Predicting Wheat Yield Gap and its Determinants in Rainfed Mediterranean Climate of Morocco: Using Ground Information, Satellite Images and Machine Learning

Wheat is the main food crop grown in more than 2.8 million ha in Morocco and almost 16.8 million ha in 21 Middle East and North Africa (MENA) region countries. It is primarily grown in rainfed conditions in the country and in MENA region, with diverse soil and climatic conditions and a varying ra...

Descripción completa

Detalles Bibliográficos
Autores principales: Devkota, Krishna, Devkota Wasti, Mina Kumari, Bouasria, Abdelkrim
Formato: Internal Document
Lenguaje:Inglés
Publicado: International Center for Agricultural Research in the Dry Areas 2022
Materias:
Acceso en línea:https://hdl.handle.net/10568/128099
_version_ 1855526954102423552
author Devkota, Krishna
Devkota Wasti, Mina Kumari
Bouasria, Abdelkrim
author_browse Bouasria, Abdelkrim
Devkota Wasti, Mina Kumari
Devkota, Krishna
author_facet Devkota, Krishna
Devkota Wasti, Mina Kumari
Bouasria, Abdelkrim
author_sort Devkota, Krishna
collection Repository of Agricultural Research Outputs (CGSpace)
description Wheat is the main food crop grown in more than 2.8 million ha in Morocco and almost 16.8 million ha in 21 Middle East and North Africa (MENA) region countries. It is primarily grown in rainfed conditions in the country and in MENA region, with diverse soil and climatic conditions and a varying range of rainfall patterns, mainly characterized by drought due to poor rainfall distribution within the season. Large disparities in attainable yield and profit gaps have been reported, and closing these gaps is important for meeting domestic demand and reducing imports. The main aim of this study was to determine field- and landscape-level yield and yield gaps for wheat and its drivers in the Central region of Morocco using ground information, remote sensing and machine learning approaches. To this end, we prepared a time series map of six vegetation indices (EVI2, CGVI, MSR, NDVI, OSAVI, and RVI) derived from Sentinel-2 images (10 m) over three consecutive crop seasons (2018-2019, 2019-2020, and 2020-2021). Vegetation indices datasets were combined with the climate, soil, and crop management, and the random forest model was calibrated and validated for each cropping season. The models that gave good performance were applied to predict actual yield, potential yield, and the yield gaps at the plot level. The models were used for mapping yield at the regional scale, Rabat-Sale-Kenitra region of Morocco. Based on those datasets, the main drivers of this gap were determined. The findings reveal that RVI, EVI2, and GCVI vegetation indices well predicted wheat yield for the 2018-2019, 2019-2020, and 2020-2021 seasons with R2 of 0.869, 0.863, and 0.844, respectively. The predicted rainfed potential wheat yields were 5.99, 1.53, and 4.66 t per ha, respectively for three crop seasons. Combined over all three seasons, the most important yield determinants are soil moisture, cumulative rainfall during the crop growing period, followed by actual evapotranspiration, and silt content of the soil. When combining soil, climate and management practices in 2019-2020, the major determinants are still soil moisture and the variables of climate followed by the management practices and soil texture. The results and maps produced are of great importance for predicting wheat yield in advance using in-season vegetation indices which is important for the farmers and policymakers for planning at regional and national scales.
format Internal Document
id CGSpace128099
institution CGIAR Consortium
language Inglés
publishDate 2022
publishDateRange 2022
publishDateSort 2022
publisher International Center for Agricultural Research in the Dry Areas
publisherStr International Center for Agricultural Research in the Dry Areas
record_format dspace
spelling CGSpace1280992026-01-23T02:08:19Z Predicting Wheat Yield Gap and its Determinants in Rainfed Mediterranean Climate of Morocco: Using Ground Information, Satellite Images and Machine Learning Devkota, Krishna Devkota Wasti, Mina Kumari Bouasria, Abdelkrim wheat morocco machine learning yield prediction random forest yield gaps vegetation indices Wheat is the main food crop grown in more than 2.8 million ha in Morocco and almost 16.8 million ha in 21 Middle East and North Africa (MENA) region countries. It is primarily grown in rainfed conditions in the country and in MENA region, with diverse soil and climatic conditions and a varying range of rainfall patterns, mainly characterized