Comparative Analysis of Multiple Gridded Precipitation Datasets for Assessing Their Reliability across Various Physiographic Zones in Odisha, India
Tushar Ranjan Mohanty
All India Coordinated Research Project on Agrometeorology, Odisha University of Agriculture and Technology, Bhubaneswar -751003, Odisha, India.
Chinmaya Kumar Sahu *
Gramin Krishi Mausam Sewa -IMD, Odisha University of Agriculture and Technology, Bhubaneswar -751003, Odisha, India.
Gourisankar Panigrahi
Gramin Krishi Mausam Sewa -IMD, Odisha University of Agriculture and Technology, Bhubaneswar -751003, Odisha, India.
Deepanjali Dugal
Gramin Krishi Mausam Sewa -IMD, Odisha University of Agriculture and Technology, Bhubaneswar -751003, Odisha, India.
*Author to whom correspondence should be addressed.
Abstract
Aims: To evaluate and intercompare six prominent gridded precipitation products—IMD (gauge-based interpolated), CHIRPS, NASA-POWER, PERSIANN-CDR (satellite-derived), ERA5, and ERA5-Land (reanalysis)—against IMD rain gauge observations across four physiographically diverse districts of Odisha, India, and to assess the effectiveness of linear scaling bias correction for agroclimatic and hydrological applications.
Study Design: A comparative validation study was conducted using daily precipitation data for 2000–2023. Both raw and linear scaling bias-corrected datasets were analysed to evaluate the impact of bias correction on product performance.
Place and Duration of Study: The study covered Keonjhar (northern plateau), Sambalpur (central tableland), Koraput (Eastern Ghats), and Khordha (coastal plain), Odisha, India, over a 24-year period (2000–2023).
Methodology: Product performance was evaluated against IMD rain gauge observations using RMSE, NRMSE, MAE, Pearson correlation coefficient (r), R², index of agreement (d), and Kling–Gupta Efficiency (KGE). Heatmaps and Taylor diagrams were used for comparative visualisation.
Results: Considerable spatial variation was observed among products and regions. NASA-POWER showed the best overall performance, particularly in Khordha (r = 0.87; KGE = 0.83) and Koraput (r = 0.87; KGE = 0.87), with relatively low RMSE values (5.2–9.5). Keonjhar showed poor agreement across all products (r = 0.15–0.32), indicating challenges in representing rainfall over complex terrain. Linear scaling generally improved most datasets without substantially reducing statistical skill.
Conclusion: NASA-POWER, particularly after bias correction, emerged as the most reliable dataset for agroclimatic and hydrological applications in Odisha. The findings highlight the importance of physiography-specific validation, bias correction, and multi-product ensembles to reduce rainfall estimation uncertainty.
Keywords: Precipitation, heatmap, physiographic zones, taylor diagram, bias correction