: Angelini, Marcos Esteban. Instituto Nacional de Tecnología Agropecuaria. Centro Regional Buenos Aires; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentina Fil: Damiano, Francisco. Instituto Nacional de Tecnología Agropecuaria. Centro Regional Buenos Aires; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentina Fil: Groenendijk, Piet. Wageningen University And Research Centre; Países Bajos Fil: Wesseling, Jan. Wageningen University And Research Centre; Países Bajos Fil: Veldhuizen, Ab. Wageningen University And Research Centre; Países Bajos

For calculating the effects of hydrological measures on agricultural production in the Netherlands a new comprehensive and climate proof method is being developed: WaterVision Agriculture (in Dutch: Waterwijzer Landbouw). End users have asked for a method that considers current and future climate, that can quantify the differences between years and also the effects of extreme weather events. Furthermore they would like a method that considers current farm management and that can distinguish three different causes of crop yield reduction: drought, saline conditions or too wet conditions causing oxygen shortage in the root zone. WaterVision Agriculture is based on the hydrological simulation model SWAP and the crop growth model WOFOST. SWAP simulates water transport in the unsaturated zone using meteorological data, boundary conditions (like groundwater level or drainage) and soil parameters. WOFOST simulates crop growth as a function of meteorological conditions and crop parameters. Using the combination of these process-based models we have derived a meta-model, i.e. a set of easily applicable simplified relations for assessing crop growth as a function of soil type and groundwater level. These relations are based on multiple model runs for at least 72 soil units and the possible groundwater regimes in the Netherlands. So far, we parameterized the model for the crops silage maize and grassland. For the assessment, the soil characteristics (soil water retention and hydraulic conductivity) are very important input parameters for all soil layers of these 72 soil units. These 72 soil units cover all soils in the Netherlands. This paper describes (i) the setup and examples of application of the process-based model SWAP-WOFOST, (ii) the development of the simplified relations based on this model and (iii) how WaterVision Agriculture can be used by farmers, regional government, water boards and others to assess crop yield reduction as a function of groundwater characteristics or as a function of the salt concentration in the root zone for the various soil types.

International audience; The aggregation of simulated gridded crop yields to national or regional scale requires information on temporal and spatial patterns of crop-specific harvested areas. This analysis estimates the uncertainty of simulated gridded yield time series related to the aggregation with four different harvested area data sets. We compare aggregated yield time series from the Global Gridded Crop Model Intercomparison project for four crop types from 14 models at global, national, and regional scale to determine aggregation-driven differences in mean yields and temporal patterns as measures of uncertainty.The quantity and spatial patterns of harvested areas differ for individual crops among the four data sets applied for the aggregation. Also simulated spatial yield patterns differ among the 14 models. These differences in harvested areas and simulated yield patterns lead to differences in aggregated productivity estimates, both in mean yield and in the temporal dynamics.Among the four investigated crops, wheat yield (17% relative difference) is most affected by the uncertainty introduced by the aggregation at the global scale. The correlation of temporal patterns of global aggregated yield time series can be as low as for soybean (r = 0.28).For the majority of countries, mean relative differences of nationally aggregated yields account for 10% or less. The spatial and temporal difference can be substantial higher for individual countries. Of the top-10 crop producers, aggregated national multi-annual mean relative difference of yields can be up to 67% (maize, South Africa), 43% (wheat, Pakistan), 51% (rice, Japan), and 427% (soybean, Bolivia). Correlations of differently aggregated yield time series can be as low as r = 0.56 (maize, India), r = 0.05 (wheat, Russia), r = 0.13 (rice, Vietnam), and r = −0.01 (soybean, Uruguay). The aggregation to sub-national scale in comparison to country scale shows that spatial uncertainties can cancel out in countries with large harvested areas per crop type. We conclude that the aggregation uncertainty can be substantial for crop productivity and production estimations in the context of food security, impact assessment, and model evaluation exercises.

