Process-based Crop Models Research Articles

Compound extreme climate events intensify yield anomalies of winter wheat in France

Abstract Compound extreme climate events (ECEs) are increasingly recognized for their potential to exacerbate food insecurity risks beyond those posed by isolated events. The notion of ‘compound event’ encompasses not only co-occurring ECEs but also multiple ECEs across (different) growth stages (mECEs). The additional effects of these mECEs on crop yield, particularly considering various types of ECEs and regional scales, remain poorly understood. To close this knowledge gap, we consider droughts, pluvials, heatwaves, and coldwaves, and further identify which types of compound events have additional effects on winter wheat yield in France, using statistical methods and datasets encompassing 94 counties over a 68-year period. Our results indicate co-occurring drought heatwaves in summer and spring, along with co-occurring pluvial heatwaves and pluvial coldwaves in winter, have negative additional effects on yield compared with single ECEs. We further identify the types of mECEs that have intensified effects, with the majority showing negative effects on yield. Key interactions leading to intensified yield loss include droughts in winter or spring combined with summer co-occurring drought heatwaves, pluvials across multiple growth stages, pluvials combined with coldwaves, and the transition between droughts and pluvials, with the most severe anomaly attaining −17.2%. Coldwaves are the main ECE related to intensified yield increases, while their frequency is decreasing. Overall, this study stresses the interactions among ECEs on crop yield, and the identified types of mECEs could serve as foundational information for designing control experiments and improving process-based crop models.

Optimization of an N2O Emission Flux Model Based on a Variable-Step Drosophila Algorithm

The application of intelligent process-based crop model parameter optimization algorithms can effectively improve both the model simulation accuracy and applicability. Based on measured values of soil N2O emission flux in wheat fields from 2020 to 2022, and meteorological data from 1971 to 2022, five parameters of the N2O emission flux module in the APSIM model were optimized using the variable step Fruit Fly algorithm (VSS-FOA). The optimized parameters were the soil nitrification potential, the range of concentrated KNH4 of ammonia and nitrogen at semi-maximum utilization efficiency, the proportion of nitrogen loss to N2O during the nitrification process, the denitrification coefficient, and the Power term P for calculating the denitrification water coefficient. Contrasting the optimized parameters using the VSS-FOA algorithm versus the default values supplied with the model substantially improved the goodness-of-fit to field measurements with the overall R2 increasing from 0.41 to 0.74, and a decrease in NRMSE from 17.1% to 11.4%. This work demonstrates that the VSS-FOA algorithm affords a straightforward mechanism for the optimization of parameters in models such as APSIM to enhance the accuracy of model N2O emission flux estimates.

Improving the simulation accuracy of summer maize growth and yield by pixel-based parameterization based on assimilating upscaled MODIS LAI

Accurately simulating maize (Zea mays L.) yield at the regional scale is of paramount importance for making informed policy adjustments and for ensuring food security. Sobol sensitivity analysis was used in this study to screen sensitive parameters of a process-based crop model (Chinese Agricultural Meteorological Model, CAMM). An upscaling approach was utilized to reduce the intra-pixel heterogeneity error of MODIS LAI. Then upscaled MODIS LAI data were assimilated into the CAMM model through a 4DVar assimilation algorithm to optimize pixel-level sensitive parameters, thereby improving the simulation accuracy of the summer maize growth process and yield. Specific leaf area (SLATB_0, SLATB_1) during the period from emergence to tasseling of summer maize exhibited the strongest impact on summer maize yield. Additionally, the average LAI value within the 85–95 % range of ordered LAI values for small pixels within large pixels effectively reduced the intra-pixel heterogeneity error of MODIS LAI, and pixel-based parameterization of SLATB_0 and SLATB_1 at a larger pixel scale (0.0625°) was achieved. Based on yields recorded at agrometeorological stations from 2015 to 2020, assimilated yields in both data assimilation scheme 1 (DA1, optimization of only the sensitive parameter SLATB_0) and scheme 2 (DA2, simultaneous optimization of sensitive parameters SLATB_0 and SLATB_1) exhibited higher accuracy than schemes without data assimilation (with r values of 0.41–0.72 and NRMSE values of 19–30 %). Furthermore, DA2 showed greater simulation accuracy (r: 0.64–0.93, NRMSE: 9–21 %) than DA1 (r: 0.61–0.91, NRMSE: 12–23 %). Upscaling remotely sensed LAI products can significantly reduce the uncertainties of LAI data at a larger pixel scale, and assimilating these LAI data into crop models can effectively increase the simulation accuracy of crop growth and development processes at the regional scale.