by drought due to poor rainfall distribution within the season. Large disparities in attainable yield and profit gaps have been reported, and closing these gaps is important for meeting domestic demand and reducing imports. The main aim of this study was to determine field- and landscape-level yield and yield gaps for wheat and its drivers in the Central region of Morocco using ground information, remote sensing and machine learning approaches. To this end, we prepared a time series map of six vegetation indices (EVI2, CGVI, MSR, NDVI, OSAVI, and RVI) derived from Sentinel-2 images (10 m) over three consecutive crop seasons (2018-2019, 2019-2020, and 2020-2021). Vegetation indices datasets were combined with the climate, soil, and crop management, and the random forest model was calibrated and validated for each cropping season. The models that gave good performance were applied to predict actual yield, potential yield, and the yield gaps at the plot level. The models were used for mapping yield at the regional scale, Rabat-Sale-Kenitra region of Morocco. Based on those datasets, the main drivers of this gap were determined. The findings reveal that RVI, EVI2, and GCVI vegetation indices well predicted wheat yield for the 2018-2019, 2019-2020, and 2020-2021 seasons with R2 of 0.869, 0.863, and 0.844, respectively. The predicted rainfed potential wheat yields were 5.99, 1.53, and 4.66 t per ha, respectively for three crop seasons. Combined over all three seasons, the most important yield determinants are soil moisture, cumulative rainfall during the crop growing period, followed by actual evapotranspiration, and silt content of the soil. When combining soil, climate and management practices in 2019-2020, the major determinants are still soil moisture and the variables of climate followed by the management practices and soil texture. The results and maps produced are of great importance for predicting wheat yield in advance using in-season vegetation indices which is important for the farmers and policymakers for planning at regional and national scales. 2022-12-30 2023-01-24T17:29:42Z 2023-01-24T17:29:42Z Internal Document https://hdl.handle.net/10568/128099 en Limited Access application/pdf International Center for Agricultural Research in the Dry Areas Krishna Devkota, Mina Kumari Devkota Wasti, Abdelkrim Bouasria. (30/12/2022). Predicting Wheat Yield Gap and its Determinants in Rainfed Mediterranean Climate of Morocco: Using Ground Information, Satellite Images and Machine Learning. Lebanon: International Center for Agricultural Research in the Dry Areas (ICARDA).
spellingShingle wheat
morocco
machine learning
yield prediction
random forest
yield gaps
vegetation indices
Devkota, Krishna
Devkota Wasti, Mina Kumari
Bouasria, Abdelkrim
Predicting Wheat Yield Gap and its Determinants in Rainfed Mediterranean Climate of Morocco: Using Ground Information, Satellite Images and Machine Learning
title Predicting Wheat Yield Gap and its Determinants in Rainfed Mediterranean Climate of Morocco: Using Ground Information, Satellite Images and Machine Learning
title_full Predicting Wheat Yield Gap and its Determinants in Rainfed Mediterranean Climate of Morocco: Using Ground Information, Satellite Images and Machine Learning
title_fullStr Predicting Wheat Yield Gap and its Determinants in Rainfed Mediterranean Climate of Morocco: Using Ground Information, Satellite Images and Machine Learning
title_full_unstemmed Predicting Wheat Yield Gap and its Determinants in Rainfed Mediterranean Climate of Morocco: Using Ground Information, Satellite Images and Machine Learning
title_short Predicting Wheat Yield Gap and its Determinants in Rainfed Mediterranean Climate of Morocco: Using Ground Information, Satellite Images and Machine Learning
title_sort predicting wheat yield gap and its determinants in rainfed mediterranean climate of morocco using ground information satellite images and machine learning
topic wheat
morocco
machine learning
yield prediction
random forest
yield gaps
vegetation indices
url https://hdl.handle.net/10568/128099
work_keys_str_mv AT devkotakrishna predictingwheatyieldgapanditsdeterminantsinrainfedmediterraneanclimateofmoroccousinggroundinformationsatelliteimagesandmachinelearning
AT devkotawastiminakumari predictingwheatyieldgapanditsdeterminantsinrainfedmediterraneanclimateofmoroccousinggroundinformationsatelliteimagesandmachinelearning
AT bouasriaabdelkrim predictingwheatyieldgapanditsdeterminantsinrainfedmediterraneanclimateofmoroccousinggroundinformationsatelliteimagesandmachinelearning