Prior studies evidenced that crop yield prediction was mostly focused on cereals, including wheat, corn and rice. However, combining the use of field and satellite data to build regression models to predict chickpea yield has not been studied. This study aims to estimate rain-fed chickpea yield using field-measured data and remote sensing vegetation indices in Kermanshah County, the most productive chickpea region, in Iran. The ground data were collected from 14 fields near the rain gauge stations in 2022. The in-situ data measured include soil moisture, soil texture, and leaf area index (LAI). In addition, the remotely-sensed based indices, including Normalized Difference Vegetation Index (NDVI), Soil Adjusted Vegetation Index (SAVI), and Temperature Vegetation Dryness Index (TVDI) acquired from Landsat 8 and 9 satellites were employed to integrate into the regression models. The most optimal model was determined by applying the Bayes Information criterion (BIC), Mallows’ Cp, LASSO and Elastic Net Regression. The results indicated that the regression model based on the BIC and Cp statistics, incorporating two predictor variables, LAI and soil sand percentage, is the optimal model from the emergence to flowering stage. This model accurately predicted 87% of the variation in chickpea yield during testing, with RMSE = 46.4 kg.ha− 1 and MAE = 40.4 kg.ha− 1. The results revealed that the Bic and Cp can be employed to select the most accurate model with the smallest error for estimating the rain-fed chickpea yield. In addition, Agricultural import and export officials can utilize this simplistic model for planning purposes.