Vulnerability of barley yields in Québec.

Climate projections across Québec indicate increased water stress and recurrent vulnerability of cropping systems. In recent years, reports of droughts and water stress have been recorded across the province. Many parts of Québec have experienced droughts in the past few years, which have had uninvestigated impacts on crops. These droughts have been described as some of the most significant in the last 80 years. On the positive side, climate change is likely to trigger shorter winters and therefore longer growing seasons for several crops. However, for crops like maple syrup, the regions suitable for their cultivation will shift northwards. Despite these projections, studies monitoring the susceptibility of barley to environmental changes, climate variability, and adaptive capacity across Québec are still limited. This study aims to provide a provincial-scale portrait of vulnerability of barley in Québec based on historical growing season precipitation, barley yield, and socioeconomic data (literacy and poverty rates) using a composite statistical model. Growing season precipitation data were downloaded from Ouranos. Barley yield data were collected from the Institut de la Statistique du Québec, and the socio-demographic data were collected from the Advisory Council of Poverty and the Institut de la Statistique du Québec. A vulnerability index with sub-indices (sensitivity, exposure, and adaptive capacity) is deployed. It is hypothesised that (1) vulnerability is inversely associated with adaptive capacity, and (2) the peripheral regions of Québec are more vulnerable and less adaptive to climate stressors. Initial results show that when the vulnerability index for barley is more prominent, the associated index of adaptive capacity is relatively lower. A significant gradient between the peripheral and southern regions of Québec is observed, with vulnerability lowest around Montreal/Laval and gradually increasing towards the peripheral regions. A better understanding of vulnerability warrants a change in approach from focusing solely on climate-related variables to integrating socioeconomic proxies. The challenge, however, has been how to introduce these socioeconomic proxies into empirical and process-based crop models.

Algorithm for estimating cultivar-specific parameters in crop models for newer crop cultivars

Model parameters in mechanistic process-based crop models are broadly classified into general parameters (environmental, management, and parameters related to soil processes), species-specific parameters, and cultivar-dependent parameters. While general and species-specific parameters typically remain unchanged, identifying cultivar-dependent parameters is crucial during calibration. With the continuous development of new cultivars, updating cultivar-specific parameters in crop models is necessary. However, gathering data on the behavior of these newer cultivars and calibrating the model to accommodate them is challenging. Currently, there is no standard approach for determining cultivar-specific parameters in crop models due to the difficulty of identifying variations in the growth and development process in newer cultivars, as well as the variations in the model structure, correlations between sub-model components and between different parameters in the crop models. The present study develops a standard methodology for determining cultivar-dependent parameters based on experiments and incorporating them into process-based crop growth models. The developed methodology is demonstrated using the cotton crop and the GOSSYM, a mechanistic process-based cotton crop simulation model. Experiments are conducted on 40 major cotton cultivars grown in the USA to quantify their growth and development under the same environmental and management conditions. Cultivar-dependent parameters are derived and integrated into the GOSSYM model based on the experimental data. By establishing a standardized methodology and demonstrating its application in the context of cotton and the GOSSYM model, this study provides practical guidance for incorporating cultivar-dependent parameters into crop simulation models. Researchers and agronomists can effectively utilize this methodology to integrate new cultivars into their crop models.

