Hourly Agro-Meteorological Grid Square Data: Creation and accuracy evaluation

The Agro-Meteorological Grid Square Data (AMGSD) system provides daily and hourly meteorological data for Japan’s third-level mesh regions, supporting agricultural and environmental research. This study presents the methodology for generating hourly AMGSD by integrating observational data with Grid Point Value (GPV) provided by the Japan Meteorological Agency. Consistency between hourly and daily values is ensured through a unified computational process. Accuracy evaluations with five meteorological sites maintained by National Agriculture and Food Research Organization (NARO) indicate Root Mean Square Error (RMSE) values ranging from 0.6 to 1.3°C (average 0.86°C) for air temperature, 3.7 to 6.8% (average 5.3%) for relative humidity, and 20.0 to 23.1 W m-2 (average 21.1 W m-2) for downward longwave radiation (DLR). Sapporo site exhibited relatively larger temperature errors, likely due to differences in land use between the observation site and the nearest Japan Meteorological Agency (JMA) station. For forecast data, errors increased with longer lead times, with RMSE values for 9-day forecasts ranging from 1.6 to 3.3°C for air temperature, 9.9 to 16.9% for relative humidity, and 17.0 to 37.9 W m-2 for DLR. In winter, colder regions (Sapporo and Morioka) showed slightly higher errors for relative humidity and DLR. Mean Error (ME) did not exhibit significant seasonal variations for air temperature or DLR, but in Morioka, relative humidity forecasts tended to be underestimated throughout the year.