Reliable estimation of surface solar radiation is crucial for climatological analyses and land-atmosphere applications; how-ever, direct radiometric observations remain sparse across much of India. Under such data limitations, temperature-based empirical models continue to be widely used, although their seasonal robustness and spatial transferability across diverse climatic regimes remain insufficiently understood. This study assesses the comparative performance of temperature-based models in predicting solar radiation required for agricultural planning across India. Daily data on maximum and minimum temperatures and sunshine duration from 14 agrometeorological stations, covering the period 2002-2023, were used to estimate reference solar radiation using the Angstrom-Prescott model and to intercompare 16 empirical temperature-based formulations against Angstrom-Prescott reference estimates without local calibration. Model performance was assessed at a daily scale using RMSE, MAE, MBE, and R & sup2;, which were synthesized into a Global Performance Indicator (GPI) to obtain season- and location-specific rankings. Results show a clear control of seasonality and geographical setting on model skill: in winter and post-monsoon, the Hargreaves-Samani model (M-1) generally provided the lowest errors across interior and semi-arid stations (e.g., RMSE of 1.94 to 2.42 MJ m & sup2; day and R & sup2; up to 0.66 at stations like Pune and Jabalpur), whereas temperature-range-based formulations that indirectly reflect atmospheric humidity conditions such as M-5 and the Hassan-based model (M-13) performed better at coastal and high-humidity sites (e.g., achieving RMSE values as low as 1.87 MJ m & sup2; day in winter at Goa). During the monsoon, when persistent cloud cover weakens the link between diurnal temperature range and radiation, the Rao et al. model (M-2) consistently achieved the highest GPI values at most locations (with RMSE typically between 3.7 and 4.9 MJ m & sup2; day and near-zero MBE) despite overall reduced R & sup2; (often below 0.34). Across all seasons, a small subset of models (notably M-1, M-2, and M-13) emerged as robust and statistically transferable within the adopted evaluation framework and without local calibration, offering practical, temperature-only options for operational agrometeorological advisories, evapotranspiration estimation, and crop simula-tion in regions lacking direct solar radiation measurements.