On-farm evaluation of a crop forecast-based approach for season-specific nitrogen application in winter wheat

Inadequate nitrogen (N)-fertilisation practices, that fail to consider seasonally variable weather conditions and their impacts on crop yield potential and N-requirements, cause reduced crop N-use efficiency. As a result, both the ecological and economic sustainability of crop production systems are put at risk. The aim of this study was to develop a season-specific crop forecasting approach that allows for a targeted application of N in winter wheat while maintaining farm revenue compared to empirical N-fertilisation practices. The crop forecasts of this study were generated using the process-based crop model SSM in combination with state-of-the-art seasonal ensemble weather forecasts (SEAS5) for the case study region of Eastern Austria. Results from three winter wheat on-farm experiments showed a significant reduction in applied N when implementing a crop forecast-based N-application approach (-43.33 kgN ha-1, -23.42%) compared to empirical N-application approaches, without compromising revenue from high-quality grain sales. The benefit of this reduced N-application approach was quantified through the economic return to applied N (ERAN). While maintaining revenue, the lower amounts of applied N led to significant benefits of + 30.22% (+ 2.20 € kgN-1) in ERAN.

Versatile crop yield estimator

Accurate production estimates, months before the harvest, are crucial for all parts of the food supply chain, from farmers to governments. While methods have been developed to use satellite data to monitor crop development and production, they typically rely on official crop statistics or ground-based data, limiting their application to the regions where they were calibrated. To address this issue, a new method called VeRsatile Crop Yield Estimator (VeRCYe) has been developed to estimate wheat yield at the pixel and field levels using satellite data and process-based crop models. The method uses the Leaf Area Index (LAI) as the linking variable between remotely sensed data and APSIM crop model simulations. In this process, the sowing dates of each field were detected (RMSE = 2.6 days) using PlanetScope imagery, with PlanetScope and Sentinel-2 data fused into a daily 3 m LAI dataset, enabling VeRCYe to overcome the traditional trade-off between satellite data that has either high temporal or high spatial resolution. The method was evaluated using 27 wheat fields across the Australian wheatbelt, covering a wide range of pedo-climatic conditions and farm management practices across three growing seasons. VeRCYe accurately estimated field-scale yield (R2 = 0.88, RMSE = 757 kg/ha) and produced 3 m pixel size yield maps (R2 = 0.32, RMSE = 1213 kg/ha). The method can potentially forecast the final yield (R2 = 0.78–0.88) about 2 months before the harvest. Finally, the harvest dates of each field were detected from space (RMSE = 2.7 days), indicating when and where the estimated yield would be available to be traded in the market. VeRCYe can estimate yield without ground calibration, be applied to other crop types, and used with any remotely sensed LAI information. This model provides insights into yield variability from pixel to regional scales, enriching our understanding of agricultural productivity.

Comparison of predictive modeling approaches to estimate soil erosion under spatially heterogeneous field conditions

The accuracy of soil erosion models in agroecosystems with heterogeneous field conditions is challenging due to uncertainties from soil water fluxes and crop growth. In this study, we coupled two modeling methods (Freebairn and Rose) to represent soil erosion with a process-based crop and runoff models within the SIMPLACE framework. Their accuracy was compared to a statistical model developed using 16 erosion plots (each of 625 cm2) within the same field. Uncertainty analysis showed that runoff and slope angle were the most critical components for predicting sediment yield in both models, followed by soil erodibility in the Freebairn model and entrainment efficiency in the Rose model. However, due to plot size constraints, slope-length effects were not examined. The Freebairn model had a slightly higher accuracy (RMSE = 0.69 t ha−1 d−1) of sediment yield predictions than the Rose model (RMSE = 0.83 t ha−1 d−1). Both models are effective for predicting soil loss with appropriate parameter values.

