Crop Phenology Models Research Articles

Predicting rice phenology across China by integrating crop phenology model and machine learning

This study explores the integration of crop phenology models and machine learning approaches for predicting rice phenology across China, to gain a deeper understanding of rice phenology prediction. Multiple approaches were used to predict heading and maturity dates at 337 locations across the main rice growing regions of China from 1981 to 2020, including crop phenology model, machine learning and hybrid model that integrate both approaches. Furthermore, an interpretable machine learning (IML) using SHapley Additive exPlanation (SHAP) was employed to elucidate influence of climatic and varietal factors on uncertainty in crop phenology model predictions. Overall, the hybrid model demonstrated a high accuracy in predicting rice phenology, followed by machine learning and crop phenology models. The best hybrid model, based on a serial structure and the eXtreme Gradient Boosting (XGBoost) algorithm, achieved a root mean square error (RMSE) of 4.65 and 5.72 days and coefficient of determination (R2) values of 0.93 and 0.9 for heading and maturity predictions, respectively. SHAP analysis revealed temperature to be the most influential climate variable affecting phenology predictions, particularly under extreme temperature conditions, while rainfall and solar radiation were found to be less influential. The analysis also highlighted the variable importance of climate across different phenological stages, rice cultivation patterns, and geographic regions, underscoring the notable regionality. The study proposed that a hybrid model using an IML approach would not only improve the accuracy of prediction but also offer a robust framework for leveraging data-driven in crop modeling, providing a valuable tool for refining and advancing the modeling process in rice.

The necessity of coupling the legacy effect with temperature response in crop phenology models

Global climate change has changed vegetation phenology substantially around the world. However, the necessity of coupling legacy effects with temperature responses in phenological models remains unclear. The objective of this study was to demonstrate that legacy effects [based on day of year (DOY) of phenology events] have substantial positive impacts on crop phenology. Data from 1883 crop×site combinations across Germany and China were analyzed. DOY was found to be a temperature-independent factor for both vegetative (VGP) and reproductive growth periods (RGP) (based on variance inflation factor values). Partial correlation analysis suggested that DOY explained almost the same variability in date of phenology events as temperature. Akaike information criterion showed the cost-effectiveness of coupling DOY with temperature in 71.2% and 59.1% of sites in VGP and RGP, respectively. A model that coupled a linear legacy effect and a temperature response mechanism (LETM) improved fitting efficiency by an average of 57%. LETM was observed to outperform the well-calibrated WOFOST and WE models in VGP and RGP for all crops. Averaged over all crops, root mean square errors for WOFOST, WE, and LETM were 4.0, 3.9, and 3.7 d in VGP, respectively, and are 5.3, 4.6, and 4.0 d in RGP, respectively. Our results verified the necessity of coupling the legacy effect with temperature responses in phenology models. Given that the results were consistent for all of the crops investigated, we believe that our conclusions can apply to other field crops. Results of this study expand the knowledge of crop phenology responses to environment, and are helpful for accurately predicting crop growth and development responses under future global climate change.

Dual ensemble approach to predict rice heading date by integrating multiple rice phenology models and machine learning-based genetic parameter regression models

Recent research advances in crop breeding and bioinformatics have revealed several genes that determine the flowering (heading) time in some crops. In this study, we calibrated an integrated rice phenology model that uses the genetic information of the cultivar to estimate the parameters of the models. Differences in the prediction of heading stage were examined using different phenology models and parameter estimation methods. We also proposed a novel approach to construct a dual-integrated ensemble model using both multiple crop phenology and regression models to estimate their parameters based on the haplotype combinations of 11 genes related to the heading date of rice. We used the data of 144 Japanese rice (Oryza sativa L.) cultivars to compare the prediction accuracy of the proposed approach with that of individual integrated phenology models. Subsequently, 4,083 ensemble patterns were analyzed using four rice phenology models (Beta, simplified Beta, ORYZA2000, and SIMIRIW) and three machine learning-based parameter regression methods (support vector, random forest, and ridge regression). The accuracy of the individual integrated model depended on the phenology model and its parameter regression method used. In some cases, the RMSE differed by more than one day. When the individual integrated model with the highest prediction accuracy in the training data was applied to the known and unknown environmental test data sets, the prediction accuracies were about 7.0 d and 8.2 d, respectively in the cross-validation. The prediction accuracy of the dual ensemble model with five ensemble members was approximately 5.9 d and 6.5 d in known and unknown environments. Our results suggest that the dual ensemble approach can enhance crop phenology prediction accuracy and present the importance of ensemble parameterization as well as combining multiple crop models when integrating genomic prediction with crop phenology models.

