Evaluation of the use of a Dynamic versus a Simple Correction Model to correct for systematic errors in Swedish Precipitation Data

Abstract: A Dynamic Correction Model (DCM) was implemented to correct daily precipitation data from the network of precipitation gauges hosted by the Swedish Meteorological and Hydrological Institute (SMHI). The goal was to determine whether the DCM could be used to improve estimates of the influence of systematic errors on precipitation observations compared to a previous assessment made on the SMHI dataset using a Simple Correction Model (SCM). While the SCM in question only takes into account station exposure and temperature, reflecting the yearly average gauge level wind speed and the monthly average probability of snow respectively, the DCM would take into account sub-daily and daily observations of precipitation intensity, wind speed and temperature to correct precipitation according to the observed conditions at the time of the precipitation event. DCM corrections were performed on precipitation data from 165 stations throughout Sweden.The long term average result of aggregating daily DCM corrected precipitation was compared to the previous SCM correction for 24 stations while the daily corrected solid precipitation was evaluated against snow depth observations for 59 stations throughout Sweden.Since the systematic errors almost exclusively cause an underestimation of ground true precipitation the correction was expected to produce an increase in total precipitation. Since this undercatch is greater during solid precipitation, a pattern of higher correction was expected to be found in winter months compared to summer months, a pattern which could be expected to be stronger the further north in the country one looked. Furthermore, superimposed on this, areas which are more exposed to wind can be expected to require a larger correction due to the undercatch being also strongly correlated to wind speed. At the sub monthly time scale a variation in the undercatch could be expected when using the DCM with increased corrections required on days when precipitation falls as snow and on days when wind speeds are higher.For the comparison with previous SCM estimates the expectation was that long term totals (30 year average corrected precipitation) would be similar if the two methods are assumed to be equally valid. The DCM was further expected to improve correlations between newly fallen snow and change in snow depth.While the qualitative patterns in the corrections found met the expectations, large differences in the quantitative results of using the two different methods were found; the DCM had a tendency to overcorrect compared to the SCM, especially for the more exposed stations. This calls into question the assumption that both methods are equally valid and differ only in their temporal resolution, and further raises the question which if any of the methods is more valid and why. The evaluation of DCM corrected solid precipitation using snow depth showed a decrease in the correlation between precipitation and change in snow depth which was contrary to expectations. The greatest uncertainty in the DCM was attributed to the methods for assessing gauge level wind speeds.