The importance of model structure and soil data detail on the simulations of crop growth and water use: A case study for sugarcane

Process-based crop models have faced rapid development over the last years, and many modelling platforms are now available and can be used in a wide range of conditions. Whilst the selection of a model should be suited to the purpose of its application, very few studies focused on the impact of choosing different model structures and data details on the simulation outputs. One important aspect is the soil water dynamics, which can be simulated at different levels of details in terms of data and approaches. In this study, we investigated the impact of model structure and data detail on simulations of sugarcane growth and irrigation scheduling. Three different soil water routines (Standalone, Tipping-Bucket, SWAP) were coupled with the SAMUCA model and calibrated with a comprehensive field experiment dataset. We also tested the influence of using simplified homogeneous (SL) and detailed (DL) soil profile information in model performance. The model framework was evaluated against independent field experiments across Brazil and used to simulate long-term sugarcane growth and irrigation scheduling. After calibration, the SWAP-DL showed the highest accuracy in soil moisture predictions, with a 6 % error (RRMSE), but the difference from TippingBucket-DL was small (8 %). While the performance of stalk dry mass, LAI and water-use efficiency simulations were within the range found in literature, comprehensive field experiments showing significant impacts of drought on sugarcane growth are still lacking for a more rigorous evaluation. Both SWAP and tipping-bucket approaches showed higher robustness to soil data detail as compared to the Standalone method, which should be avoided when soil water is critical for sugarcane growth. The use of tipping-bucket method may still be preferred when the research goal is focused on crop growth and soil parameters are limited. SWAP-SAMUCA may provide an extended ability to represent agrohydrological processes in sugarcane plantations and process understanding.

Improving prediction accuracy of CSM-CERES-Wheat model for water and nitrogen response using a modified Penman-Monteith equation in a semi-arid region

Context or problemThe implementation of crop models for water and nitrogen management are important in preserving water resources and reducing groundwater pollution. The Decision Support System for Agrotechnology Transfer (DSSAT) is a collection of field-scale process-based crop models that simulate crop growth and yield in response to water and nitrogen management as well as many other factors. Objective or research questionThe objective of this study was to improve the simulation of irrigation and nitrogen fertilizer effects on growth and yield of winter wheat using locally modified Penman-Monteith (PM) equation. Specifically, we focused on improving the estimation of ETo using a locally modified Penman-Monteith (PM) equation in the CSM-CERES-Wheat model. MethodsField measured data from a two-year experiment (2009–2010 and 2010–2011) in a semi-arid region were used to calibrate and evaluate the CSM-CERES-Wheat model in a region with strong advective condition. In this experiment, the irrigation treatments were 1.2, 1.0, 0.8, 0.5 times by full irrigation requirements and rain-fed, and nitrogen treatments were 0, 46, 92, and 136 kg N ha−1. ResultsUsing the CSM-CERES-Wheat model with the unmodified PM equation resulted in some growth variables such as: aboveground dry matter, grain yield, and grain nitrogen uptake with high normalized root mean square error (NRMSE) of 0.51, 0.35, and 0.32, respectively. However, using the modified PM equation for ETo module under semi-arid condition improved the soil water content estimation and the model outputs, such as the above ground dry matter, grain yield and grain nitrogen uptake with NRMSE of 0.2, 0.15 and 0.16, respectively, that are accurate. This modification provided more accurate estimations, particularly in our experimental site in semi-aid region with strong advective conditions. ConclusionsThe findings of this study indicated that the locally modified PM equation in the CSM-CERES-Wheat model enhanced the accuracy of estimating irrigation and nitrogen fertilizer on the growth and yield of winter wheat. By improving the estimation of ETo value, the model's performance in simulating crop responses to water and nitrogen management was significantly enhanced. Implications or significanceThe utilization of the accurately estimated ETo value of the modified PM equation has significant importance in the practical application of the CSM-CERES-Wheat model. Furthermore, this modification greatly enhances the accuracy of model outputs, particularly at our experimental site located in a semi-arid region with robust advective conditions.

