Assessment of Biases in Reanalyses in Representing Monsoon Dynamics Over the Indian Subcontinent

Reliable vertical profiles of temperature, humidity, and wind over the Indian Summer Monsoon (ISM) region are essential for diagnosing monsoon dynamics and for downstream forecasting and climate‐service applications, yet remain uncertain over data‐sparse and orographically complex regions. In this study, we evaluate three modern reanalyses—ERA5, JRA‐3Q, and MERRA‐2—against collocated RS41 radiosonde profiles from 11 Indian stations during the 2016–2019 monsoon seasons. The analysis covers temperature and wind from the lower troposphere to 100 hPa and relative humidity to 300 hPa. Beyond mean bias and normalized root‐mean‐square error (NRMSE), we decompose total error into systematic bias and non‐systematic pattern error components and examine lower‐ and upper‐quintile observed regimes. All three reanalyses show near‐surface cold biases (−0.2 to −2 K). Relative humidity has the largest total errors (NRMSE = 0.4–1.2), mainly due to pattern errors associated with variability and phase/timing mismatches. The quintile analysis shows that reanalyses weaken observed temperature, humidity, and wind speed extremes by overestimating lower‐quintile values and underestimating upper‐quintile values. Pattern error is the larger component in about 88% of valid variable‐, level‐, station‐, and regime‐dependent comparisons and commonly contributes around 80% of the total squared error, although its contribution varies among individual cases. ERA5 shows the most robust overall performance, but errors remain variable‐, level‐, station‐, and regime‐dependent. All reanalyses underestimate low‐level vertical shear and misrepresent stability. These uncertainties indicate that reanalyses should be used cautiously for humidity variability, wind‐shear diagnostics, and convection‐related applications over the Indian monsoon region.