Proposal and extensive test of a calibration protocol for crop phenology models

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

Rice Crop Phenology Model to Monitor Rice Planting and Harvesting Time using Remote Sensing Approach

Rice is one of the most significant food commodity products in Indonesia. The production of rice in 2019 reached 49.8 million tons. On a global scale, rice is consumed by half of the human population around the world. This study will support the development of sustainable natural resources management, which is an important thing to the realization of the Sustainable Development Goals in zero poverty and zero hunger. Remote sensing is a useful instrument to monitor natural resources. This study used Sentinel-2 imageries to extract rice phenology using vegetation indices (NDVI and NDWI), then acquired the planting and harvesting time using the temporal analysis. The NDVI value is showing a parabolic curve regarding the planting stage of the rice. The value of NDVI is high in the transplanting stage but decreases in the harvesting phase. Besides that, in the seedling and transplanting stage, NDWI has a higher value than NDVI. However, in tillering until the harvesting phase, NDWI has a similar characteristic but lower value than NDVI. Based on the spatial and temporal distribution of rice planting and harvesting date, it is known that climate is not a resistant factor, especially the irrigated rice field. Nevertheless, in the rainfed rice field, the planting time depends on climate conditions.

Clisagri: An R package for agro-climate services

Crop yields are affected by unfavourable/extreme weather and climate events occurring during sensitive growth stages. Understanding the risks associated with theseevents is essential to adapt agro-management decisions and reduce losses.For this purpose, we propose a targeted climate service integrating a dynamic crop phenology model with an approach based on dedicated agro-climate risk indicators. The initial set of these indicators has been developed in aco-design approach with agronomists and durum wheat farmers participating as end-users in the H2020-MedGOLD project. Four groups of indicators characterize drought events, excessive wetness, cold stress and heat stress during sensitive growth stages.The proposed approach has been fully implemented as an R-package named Clisagri.The packageis complemented with a set of optimization functions, which target optimal variety selection in terms of crop cycle duration.

Prediction and parameter uncertainty for winter wheat phenology models depend on model and parameterization method differences

Crop phenological development models are fundamental tools that can be used for scheduling agricultural practices and predicting crop yields. In previous research, different crop phenology models were compared, and different parameterization methods for crop model calibration were evaluated, which revealed that both model structures and parameterization methods were important for crop modeling. However, few studies have considered the combination of both factors and compared these influences simultaneously. Therefore, information regarding the extent of variation in model accuracy and uncertainty depending on model structure and parameterization method is lacking. In this study, we developed three winter wheat phenology models with different structures, i.e., the Agricultural Production System Simulator model for wheat (APSIM-wheat), Wang and Engel model, and sigmoid and exponential function based model, to predict the heading date as a case study. We calibrated these models using three different parameterization methods (augmented Lagrange multiplier method, Nelder-Mead method, and Bayesian optimization with Gaussian process) to investigate their effects on model accuracy and uncertainty. Six-fold cross validation of nine combinations of model calibration (3 models × 3 parameterizations) and their validation revealed that accuracies ranged mostly from 2 to 7 days in the root mean square error (RMSE). The coefficient of variation of RMSE varied widely in among model structures and parameterization methods (∼0.01–0.6). Furthermore, the coefficient of variation of model parameters also varied substantially depending both on model structure and parameterization method. Especially for the model with more parameters, we found that the prediction and parameter stability varied depending on parameterization methods. These findings suggest that both prediction and parameter uncertainty varied with model structure and parameterization method and emphasize the importance of which models and parameterization methods modelers use for robust crop phenology model.

Uncertainty in wheat phenology simulation induced by cultivar parameterization under climate warming

Abstract Rigorous calibration of crop phenology models, providing both best-estimate parameters and estimates of parameter uncertainty, is essential for evaluating how crops will respond to future environmental and management changes. Least squares parameter estimation is a widely used approach to calibration of nonlinear models, and there are many software packages available for implementing this approach. However, these packages are rarely if ever used for complex phenology models because of technical difficulties. The purpose of this research is to overcome these difficulties, in particular the issue of a model which is a discontinuous function of the parameters. The calculations were conducted with the WheatGrow phenology model, but the approach is applicable to other complex phenology models. The approach was used to calibrate WheatGrow phenology for 4 widely used cultivars in the main winter wheat production region of China. The resulting fit to the data was quite good (root mean squared error (RMSE) of 3–4 days for flowering and maturity). The coefficients of variation (CV) of the parameters ranged from 6% to 40%. Furthermore, the model was used to predict the effect of warming on phenology, and the uncertainty in those predictions. The results showed that each degree of warming reduced the time from sowing to flowering by 7–8 days for the spring cultivars and 3–4 days for the winter cultivars. The time form flowering to maturity is hardly affected. In addition, the higher the temperature, the larger the uncertainty in the predictions. Comparison with variability in multi-model ensembles suggests that parameter uncertainty is less than the model uncertainty.