In the period 1999-2009 ten-day SPOT-VEGETATION products of the Normalized Difference Vegetation Index (NDVI) and Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) at 1 km spatial resolution were used in order to estimate and forecast the wheat yield over Europe. The products were used together with official wheat yield statistics to fine-tune a statistical model for each NUTS2 region, based on the Partial Least Squares Regression (PLSR) method. This method has been chosen to construct the model in the presence of many correlated predictor variables (10-day values of remote sensing indicators) and a limited number of wheat yield observations. The model was run in two different modalities: the "monitoring mode", which allows for an overall yield assessment at the end of the growing season, and the "forecasting mode", which provides early and timely yield estimates when the growing season is on-going. Performances of yield estimation at the regional and national level were evaluated using a cross-validation technique against yield statistics and the estimations were compared with those of a reference crop growth model. Models based on either NDVI or FAPAR normalized indicators achieved similar results with a minimal advantage of the model based on the FAPAR product. Best modelling results were obtained for the countries in Central Europe (Poland, North-Eastern Germany) and also Great Britain. By contrast, poor model performances characterise countries as follows: Sweden, Finland, Ireland, Portugal, Romania and Hungary. Country level yield estimates using the PLSR model in the monitoring mode, and those of a reference crop growth model that do not make use of remote sensing information showed comparable accuracies. The largest estimation errors were observed in Portugal, Spain and Finland for both approaches. This convergence may indicate poor reliability of the official yield statistics in these countries., JRC.H.4-Monitoring Agricultural Resources

It is well known that timely crop growth monitoring and accurate crop yield estimation at a fine scale is of vital importance for agricultural monitoring and crop management. Crop growth models have been widely used for crop growth process description and yield prediction. In particular, the accurate simulation of important state variables, such as leaf area index (LAI) and root zone soil moisture (SM), is of great importance for yield estimation. Data assimilation is a useful tool that combines a crop model and external observations (often derived from remote sensing data) to improve the simulated crop state variables and consequently model outputs like crop total biomass, water use and grain yield. In spite of its effectiveness, applying data assimilation for monitoring crop growth at the regional scale in China remains challenging, due to the lack of high spatiotemporal resolution satellite data that can match the small field sizes which are typical for agriculture in China. With the accessibility of freely available images acquired by Sentinel satellites, it becomes possible to acquire data at high spatiotemporal resolution (10&ndash, 30 m, 5&ndash, 6 days), which offers attractive opportunities to characterize crop growth. In this study, we assimilated remotely sensed LAI and SM into the Word Food Studies (WOFOST) model to estimate winter wheat yield using an ensemble Kalman filter (EnKF) algorithm. The LAI was calculated from Sentinel-2 using a lookup table method, and the SM was calculated from Sentinel-1 and Sentinel-2 based on a change detection approach. Through validation with field data, the inverse error was 10% and 35% for LAI and SM, respectively. The open-loop wheat yield estimation, independent assimilations of LAI and SM, and a joint assimilation of LAI + SM were tested and validated using field measurement observation in the city of Hengshui, China, during the 2016&ndash, 2017 winter wheat growing season. The results indicated that the accuracy of wheat yield simulated by WOFOST was significantly improved after joint assimilation at the field scale. Compared to the open-loop estimation, the yield root mean square error (RMSE) with field observations was decreased by 69 kg/ha for the LAI assimilation, 39 kg/ha for the SM assimilation and 167 kg/ha for the joint LAI + SM assimilation. Yield coefficients of determination (R2) of 0.41, 0.65, 0.50, and 0.76 and mean relative errors (MRE) of 4.87%, 4.32%, 4.45% and 3.17% were obtained for open-loop, LAI assimilation alone, SM assimilation alone and joint LAI + SM assimilation, respectively. The results suggest that LAI was the first-choice variable for crop data assimilation over SM, and when both LAI and SM satellite data are available, the joint data assimilation has a better performance because LAI and SM have interacting effects. Hence, joint assimilation of LAI and SM from Sentinel-1 and Sentinel-2 at a 20 m resolution into the WOFOST provides a robust method to improve crop yield estimations. However, there is still bias between the key soil moisture in the root zone and the Sentinel-1 C band retrieved SM, especially when the vegetation cover is high. By active and passive microwave data fusion, it may be possible to offer a higher accuracy SM for crop yield prediction.

