An alternative method for predicting relative humidity for climate change studies

The reliable prediction of climate variables at finer temporal resolutions, e.g. daily scale, is essential for climate change impact analysis. However, humidity variables remain less focused, although probable changes under changing climate conditions and their effect on impact studies can be significant. The potential of the daily minimum air temperature (T_n) downscaled from general circulation models (GCMs) to predict ground-scale relative humidity in climate change studies is presented. Daily data over 50 years for Sendai city were obtained from the Japan Meteorological Agency. For GCM data, the INMCM4 model developed by the Institute for Numerical Mathematics, Russia, was selected. Statistically downscaled T_n was used with the linear or second order polynomial relationships developed for each calendar month using observed relative humidity and T_n. This produced predicted relative humidity using the alternative method. Furthermore, relative humidity was directly downscaled using the relative humidity results from the GCM. The results from the two methods were compared with the observed relative humidity data based on the root mean square error and the Kolmogorov–Smirnov test. Results based on root mean square error indicate that the predictions made using the downscaled T_n were less satisfied only for June and September compared to the results from the direct downscaling method. Moreover, the predicted relative humidity using the downscaled T_n matched rather well with the observations for 8 months (except for March, July, August and October) with calculated p values above the 0.05 level. Notably, the proposed method achieved better performance in the winter months when the directly downscaled method failed to satisfy the Kolmogorov–Smirnov test.