Search

Accurate seasonal forecasting of cereal yields is an important decision support tool for countries, such as Morocco, that are not self-sufficient in order to predict, as early as possible, importation needs. This study aims to develop an early forecasting model of cereal yields (soft wheat, barley and durum wheat) at the scale of the agricultural province considering the 15 most productive over 2000–2017 (i.e., 15 × 18 = 270 yields values). To this objective, we built on previous works that showed a tight linkage between cereal yields and various datasets including weather data (rainfall and air temperature), regional climate indices (North Atlantic Oscillation in particular), and drought indices derived from satellite observations in different wavelengths. The combination of the latter three data sets is assessed to predict cereal yields using linear (Multiple Linear Regression, MLR) and non-linear (Support Vector Machine, SVM; Random Forest, RF, and eXtreme Gradient Boost, XGBoost) machine learning algorithms. The calibration of the algorithmic parameters of the different approaches are carried out using a 5-fold cross validation technique and a leave-one-out method is implemented for model validation. The statistical metrics of the models are first analyzed as a function of the input datasets that are used, and as a function of the lead times, from 4 months to 2 months before harvest. The results show that combining data from multiple sources outperformed models based on one dataset only. In addition, the satellite drought indices are a major source of information for cereal prediction when the forecasting is carried out close to harvest (2 months before), while weather data and, to a lesser extent, climate indices, are key variables for earlier predictions. The best models can accurately predict yield in January (4 months before harvest) with an R2 = 0.88 and RMSE around 0.22 t. ha−1. The XGBoost method exhibited the best metrics. Finally, training a specific model separately for each group of provinces, instead of one global model, improved the prediction performance by reducing the RMSE by 10% to 35% depending on the provinces. In conclusion, the results of this study pointed out that combining remote sensing drought indices with climate and weather variables using a machine learning technique is a promising approach for cereal yield forecasting. [ABSTRACT FROM AUTHOR]

In Morocco, cereal production shows high interannual variability due to uncertain rainfall and recurrent drought periods. Considering the socioeconomic importance of cereal for the country, there is a serious need to characterize the impact of drought on cereal yields. In this study, drought is assessed through (1) indices derived from remote sensing data (the vegetation condition index (VCI), temperature condition index (TCI), vegetation health ind ex (VHI), soil moisture condition index (SMCI) and soil water index for different soil layers (SWI)) and (2) key land surface variables (Land Area Index (LAI), soil moisture (SM) at different depths, soil evaporation and plant transpiration) from a Land Data Assimilation System (LDAS) over 2000–2017. A lagged correlation analysis was conducted to assess the relationships between the drought indices and cereal yield at monthly time scales. The VCI and LAI around the heading stage (March-April) are highly linked to yield for all provinces (R = 0.94 for the Khemisset province), while a high link for TCI occurs during the development stage in January-February (R = 0.83 for the Beni Mellal province). Interestingly, indices related to soil moisture in the superficial soil layer are correlated with yield earlier in the season around the emergence stage (December). The results demonstrate the clear added value of using an LDAS compared with using a remote sensing product alone, particularly concerning the soil moisture in the root-zone, considered a key variable for yield production, that is not directly observable from space. The time scale of integration is also discussed. By integrating the indices on the main phenological stages of wheat using a dynamic threshold approach instead of the monthly time scale, the correlation between indices and yield increased by up to 14%. In addition, the contributions of VCI and TCI to VHI were optimized by using yield anomalies as proxies for drought. This study opens perspectives for the development of drought early warning systems in Morocco and over North Africa, as well as for seasonal crop yield forecasting. [ABSTRACT FROM AUTHOR]

In Morocco, cereal production shows a high interannual variability due to uncertain rainfall and recurrent drought periods that may occur at key phenological stages. In view of the importance of this resource to the country's economy, it is important to better characterize the impact of drought strength and duration on yields in order to better anticipate production at the seasonal scale. While meteorological drought is usually derived from precipitation anomalies, the monitoring of agronomical drought may be more complex because of data availability issues. In this study, drought is assessed through, on one hand, classical indices derived from remote sensing data (Vegetation Condition Index -VCI-, Temperature Condition Index -TCI- and Vegetation Health Index -VHI-) and on the other hand, the outputs of a land data assimilation system (LDAS). NDVI and land surface temperature MODIS products are used to compute VCI, TCI and VHI. The LDAS has been developed by the CNRM/Meteo-France and is composed of the ISBA-A-gs land surface model forced by ERA-5 re-analysis constrained by LAI product and a surface soil moisture derived from scatterometer data. LAI, Transpiration and soil moisture at 20 and 40 cm depth are used. The study period is 2001-2017. The first results obtained on the Settat agricultural province which is the most productive province of cereal in Morocco exhibit a strong correlation between cereal production obtained from agricultural statistics and VCI and VHI during February to March (R=0.8-0.9) corresponding to the critical flowering and grain-filling stages. By contrast, the correlation with TCI was found to be significant during the tillering stage (December-January). TCI is a normalized temperature index computed from the observed extreme values of the time series. It thus represents the anomaly due to air temperature condition. A modified temperature index hereafter called Temperature Hydric Conditions Index (THCI) for which extreme temperature values are computed spatially over each agricultural province at the daily time step is proposed. In order to discard temperature change due to elevation conditions, pixel are filtered using the SRTM MNT. A higher correlation (R=0.87) was observed between THCI and cereal production during the grain maturity stage in April. Finally, outputs from the LDAS appeared very promising as strong correlation were obtained with LAI assimilated in March (R=0.92), with transpiration from February until March (R=0.85-0.9) and for soil moisture at 20 and 40 cm deep at the beginning of crop season (December R=0.86) during emergence. Our current work is focused on the development of empirical models for early forecasting cereal production at the provincial and national scales based on linear and non linear (support vector machine regression) approaches. [ABSTRACT FROM AUTHOR]