Rubber is a perennial plant grown to produce natural rubber. It is a raw material for industrial and non-industrial products important to the world economy. The sustainability of natural rubber production is, therefore, critical for smallholder livelihoods and economic development. To maintain price stability, it is important to estimate the yields in advance. Remote sensing technology can effectively provide large-scale spatial data; however, productivity estimates need to be processed from high spatial resolution data generated from satellites with high accuracy and reliability, especially for smallholder livelihood areas where smaller plots contrast with large farms. This study used reflectance data from Sentinel-2 satellite imagery acquired for the 12 months between December 2020 and November 2021. The imagery included 213 plots where data on rubber production in smallholder agriculture were collected. Six vegetation indices (Vis), namely Green Soil Adjusted Vegetation Index (GSAVI), Modified Simple Ratio (MSR), Normalized Burn Ratio (NBR), Normalized Difference Vegetation Index (NDVI), Normalized Green (NR), and Ratio Vegetation Index (RVI) were used to estimate the rubber yield. The study found that the red edge spectral band (band 5) provided the best prediction with R² = 0.79 and RMSE = 29.63 kg/ha, outperforming all other spectral bands and VIs. The MSR index provided the highest coefficient of determination, with R² = 0.62 and RMSE = 39.25 kg/ha. When the red edge reflectance was combined with the best VI, MSR, the prediction model only slightly improved, with a coefficient determination of (R²) of 0.80 and an RMSE of 29.42 kg/ha. The results demonstrated that the Sentinel-2 data are suitable for rubber yield prediction for smallholder farmers. The findings of this study can be used as a guideline to apply in other countries or areas. Future studies will require the use of reflectance and vegetation indices derived from satellite data in combination with meteorological data, as well as the application of complex models, such as machine learning and deep learning.

Predictions of crop production in the Chi basin are of major importance for decision support tools in countries such as Thailand, which aims to increase domestic income and global food security by implementing the appropriate policies. This research aims to establish a predictive model for predicting crop production for an internal crop growth season prior to harvest at the province scale for fourteen provinces in Thailand’s Chi basin between 2011 and 2019. We provide approaches for reducing redundant variables and multicollinearity in remotely sensed (RS) and meteorological data to avoid overfitting models using correlation analysis (CA) and the variance inflation factor (VIF). The temperature condition index (TCI), the normalized difference vegetation index (NDVI), land surface temperature (LST_nighttime), and mean temperature (Tmean) were the resulting variables in the prediction model with a p-value < 0.05 and a VIF < 5. The baseline data (2011–2017: June to November) were used to train four regression models, which revealed that eXtreme Gradient Boosting (XGBoost), random forest (RF), and XGBoost achieved R2 values of 0.95, 0.94, and 0.93, respectively. In addition, the testing dataset (2018–2019) displayed a minimum root-mean-square error (RMSE) of 0.18 ton/ha for the optimal solution by integrating variables and applying the XGBoost model. Accordingly, it is estimated that between 2020 and 2022, the total crop production in the Chi basin region will be 7.88, 7.64, and 7.72 million tons, respectively. The results demonstrated that the proposed model is proficient at greatly improving crop yield prediction accuracy when compared to a conventional regression method and that it may be deployed in different regions to assist farmers and policymakers in making more informed decisions about agricultural practices and resource allocation.

Predictions of crop production in the Chi basin are of major importance for decision support tools in countries such as Thailand, which aim to increase domestic income and global food security by implementing the appropriate policies. This research aims to establish a predictive model for predicting crop production for an internal crop growth season prior of harvest at the province scale for fourteen provinces in Thailand's Chi basin between 2011 and 2019. We provide approaches for reducing redundant variables and multicollinearity in remotely sensed (RS) and meteorological data to avoid overfitting models using correlation analysis (CA) and variance inflation factor (VIF). Temperature condition index (TCI), normalized difference vegetation index (NDVI), land surface temperature (LSTnight), and mean temperature (Tmean) were the resulting variables in the prediction model with a p-value < 0.05 and a VIF < 5. The baseline data (2011–2017: June to November) were used to train four regression models, which revealed that eXtreme Gradient Boosting (XGBoost), random forest (RF), and XGBoost achieved R2 values of 0.95, 0.94, and 0.93, respectively. In addition, the testing dataset (2018–2019) displayed a minimum root mean square error (RMSE) of 0.18 ton/ha for the optimal solution by integrating variables and applying the XGBoost model. Accordingly, it is estimated that between 2020 and 2022, the total crop production in the Chi basin region would be 7.88, 7.64, and 7.72 million tons, respectively. The results demonstrated that the proposed model is proficient at greatly improving crop yield prediction accuracy when compared to a conventional regression method and that it may be deployed in different regions to assist farmers and policymakers in making more informed decisions about agricultural practices and resource allocation.

The Northeast region of Thailand covers approximately 16.89 million hectares, with about 6.17 million hectares of seasonal rice cultivation and 2.85 million hectares affected by soil salinity—a major constraint to agricultural productivity in this region. This study develops an integrated data fusion framework combining multi-temporal Landsat-8 and Sentinel-2 imagery to train machine learning (ML) models for the prediction of rice yield and soil salinity, allowing for an analysis of their relationship. The field data comprised 380 rice yield and 625 soil electrical conductivity (EC) samples collected in 2023. Three ML models—Random Forest (RF), Classification and Regression Trees (CART), and Support Vector Regression (SVR)—were applied for variable reduction and optimal predictor selection. RF achieved the highest accuracy for yield prediction (R² = 0.86, RMSE = 0.19 t ha⁻¹) and salinity estimation (R² = 0.93, RMSE = 0.87 dS/m) when using fused Landsat–Sentinel data. Spatial analysis of 5000 matched points showed a strong negative relationship between seedling stage EC and yield (R² = 0.71), with yields declining sharply above 5 dS/m and remaining below 1.5 t ha⁻¹ beyond 15 dS/m. These results demonstrate the potential of multi-sensor fusion and ensemble ML approaches for precise soil salinity monitoring and sustainable rice production.

Accurate mapping of commercial durian plantations is essential for agricultural inventory, precision agriculture, and sustainable land management but remains challenging because of spectral similarity with other evergreen vegetation. This study developed a multisource remote sensing approach integrating multi-temporal Sentinel-2 imagery, Sentinel-1 Synthetic Aperture Radar (SAR), DEM-derived terrain variables, recursive feature selection, and machine learning for binary durian plantation classification in Sisaket Province, Thailand. A total of 2430 field reference samples were used to evaluate Random Forest (RF), Support Vector Machine (SVM), and Classification and Regression Tree (CART). RF achieved the strongest overall performance using the integrated Sentinel-2, Sentinel-1, and DEM dataset, with an Overall Accuracy of 90.38%, an F1-score of 90.59%, a Kappa coefficient of 0.81, and an AUC of 0.991. Recursive feature selection reduced the predictor set from 53 to 12 variables (77.4%) while retaining high classification performance. The retained 12-predictor subset comprised nine Sentinel-2-derived variables (five seasonal EVI and four NDBI variables), two Sentinel-1 VV/VH ratio variables, and one DEM-derived elevation variable, corresponding to 75.0%, 16.7%, and 8.3% of the retained predictors, respectively. The proposed approach provides an accurate and computationally efficient method for durian plantation mapping within the study area, with broader applicability requiring further validation.