Modeling the effects of tropospheric ozone on the growth and yield of global staple crops with DSSAT v4.8.0

Abstract. Elevated surface ozone (O3) concentrations can negatively impact growth and development of crop production by reducing photosynthesis and accelerating leaf senescence. Under unabated climate change, future global O3 concentrations are expected to increase in many regions, adding additional challenges to global agricultural production. Presently, few global process-based crop models consider the effects of O3 stress on crop growth. Here, we incorporated the effects of O3 stress on photosynthesis and leaf senescence into the Decision Support System for Agrotechnology Transfer (DSSAT) crop models for maize, rice, soybean, and wheat. The advanced models reproduced the reported yield declines from observed O3-dose field experiments and O3 exposure responses reported in the literature (O3 relative yield loss RMSE <10 % across all calibrated models). Simulated crop yields decreased as daily O3 concentrations increased above 25 ppb, with average yield losses of 0.16 % to 0.82 % (maize), 0.05 % to 0.63 % (rice), 0.36 % to 0.96 % (soybean), and 0.26 % to 1.23 % (wheat) per ppb O3 increase, depending on the cultivar O3 sensitivity. Increased water deficit stress and elevated CO2 lessen the negative impact of elevated O3 on crop yield, but potential yield gains from CO2 concentration increases may be counteracted by higher O3 concentrations in the future, a potentially important constraint to global change projections for the latest process-based crop models. The improved DSSAT models with O3 representation simulate the effects of O3 stress on crop growth and yield in interaction with other growth factors and can be run in the parallel DSSAT global gridded modeling framework for future studies on O3 impacts under climate change and air pollution scenarios across agroecosystems globally.

Enhancing in-season yield forecast accuracy for film-mulched wheat: A hybrid approach coupling crop model and UAV remote-sensing data by ensemble learning technique

Accurate in-season yield forecasts for field-scale crops are crucial for both farmers and decision-makers. Common methods for yield prediction are limited by the availability of unknown weather data (process-based crop models) and the failure to consider yield formation processes (statistical models based on unmanned aerial vehicle (UAV) images), respectively. Furthermore, previous studies focused only on crops without mulching, yet mulching is an important agronomic approach to increase grain yield in the arid areas of northwest China. We aim to develop a hybrid approach coupling crop model and UAV data through ensemble learning to achieve in-season yield forecasts for film-mulched wheat. A four-year field experiment was constructed (2018–2020 and 2021–2023). We first calibrated AquaCrop using data from 2018 to 2020, and historical weather data were employed to drive AquaCrop for predicting yields in 2021–2023. Next, statistical models were constructed to predict yields based on spectral and textural indices calculated from UAV images. Finally, a hybrid approach coupling the AquaCrop model and remote-sensing data was developed using ensemble learning technique. Quantifying the relative contribution of features used SHapley Additive exPlanations values. The results indicated that AquaCrop yield forecasts exhibited considerable uncertainties (R2: 0.53–0.63; NRMSE: 16.54%–14.83%). The interpretation of yield for remote-sensing data was influenced by background and saturation effects, reaching its highest accuracy at the heading stage (R2 was 0.80, NRMSE was 11.88%). Ensemble learning demonstrated strong performance compared to machine learning algorithms. The coupling model combined the advantages of crop and statistical models by the ensemble learning algorithm, achieving accurate yield predictions more than 40 days before harvest (heading stage) based on AdaBoost regression (R2 was 0.88, NRMSE was 8.40%). The most important forecasting factors affecting yield prediction were the textural indices, followed by the AquaCrop simulated values. Overall, the coupled model showed good performance in predicting the in-season yield of film-mulched wheat, which provided new insights into farm-scale yield prediction. Further validation of the generalizability of the coupled model in different scenarios is required in the future to improve the applicability of the model in actual production practice.

Water footprints and crop water use of 175 individual crops for 1990–2019 simulated with a global crop model

The water footprint of a crop (WF) is a common metric for assessing agricultural water consumption and productivity. To provide an update and methodological enhancement of existing WF datasets, we apply a global process-based crop model to quantify consumptive WFs of 175 individual crops at a 5 arcminute resolution over the 1990–2019 period. This model simulates the daily crop growth and vertical water balance considering local environmental conditions, crop characteristics, and farm management. We partition WFs into green (water from precipitation) and blue (from irrigation or capillary rise), and differentiate between rainfed and irrigated production systems. The outputs include gridded datasets and national averages for unit water footprints (expressed in m3 t−1 yr−1), water footprints of production (m3 yr−1), and crop water use (mm yr−1). We compare our estimates to other global studies covering different historical periods and methodological approaches. Provided outputs can offer insights into spatial and temporal patterns of agricultural water consumption and serve as inputs for further virtual water trade studies, life cycle and water footprint assessments.