Response of double cropping suitability to climate change in the United States

In adapting US agriculture to the climate of the 21st century, a key unknown is whether cropping frequency may increase, helping to offset projected negative yield impacts in major production regions. Combining daily weather data and crop phenology models, we find that cultivated area in the US suited to dryland winter wheat–soybeans, the most common double crop (DC) system, increased by up to 28% from 1988 to 2012. Changes in the observed distribution of DC area over the same period agree well with this suitability increase, evidence consistent with climate change playing a role in recent DC expansion in phenologically constrained states. We then apply the model to projections of future climate under the RCP45 and RCP85 scenarios and estimate an additional 126–239% increase, respectively, in DC area. Sensitivity tests reveal that in most instances, increases in mean temperature are more important than delays in fall freeze in driving increased DC suitability. The results suggest that climate change will relieve phenological constraints on wheat–soy DC systems over much of the United States, though it should be recognized that impacts on corn and soybean yields in this region are expected to be negative and larger in magnitude than the 0.4–0.75% per decade benefits we estimate here for double cropping.

Impact of Climate Change Effects on Contamination of Cereal Grains with Deoxynivalenol

Climate change is expected to aggravate feed and food safety problems of crops; however, quantitative estimates are scarce. This study aimed to estimate impacts of climate change effects on deoxynivalenol contamination of wheat and maize grown in the Netherlands by 2040. Quantitative modelling was applied, considering both direct effects of changing climate on toxin contamination and indirect effects via shifts in crop phenology. Climate change projections for the IPCC A1B emission scenario were used for the scenario period 2031-2050 relative to the baseline period of 1975-1994. Climatic data from two different global and regional climate model combinations were used. A weather generator was applied for downscaling climate data to local conditions. Crop phenology models and prediction models for DON contamination used, each for winter wheat and grain maize. Results showed that flowering and full maturity of both wheat and maize will advance with future climate. Flowering advanced on average 5 and 11 days for wheat, and 7 and 14 days for maize (two climate model combinations). Full maturity was on average 10 and 17 days earlier for wheat, and 19 and 36 days earlier for maize. On the country level, contamination of wheat with deoxynivalenol decreased slightly, but not significantly. Variability between regions was large, and individual regions showed a significant increase in deoxynivalenol concentrations. For maize, an overall decrease in deoxynivalenol contamination was projected, which was significant for one climate model combination, but not significant for the other one. In general, results disagree with previous reported expectations of increased feed and food safety hazards under climate change. This study illustrated the relevance of using quantitative models to estimate the impacts of climate change effects on food safety, and of considering both direct and indirect effects when assessing climate change impacts on crops and related food safety hazards.

Minimising cold damage during reproductive development among temperate rice genotypes. I. Avoiding low temperature with the use of appropriate sowing time and photoperiod-sensitive varieties

Multiple-sown field trials in 4 consecutive years in the Riverina region of south-eastern Australia provided 24 different combinations of temperature and day length, which enabled the development of crop phenology models. A crop model was developed for 7 cultivars from diverse origins to identify if photoperiod sensitivity is involved in determining phenological development, and if that is advantageous in avoiding low-temperature damage. Cultivars that were mildly photoperiod-sensitive were identified from sowing to flowering and from panicle initiation to flowering. The crop models were run for 47 years of temperature data to quantify the risk of encountering low temperature during the critical young microspore stage for 5 different sowing dates. Cultivars that were mildly photoperiod-sensitive, such as Amaroo, had a reduced likelihood of encountering low temperature for a wider range of sowing dates compared with photoperiod-insensitive cultivars. The benefits of increased photoperiod sensitivity include greater sowing flexibility and reduced water use as growth duration is shortened when sowing is delayed. Determining the optimal sowing date also requires other considerations, e.g. the risk of cold damage at other sensitive stages such as flowering and the response of yield to a delay in flowering under non-limiting conditions. It was concluded that appropriate sowing time and the use of photoperiod-sensitive cultivars can be advantageous in the Riverina region in avoiding low temperature damage during reproductive development.