We investigate in this study (i) a redefinition of crop variety zonations at a spatial scale of 10x10 km, and (ii) the influence of recalibrated crop parameters on regional yield forecasting of winter wheat and grain maize in western Europe. The baseline zonation and initial crop parameter set was derived from the operational European crop growth monitoring system (CGMS) which involves the agrometeorological model WOFOST. Air temperature data from 325 weather stations over the 1992-2007 period were used to define new zonations in a 300 x 300 km test site. Two parameters which influenced mostly the phenological development stages (i.e. TSUM1 and TSUM2, the effective air temperature sums from emergence to anthesis, and from anthesis to maturity, respectively) were chosen and calibrated. The CGMS was finally run based on these new recalibrated parameters and simulated crop status indicators were compared with official statistics over the 2000-2007 period. Our results showed that the days of anthesis and maturity were simulated with coefficients of determination (R2) ranging from 0.22 to 0.87 for both crops over the study site. A qualitative assessment of maximum leaf area index and harvest index also revealed a more consistent spatial pattern than the initial zonation in the simulation results. Finally, recalibrated TSUM1 and TSUM2 led to improved relationships between official yield and simulated crop indicators (significant R2 in 17 out of 28 and in 14 out of 59 NUTS3 regions with respect to the best predictor for grain maize and winter wheat, respectively).

This report describes a soil nitrogen module (Soil-N), which is combined with the agro-hydrological model, SWAP, and the crop growth model, WOFOST. The core of the Soil-N module is a description of the nitrogen cycle, which is coupled to the organic matter cycle based upon the RothC-26.3 model. Nitrogen can be supplied to the soil as different types of fertilizer applications and through mineralisation of organic nitrogen. Ammonium and nitrate balances are calculated including uptake by plant roots, de-nitrification and leaching of nitrate. Data exchange is on a daily base. The partitioning of nitrogen within crops and the nitrogen contents of crop residues are calculated by WOFOST and passed to the Soil-N module. SWAP generates the data for establishing the water balance of the soil compartment for which the Soil-N perform the simulations. Nitrogen uptake by the crop is calculated as the minimum of the demand by the crop and the availability of nitrogen in the soil. The crop production rate is reduced when the mineral nitrogen stock is limited. Nitrogen-fixation is based on a simple approach. An improved sub-model for phenological stages of soybean was implemented. Increasing atmospheric CO2 concentrations can be accounted for. The innovated integrated model was tested using data sets from The Netherlands, China and Argentina, for which examples are given. This new model can be used as a tool in studies, in which both water and nitrogen can be limited for crop growth.

Remotely sensed vegetation indices are widely used to detect greening and browning trends; especially the global coverage of time-series normalized difference vegetation index (NDVI) data which are available from 1981. Seasonality and serial auto-correlation in the data have previously been dealt with by integrating the data to annual values; as an alternative to reducing the temporal resolution, we apply harmonic analyses and non-parametric trend tests to the GIMMS NDVI dataset (1981–2006). Using the complete dataset, greening and browning trends were analyzed using a linear model corrected for seasonality by subtracting the seasonal component, and a seasonal non-parametric model. In a third approach, phenological shift and variation in length of growing season were accounted for by analyzing the time-series using vegetation development stages rather than calendar days. Results differed substantially between the models, even though the input data were the same. Prominent regional greening trends identified by several other studies were confirmed but the models were inconsistent in areas with weak trends. The linear model using data corrected for seasonality showed similar trend slopes to those described in previous work using linear models on yearly mean values. The non-parametric models demonstrated the significant influence of variations in phenology; accounting for these variations should yield more robust trend analyses and better understanding of vegetation trends.