Quest to find compromised spatial and temporal resolutions for integrating remote sensing data with an agro-ecosystem model for grasslands

This paper addressed one of the main challenges in assimilating remote sensing derived variables into process-based crop model simulations, which is the inconsistent spatial and temporal resolution between information obtained from remote sensing and the outputs of process-based agroecosystem model. We proposed an applied method to reduce the number of required simulations by identifying (i) optimal points in time where additional information from remote sensing has the largest positive influence on the model performance and (ii) options to cluster 10 m grid cells to larger cells without compromising their information content. The MONICA (Model for Nitrogen and Carbon) model was applied to simulate above and below ground biomass in two grassland sites located in southern and eastern parts of Germany. The model was calibrated using LAI values obtained from Sentinel-2 and the sensitivity of output variables to two key root parameters (Root Form factor and specific root length) was evaluated. Our results showed that one or two satellite images covering the critical time periods right after cutting events significantly improved the predictions of grass yields produced by a mechanistic agroecosystem models (by up to 30 %). A larger number of images at other grass growth stages would not further improve the predictive power of the model. We also found that the sensitivity to these critical time periods was independent of model parameters. The mixed-resolution scheme (between 10 and 50 m) achieved better results compared with the high-resolution standalone state-updating method, yet it reduced computational costs by more than 50 %. In conclusion, we proposed a methodology to reduce the number of required simulations for data assimilation by aggregating data from fine to coarse resolutions. Our method was promising for applying data assimilation over large areas and benefiting more from satellite information for real-time prediction of agricultural productivity.

A Comparison of Yield Prediction Approaches Using Long-Term Multi-Crop Site-Specific Data

Highlights Long-term site-specific multiple-crop yield data is used in two modeling approaches. Different sources of data, including soil properties, topography, weather, and multispectral data, are tested. Linear modeling and Bayesian hierarchical modeling (BHM) are evaluated by examining predictions of relative yield. BHM with spatio-temporal effects provides the best estimates, handles missing data, and provides uncertainty estimates in time and space. Abstract. Growers in the inland Pacific Northwest face numerous challenges in managing cropping systems. Climate variability, soil degradation, and topography all lead to significant spatial and temporal variability in yield. Often, yield modeling approaches such as deep learning can be “black boxes” or suffer from parameter uncertainty and instability, as with process-based crop models. To explore an alternative, we examined nearly two decades of crop rotation data from the R.J. Cook Agronomy Farm, along with soil properties, topography, weather, and multispectral data. We then tested two modeling approaches to estimate yield: linear modeling (LM) and Bayesian hierarchical modeling (BHM). We found BHM with spatial and temporal random effects performed best in predicting relative yield, both using soil variables as predictors or remotely sensed data. Since the BHM approach handles missing data, offers the possibility of farmer knowledge to be incorporated into prior probabilities, and gives uncertainty, this methodology lends itself well to decision support tools and on-farm study design. Keywords: Bayesian, Long-term, Modeling, Remote sensing, Yield.

Sensitivity analysis of simulated Lycium barbarum L. yield in the WOFOST model under different climate conditions

Calibrating process-based crop models in a timely manner is often challenging due to the extensive range of parameters involved. Sensitivity analysis (SA) is a useful practice for analyzing model uncertainty in these parameters. In this study, the Morris and extended Fourier Amplitude Sensitivity Test (E-FAST) methods were employed in sub-model level experiments. The yield outputs of WOFOST model under different climate conditions were used to evaluate the impact of altering model parameters in the modeling of Lycium barbarum L. The results indicate that parameters associated with CO2 assimilation rate, leaf area expansion and thermal time during specific periods have a significant impact on the simulated yield. The sensitive parameter rankings obtained from E-FAST exhibited good concordance across each planting site, and both SA methods consistently revealed similar rankings. This study provided a strategy to efficiently verify the applicability of new crop model parameters in different regions.

