Fig 1.
Representative crop yield disaggregation methodologies.
Table 1.
Five-year records of wheat and mustard (2018-2023) from the Directorate of Economics and Statistics confirm this stability of crop dominance in the study area.
Fig 2.
Study area and spatial distribution of village level wheat and mustard yields in Hisar and Bhiwani districts, Haryana, India, during the rabi season 2022−23.
The spatial yield maps are author-generated outputs using QGIS. Village boundaries were obtained from Survey of India (SoI) village digital boundary shape files (Product Code: OVSF/-/10; https://onlinemaps.surveyofindia.gov.in/Digital_Product_Show.aspx; accessed Jan 2024).
Table 2.
Overview of datasets used in the study.
Fig 3.
Dataset preparation and integration workflow for crop yield disaggregation, illustrating representative spatial layers for the Hisar and Bhiwani districts used in the study.
The figure shows (from left to right) village administrative boundaries (Survey of India, Product Code: OVSF/-/10; https://onlinemaps.surveyofindia.gov.in/Digital_Product_Show.aspx; accessed Jan 2024) with yield, crop mask (wheat and mustard), Sentinel-2 RGB imagery, representative soil property raster’s (sand, silt, and clay), and selected weather raster’s (air temperature at 2 m, soil temperature level 1, and surface net solar radiation). The workflow diagram layout was created by the authors, and map visualizations were prepared in QGIS.
Table 3.
List of dataset combination and reference combination ID used in the study.
Table 4.
List of algorithms used and reference Algo. ID used in the study.
Fig 4.
Flow diagram for village-to-pixel crop yield disaggregation.
The yield prediction, residual, and corrected yield maps of Hisar and Bhiwani shown in the figure are author-generated (Python) and visualized using QGIS are included to illustrate the methodological process.
Fig 5.
DL based yield disaggregation model for temporal and static inputs.
GRU and LSTM models share identical architectures, differing only in the recurrent cell type.
Table 5.
Hyper parameter settings used for model implementation.
Fig 6.
RMSE, R2 and adjusted R2 across datasets and models for test data.
Table 6.
Comparison of average rank difference among different datasets.
Fig 7.
Comparison of the average rank of different datasets for testing the significance based on RMSE and adjusted R2.
Fig 8.
Comparison of the average rank of different models for testing the significance based on RMSE.
Fig 9.
Comparison training and validation MAE and loss for LSTM and GRU models.
Fig 10.
Comparison among traditional yield disaggregation methods for Hisar and Bhiwani districts: i. wheat map and ii. mustard map.
These maps are analytical raster outputs (Python).
Fig 11.
Comparison among different ML, and DL yield disaggregation methods for Hisar and Bhiwani districts: i. wheat map and ii. mustard map.
These maps are analytical raster outputs (Python) produced in this study and visualized using QGIS.
Table 7.
Cross-validation RMSE (quintals ha−1) for candidate variogram models and selected variogram type for residual kriging across crops, districts, and predictive models.
Fig 12.
Spatial distribution of residuals after kriging for yield prediction in Hisar and Bhiwani districts: (i) wheat residual maps and (ii) mustard residual maps.
These maps are analytical raster outputs (Python) produced in this study and visualized using QGIS.
Table 8.
Comparison of residual error, kriged residual error, and residual spatial autocorrelation across crops, districts, and predictive models.
Fig 13.
Comparison among geostatistically corrected Hisar and Bhiwani crop map of ML- kriging and DL-kriging, i. wheat map ii. mustard map.
These maps are analytical raster outputs (Python) produced in this study and visualized using QGIS.
Table 9.
Village and Block level performance metrics of crop yield disaggregation models.
Table 10.
Comparative performance of block levels predicted aggregated wheat yields across algorithms in metric quintal per hectare.
Table 11.
Comparative performance of block levels predicted aggregated mustard yields across algorithms in metric quintal per hectare.