systematically underestimated. The bias is not as pronounced for the non glacial rivers.
Com-
bining synoptic-scale and basin-scale predictors (method 3) leads to a substantial improvement
compared to the use of MSLP fields alone (method 1). Analogue forecasts become similar or
better than persistence, depending on catchment and lead time. Usually, persistence performs
better for T=1 day and then method 3
/media/vedurstofan/utgafa/skyrslur/2013/VI_2013_008.pdf