Land cover and land use are important information sources for environmental issues. One of the most important changes at the Earth's surface concerns land cover and land use. Knowledge about the location and type of these changes is essential for environmental modeling and management. Remote sensing data in combination with additional spatial data are recognized as an important source of information to detect these land cover and land use changes. In this paper, we introduce the Dutch multi-date land use database (LGN). Today there are six versions of the LGN database (LGN1–LGN6) based on satellite imagery of respectively 1986, 1992/1994, 1995/1997, 1999/2000, 2003/2004 and 2007/2008. With the completion of LGN6 a time-series of land use databases covering 20 years became available. The different databases were produced according to different methodologies. The resulting inconsistencies for monitoring land use changes and possible solutions are the main themes of this paper. Emerging user requirements, increased data availability and technical improvements lead to methodological differences between the LGN5 and LGN6 Dutch land use database. Monitoring of land use changes by integrating independent spatial datasets results in a mixture of real and methodological changes. A methodology is applied to detect real land use changes for a limited number of land use monitoring classes. The detected real changes have a high probability (almost 95%) that they are real changes. Next to these real changes, differences exist between LGN5 and LGN6, i.e. so-called methodological land use changes, being the result of methodological adaptations over time.

By governing water transfer between vegetation and atmosphere, evapotranspiration (ET) can have a strong influence on crop yields. An estimation of ET from remote sensing is proposed by the EUMETSAT ‘Satellite Application Facility’ (SAF) on Land Surface Analysis (LSA). This ET product is obtained operationally every 30 min using a simplified SVAT scheme that uses, as input, a combination of remotely sensed data and atmospheric model outputs. The standard operational mode uses other LSA-SAF products coming from SEVIRI imagery (the albedo, the downwelling surface shortwave flux, and the downwelling surface longwave flux), meteorological data, and the ECOCLIMAP database to identify and characterize the land cover. With the overall objective of adapting this ET product to crop growth monitoring necessities, this study focused first on improving the ET product by integrating crop-specific information from high and medium spatial resolution remote-sensing data. A Landsat (30 m)-based crop type classification is used to identify areas where the target crop, winter wheat, is located and where crop-specific Moderate Resolution Imaging Spectroradiometer (MODIS) (250 m) time series of green area index (GAI) can be extracted. The SVAT model was run for 1 year (2007) over a study area covering Belgium and part of France using this supplementary information. Results were compared to those obtained using the standard operational mode. ET results were also compared with ground truth data measured in an eddy covariance station. Furthermore, transpiration and potential transpiration maps were retrieved and compared with those produced using the Crop Growth Monitoring System (CGMS), which is run operationally by the European Commission's Joint Research Centre to produce in-season forecast of major European crops. The potential of using ET obtained from remote sensing to improve crop growth modelling in such a framework is studied and discussed. Finally, the use of the ET product is also explored by integrating it in a simpler modelling approach based on light-use efficiency. The Carnegie–Ames–Stanford Approach (CASA) agroecosystem model was therefore applied to obtain net primary production, dry matter productivity, and crop yield using only LSA-SAF products. The values of yield were compared with those obtained using CGMS, and the dry matter productivity values with those produced at the Flemish Institute for Technological Research (VITO). Results showed the potential of using this simplified remote-sensing method for crop monitoring

Morocco has an agricultural production system for cereals which is dominated by rainfed, low yielding cereal production which is highly vulnerable to fluctuations in rainfall. Crop yield forecasting systems could play a significant role to reduce the vulnerability of the Moroccan agriculture to weather risks in the framework of a food security strategy. This paper describes the efforts that were carried out for adapting the European Crop Growth Monitoring System to Moroccan conditions for crop monitoring and regional yield forecasting. Results demonstrate that the correlations between CGMS output and regional reported yields can be strongly improved by relatively simple changes to the parameterization of the crop model and a different strategy for the initialization of the soil water balance.

