1845–2016 gridded dataset of monthly precipitation over the upper Adda river basin: a comparison with runoff series

Abstract

A new high-resolution gridded dataset of 1845–2016 monthly precipitation series for the upper Adda river basin was computed starting from a network of high-quality and homogenised station records covering Adda basin and neighbouring areas and spanning more than two centuries. The long-term signal was reconstructed by a procedure based on the anomaly method and consisting in the superimposition of two fields which were computed independently: 1961–1990 monthly climatologies and gridded anomalies. Model accuracy was evaluated by means of station series reconstruction in leave-one-out approach and monthly relative mean absolute errors were found to range between 14 % in summer and 24 % in winter. Except for the period before the 1870s when station coverage is rather low, reconstruction errors are quite stable. The 1845–2016 monthly areal precipitation series integrated over Adda basin was finally computed. The robustness of this series was evaluated and it was investigated for long-term trend. While no significant trend emerged for precipitation, the analysis performed on 1845–2016 annual runoff values recorded at Lake Como outlet highlighted a negative trend. Runoff decrease is supposed to be mostly due to an increasing role of evapotranspiration linked to temperature increase, which is only partially compensated by the increase in glacier melting rate. In order to test the applicability of the gridded database for the reconstruction of extreme past events, the episode with the highest precipitation in Adda basin series (November 2002) was considered and the corresponding gridded fields of monthly anomalies and precipitation values were evaluated both with actual station density and with station densities corresponding to 1922 and 1882. Even considering 1882 station density, the main spatial patterns are well depicted proving the suitability of anomaly method to deal also with sparse station networks.