Calibrating the STICS soil-crop model to explore the impact of agroforestry parklands on millet growth

ContextAgroforestry systems provide critical benefits for food security and climate change mitigation. Yet, they are complex and heteregoneous sytems hard to optimize. The use of process-based crop models provides an opportunity to understand better the interactions between soil, crop, tree and climate and explore the impact of agroforestry on crop growth, for contrasting crop management. ObjectiveThe objectives of this study were to i) calibrate the soil-crop STICS model for pearl millet (Pennisetum glaucum) in order to simulate millet potential growth and impact of water and nitrogen limitations on millet growth in open fields and ii) explore the impacts of the parkland tree Faidherbia albida on millet performance for contrasting N fertilizer inputs. MethodsWe gathered a comprehensive database of 28 agronomically contrasting situations, ranging from near-potential growth to drought- and N-stress, either on-station or in a farmer’s home- or bush-fields. Parameters governing relevant plant and soil processes for grain yield were calibrated in a stepwise procedure. The calibrated model was used to explore the impact on millet growth of the widely reported benefits of Faidherbia albida, namely a minimum reduction in radiation thanks to the peculiar reverse phenology of this tree, improvement of soil water content at the beginning of the growing season and of organic nitrogen in the topsoil. ResultsModel simulations with the calibration dataset were reasonably accurate for aboveground biomass and grain yield. Normalized Root Mean Square Errors (nRMSE) for these variables were 29% and 26%, respectively; model efficiency (EF) was 0.58 for both. The nRMSE ranged from 33% to 53% for Soil Water Content (SWC), plant N uptake, grain number, and leaf-area index (LAI). Model accuracy was lower with the evaluation dataset. In the virtual experiment, millet yield decreased with incoming solar radiation, but only at levels of shading (e.g. below 80% of the radiation obtained with full sun) that do not occur under Faidherbia. The decline was greater when millet was fertilized. Increasing the initial soil water content did not affect simulated millet growth. Simulated millet aboveground biomass and grain yield increased with higher organic nitrogen contents of the topsoil, by 80% when millet was not fertilizer, but only by 25% when millet was fertilized. ImplicationsThis study provides the first set of comprehensively calibrated parameters for applying STICS to pearl millet in open cropland. A virtual experiment with historical climate suggests that the benefits of Faidherbia decrease if farmers intensify crop production by adding more mineral N fertilizer. Hence, precise fertilizer management is recommended in Faidherbia parklands. These results illustrate the benefits of process-based crop modelling for better understanding the functioning of agroforestry systems.

Examining the Corn Seedling Emergence-Temperature Relationship for Recent Hybrids: Insights from Experimental Studies.

Corn seedling emergence is a critical factor affecting crop yields. Accurately predicting emergence is crucial for precise crop growth and development simulation in process-based crop models. While various experimental studies have investigated the relationship between corn seedling emergence and temperature, there remains a scarcity of studies focused on newer corn hybrids. In the present study, statistical models (linear and quadratic functional relationships) are developed based on the seedling emergence of ten current corn hybrids, considering soil and air temperatures as influencing factors. The data used for model development are obtained from controlled soil plant atmospheric research chamber experiments focused on corn seedling emergence at five different temperatures. Upon evaluating the developed models, the quadratic model relating the air temperature with time to emergence was found more accurate for all corn hybrids (coefficient of determination (R2): 0.97, root mean square error (RMSE): 0.42 day) followed by the quadratic model based on soil temperature (R2: 0.96, RMSE: 1.42 days), linear model based on air (R2: 0.94, RMSE: 0.53 day) and soil temperature (R2: 0.94, RMSE: 0.70 day). A growing degree day (GDD)-based model was also developed for the newer hybrids. When comparing the developed GDD-based model with the existing GDD models (based on old hybrids), it was observed that the GDD required for emergence was 16% higher than the GDD used in the current models. This showed that the existing GDD-based models need to be revisited when adopted for newer hybrids and adapted to corn crop simulation models. The developed seedling emergence model, integrated into a process-based corn crop simulation model, can benefit farmers and researchers in corn crop management. It can aid in optimizing planting schedules, supporting management decisions, and predicting corn crop growth, development, and it yields more accurately.