Within this paper we describe and test a methodology to optimize winter-wheat green area index (GAI) simulated by the WOFOST crop model using MODIS estimates of GAI over the Walloon region in Belgium. Detailed crop type maps over the period 2000-2009 were used to derive time-series of crop-specific GAI by selecting only those 250m MODIS pixels that have at least 75% purity for the target crop. Two important model parameters where optimized by minimizing the difference between simulated and observed GAI for each individual pixel and year. The resulting year-specific joint parameter distributions were used to run an ensemble of crop simulations where the ensemble was initialized by sampling from the joint distribution of the corresponding year. The semi-variograms of the retrieved parameters demonstrate that the spatial patterns are consistent with agricultural practices, while seasonal characteristics of weather patterns in Wallonia can, at least partially, explain the temporal variability in the retrieved parameter distributions. Finally, the average ensemble crop simulation results were aggregated to the level of provinces and regions. A validation with EUROSTAT reported yields over the period 2000—2009 demonstrates that the assimilation of MODIS GAI leads to an improved relationship between simulation results and reported yields at regional level., JRC.H.4-Monitoring Agricultural Resources

China Agriculture Remote Sensing Monitoring System (CHARMS) was established in 2001 by the Ministry of Agriculture of China. The main tasks of CHARMS are to monitor agricultural condition with remote sensing and other techniques and to provide basic information for government decision-making and agricultural production management. The European crop growth monitoring system (EU-CGMS) has been studied, introduced, digested and absorbed into CHARMS since last year, and a new China-CGMS was built. The system is made of weather data, crop characteristics data, soil information data, field surveys data, administrative region vector data as well as agro-meteorological modelling and statistical analysis tools etc..The China-CGMS has been providing information of crop growth condition at three different levels of aggregation: counties, provinces, and regions. This paper gives an overview of China-CGMS, including basic principles, methods and regular operation of China-CGMS.

An Observing System Simulation Experiment (OSSE) has been defined to assess the potentialities of assimilating winter wheat leaf area index (LAI) estimations derived from remote sensing into the crop growth model WOFOST. Two assimilation strategies are considered: one based on Ensemble Kalman Filter (EnKF) and the second on recalibration/re-initialisation of uncertain model parameters and initial state conditions. The main objective of the OSS Experiment is to estimate the requisites for the remotely sensed LAI, in terms of accuracy and sampling frequency, to reach target of either 25 or 50% reduction of errors on the final estimation of grain yields. Our results demonstrate that EnKF is not suitable for assimilating LAI in WOFOST as the average error on final grain yields estimation globally increases. These poor results can be explained by the possible differences of phenological development existing between assimilated and modelled LAI values (difference called “phenological shift” in our study) which is not corrected by the EnKF-based assimilation strategy. On the contrary, a recalibration-based assimilation approach globally improves the estimation of final grain yields in a significant way. On average, such improvement can reach up to approximately 65% when observations are available all along the growing season. Improvements on the order of 20% can be already be attained early in the season, which is of great interest in a crop yield forecasting perspective. If the first objective (25%) of error reduction on final grain yields can be reached in a quite high number of assimilated LAI observations availabilities and uncertainty levels, the field of possibilities is significantly restricted for the second objective (50%) and implies to have LAI observations available all along the growing season, at least on a weekly basis and with an uncertainty level equal or ideally lower than 10%. These requirements are not currently met from neither a technological nor an operational point of view but the results presented here can provide guidelines for future missions dedicated to crop growth monitoring.

The main objective of this study was to investigate whether AVHRR data could be useful for crop model simulation steering by intrinsically taking the mixed pixel effects into account. The second objective was to determine if the application of a genetic algorithm could be an effective technique for crop model steering. The principles were tested for the Seville test site using synthetic data and AVHRR data from 1995 and 1996 because these years show a large contrast in crop development. The main conclusions are that a genetic algorithm is a very powerful technique for crop model optimization, but adaptations are needed to the current optimization scheme in order to be able to steer the WOFOST crop model on the basis of NOAA-AVHRR data.