Scholarlink

<p>Reliable estimation of surface solar radiation is crucial for climatological analyses and land–atmosphere applications; however, 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 R², 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&#xa0;MJ m⁻² day⁻¹ and R² 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&#xa0;MJ m⁻² day⁻¹ 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&#xa0;MJ m⁻² day⁻¹ and near-zero MBE) despite overall reduced R² (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 simulation in regions lacking direct solar radiation measurements.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Seasonal and regional evaluation of sixteen temperature-based solar radiation models for India using a global performance indicator

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • Moreshwar Karale,
  • Ashutosh Kumar Misra,
  • Kripan Ghosh,
  • Asha Latwal

摘要

Reliable estimation of surface solar radiation is crucial for climatological analyses and land–atmosphere applications; however, 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 R², 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⁻² day⁻¹ and R² 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⁻² day⁻¹ 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⁻² day⁻¹ and near-zero MBE) despite overall reduced R² (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 simulation in regions lacking direct solar radiation measurements.