Optimizing Nitrogen Fertilization to Enhance Productivity and Profitability of Upland Rice Using CSM-CERES-Rice.

Nitrogen (N) deficiency can limit rice productivity, whereas the over- and underapplication of N results in agronomic and economic losses. Process-based crop models are useful tools and could assist in optimizing N management, enhancing the production efficiency and profitability of upland rice production systems. The study evaluated the ability of CSM-CERES-Rice to determine optimal N fertilization rate for different sowing dates of upland rice. Field experimental data from two growing seasons (2018-2019 and 2019-2020) were used to simulate rice responses to four N fertilization rates (N30, N60, N90 and a control-N0) applied under three different sowing windows (SD1, SD2 and SD3). Cultivar coefficients were calibrated with data from N90 under all sowing windows in both seasons and the remaining treatments were used for model validation. Following model validation, simulations were extended up to N240 to identify the sowing date's specific economic optimum N fertilization rate (EONFR). Results indicated that CSM-CERES-Rice performed well both in calibration and validation, in simulating rice performance under different N fertilization rates. The d-index and nRMSE values for grain yield (0.90 and 16%), aboveground dry matter (0.93 and 13%), harvest index (0.86 and 7%), grain N contents (0.95 and 18%), total crop N uptake (0.97 and 15%) and N use efficiencies (0.94-0.97 and 11-15%) during model validation indicated good agreement between simulated and observed data. Extended simulations indicated that upland rice yield was responsive to N fertilization up to 180 kg N ha-1 (N180), where the yield plateau was observed. Fertilization rates of 140, 170 and 130 kg N ha-1 were identified as the EONFR for SD1, SD2 and SD3, respectively, based on the computed profitability, marginal net returns and N utilization. The model results suggested that N fertilization rate should be adjusted for different sowing windows rather than recommending a uniform N rate across sowing windows. In summary, CSM-CERES-Rice can be used as a decision support tool for determining EONFR for seasonal sowing windows to maximize the productivity and profitability of upland rice production.

Cross-Validation Strategy Impacts the Performance and Interpretation of Machine Learning Models

Abstract Machine learning algorithms are able to capture complex, nonlinear, interacting relationships and are increasingly used to predict agricultural yield variability at regional and national scales. Using explainable artificial intelligence (XAI) methods applied to such algorithms may enable better scientific understanding of drivers of yield variability. However, XAI methods may provide misleading results when applied to spatiotemporal correlated datasets. In this study, machine learning models are trained to predict simulated crop yield from climate indices, and the impact of cross-validation strategy on the interpretation and performance of the resulting models is assessed. Using data from a process-based crop model allows us to then comment on the plausibility of the “explanations” provided by XAI methods. Our results show that the choice of evaluation strategy has an impact on (i) interpretations of the model and (ii) model skill on held-out years and regions, after the evaluation strategy is used for hyperparameter tuning and feature selection. We find that use of a cross-validation strategy based on clustering in feature space achieves the most plausible interpretations as well as the best model performance on held-out years and regions. Our results provide the first steps toward identifying domain-specific “best practices” for the use of XAI tools on spatiotemporal agricultural or climatic data. Significance Statement “Explainable” or “interpretable” machine learning (XAI) methods have been increasingly used in scientific research to study complex relationships between climatic and biogeoscientific variables (such as crop yield). However, these methods can return contradictory, implausible, or ambiguous results. In this study, we train machine learning models to predict maize yield anomalies and vary the model evaluation method used. We find that the evaluation (cross validation) method used has an effect on model interpretation results and on the skill of resulting models in held-out years and regions. These results have implications for the methodological design of studies that aim to use XAI tools to identify drivers of, for example, crop yield variability.