ixrbhii/india-imd-rainfall-daily

India daily rainfall (IMD gridded 0.25°), 2015–2025

Daily rainfall for all of India from the India Meteorological Department's gridded dataset: 4,964 grid points (0.25° ≈ 27 km), 1 January 2015 – 31 December 2025 (4,018 days), in mm/day. It is interpolated from about 6,955 rain gauges, so it is measured rainfall, not a model. Ready for LULC Fetch (Library menu, then Forecast ▸ Train forecasting model) or any GIS / Python.

File	What it is
india_rain_daily_2015_2025.parquet	One row per grid point and day: point (lat_lon), lat, lon, date, rain_mm. 19,945,352 rows (79 missing values).
geotiff/india_rain_<year>.tif	The same grids as GeoTIFF (EPSG:4326), one band per day; each band's description is its date (2020-07-15).
india_rain_grid_points.geojson / .csv	The 4,964 grid points with their mean annual rainfall (2015–2025): a points layer.
catalog.json	Titles, types, rows / bands and extents, for the LULC Fetch Library.

Mean annual rainfall over all points ranges from 1,007 mm (2018) to 1,276 mm (2020). A day's value is the rain in the 24 hours ending at 08:30 IST.

Example

import pandas as pd
df = pd.read_parquet("india_rain_daily_2015_2025.parquet")
kerala = df[df.lat.between(8, 12.9) & df.lon.between(74.8, 77.4)]     # 124 points

A first 7-day forecast of all India with LightGBM (direct forecasting, 4 backtests over 2025): MAE 4.38 mm/day, 10 % better than "same as yesterday".

Source, licence and citation

Source: India Meteorological Department (IMD), Pune, "High resolution (0.25 × 0.25 degree) long period (1901–2025) daily gridded rainfall data set over India", downloaded on 2026-10-04 from https://imdpune.gov.in/cmpg/Griddata/Rainfall_25_NetCDF.html (free). IMD asks users to cite:

Pai D.S., Latha Sridhar, Rajeevan M., Sreejith O.P., Satbhai N.S. and Mukhopadhyay B., 2014: Development of a new high spatial resolution (0.25° × 0.25°) long period (1901–2010) daily gridded rainfall data set over India and its comparison with existing data sets over the region. Mausam, 65(1), 1–18.

The data are IMD's; this repository only converts their yearly NetCDF files to GeoTIFF and Parquet (values unchanged). Follow IMD's terms for any use. Converted and published by IXR (agritechixr) with LULC Fetch.