A Simple Phenological Model of Muskmelon Development

Utilizing information gathered in previous growth chamber and field experiments, we developed a simple temperature-driven crop phenology model of muskmelon (Cucumis melo L.) to help commercial growers time crop phenological events and predict harvest dates. The model quantifies vegetative development in terms of main vine node numbers which allows the model to simulate either a direct-seeded or a transplanted crop. The model operates on an hourly time-step but requires only daily weather data and a few cultivar-specific parameters including plastochron interval and thermal time requirements to reach six predefined developmental stages. The model was tested against an independent data set consisting of three muskmelon cultivars grown at five transplanting dates. Tests of the model indicate an average ability to predict main vine node numbers to within one to two nodes of observed values. Estimated harvest date predictions were more variable than those for main vine node number but an average model accuracy of 1 to 3d was obtained in model tests with a data set used to construct the model. Procedures for calibrating the model for different cultivars, cultural practices or environments are outlined.

A non-linear regression form for vegetation index-crop yield relation incorporating acquisition date normalization

A non-linear form relating vegetation indices (VI) to crop grain yields which normalizes for differences in acquisition date is suggested. It is based on the assumption that deviations in VI near the peak VI follow a quadratic behaviour. This form gave a higher R2 value than a simple VI-yield linear model on a multi-year, multi-location data set of IRS (Indian Remote Sensing Satellite-1A) LISS-I(Linear Imaging Self Scanner-I) derived near-infrared (NIR)/red radiance ratios and wheat grain yields in a study site in Madhya Pradesh (India). As the suggested model includes time of peak as a variable, it allows integration of results from other sources, such as, weather-based crop phenology model or high repetivity spectral data into the VI-yield relation.

Optimizing Multiple Cropping Systems: A Systems Approach

ABSTRACT THE objective of this study was to develop a framework for optimizing multiple cropping systems by selecting cropping sequences and their management practices as affected by weather and cropping history. Several alternative formulations of multiple cropping problems were studied with regard to their practicality for solutions. A deterministric activity network model that combined simulation and optimization techniques was developed to study this problem. Irrigation strategies of various crops were used to demonstrate the ability of the model to determine crop sequences as well as within-season management practices. Application of the model to other types of management decisions (i.e. pesticide application) was also discussed. Simulation models were needed to predict crop yield responses to management practices and to simulate the state of the field after each crop. In this case, the state of the field was the remaining soil water. A crop phenology model was used to predict the duration of each crop, depending on when it was planted and subsequent weather conditions. A crop yield response model estimated the yield of the crop based on weather and irrigation management. A soil water balance model predicted available soil water during each season and also provided the values for available water during the time when a new crop was being selected. These models were combined to simulate the decision network. Then, the K Longest Path algorithm was applied to optimize cropping sequences. The results showed that combined network optimization-simulation was an effective way to study multiple cropping systems. With appropriate simulation models, this approach can be used to study multiple cropping under various management conditions.

Estimating Emergence Date of Spring Small Grains Using Landsat Spectral Data1

AbstractRemotely acquired data using the Landsat satellite system provide a means for initiating crop growth simulation models and crop phenology models at a field level over a large geographic area. A model based on Landsat spectral data has been developed that estimates a spectral emergence of the crop at a pixel (field) level. This spectral emergence date is evaluated using dates of planting and emergence reported by farmers for their fields of spring wheat (Triticum aestivum L.) and barley (Hordeum vulgare L.) located in North and South Dakota. The differences between model predicted emergence date and farmer reported is found to be 4.9 ± 7.7 days. The model can provide a means of understanding the distribution of crop emergence across all pixels of a region. This knowledge of emergence date of spring grains from satellite data provides an ability to run crop development and yield prediction models at field level.

A crop moisture stress index for large areas and its application in the prediction of spring wheat phenology

Crop phenology models are used in remote sensing programs for interpreting changes in spectral signatures during the growing season, in yield modeling, and in early warning of episodic events that affect crop production over large areas. An agrometeorological crop phenology model for spring wheat that incorporates the photothermal concept suggested in the Robertson crop calendar model is presented and modified to incorporate a crop moisture stress index as an additional model parameter. The crop stress index parameter is derived from comparison of crop water use under unlimited water conditions. The index may be used in other applications such as yield predictions. The influence of moisture stress on phenological development was derived from the literature. Development in general is retarded when moisture stress occurs early in the vegetative phase. Moisture deficit conditions after flowering hasten senescence, possibly through an increase in leaf temperature. Validity of the model was tested in the 1980 spring wheat areas in north Dakota, South Dakota, Minnesota and Montana. The model is flexible to incorporate remotely sensed multispectral data as inputs.