Precipitation phase determination in cold region conceptual models - analysis and method development

Sammanfattning: Precipitation phase uncertainty is a known source of error in conceptual models used for many hydrological, climatological, and environmental applications. These conceptual models often use the simple approach of calibrating an air temperature threshold (TRS) over a large area irrespective of physiographic characteristics such as distance to the ocean and topographic relief. Conceptual modeling requires empirical formulas to simplify physical processes. However, there is a plethora of literature against this TRS approach. The magnitude of uncertainty caused by the use of a set TRS is greatest in areas such as Scandinavia, where an average of 39% annual station precipitation occurs in the air temperature (TA) range of -3 to 5°C. One argument for the use of set threshold temperatures in conceptual models was the reduction of computational load, but this came at the cost of accuracy. To compound this error, surface conditions only have a minor contribution to the surface precipitation phase. Instead, microphysics (air-hydrometeor energy exchanges) and properties of the air in the lower atmosphere are the major influences on the observed precipitation phase. However, without adding atmospheric data, improvements to cold region conceptual model, precipitation phase can be achieved through the use of other reported surface data. Meteorological data from 169 observing stations were used to determine percent misclassified precipitation when air temperature (TA), Wet-bulb temperature 0.5°C (TW 0.5°C) thresholds, and an air temperature adjusted by relative humidity (TRH) = 0.75+0.085'(100-RH) formula were applied. The main dataset had roughly 400,000 precipitation events between TA -3 and 5°C. When analysed by country, Norwegian stations had average misclassified precipitation of 10.8% (1.2°C) for TA, 8.3% for TW 0.5°C, and 8.7% for TRH. In comparison, Swedish stations had misclassified precipitation totals of 9.3% (0.9°C) for TA, 8.2% for TW 0.5°C, and 8.7% for TRH. TW 0.5°C resulted in the least misclassified precipitation for both countries. However, saturation vapor pressure, relative humidity (RH), and other parameters required to calculate TW are often not reported by hydrological or meteorological stations. Therefore, improvement in TA methods is preferential over TW or RH methods. Cloud base height TA thresholds were found to increase with height and could be used as a proxy for RH. Cloud base height thresholds had 10.3%, and 9.1% misclassified precipitation in Norway and Sweden respectively. This method had greater error than RH, but performed better in low cloud conditions (100m in Norway and 300m in Sweden), so combining the methods is an option. However, cloud base height is not reported by all stations. If restricted to TA methods, sub-grouping stations by physiographic characteristics within a 15km radius decreased TA misclassified precipitation by 0.5% in Norway with little change in Sweden. This is a result of the Norwegian landscape varying to a greater extent than in Sweden. TA thresholds in Norway ranged from 2.4°C for ocean platforms to 0.9°C in the hills. Particularly high misclassified precipitation rates in mountains and hills can be reduced by nearly 10% when assigning TA for different station sub-groups using 1km maximum elevation or relief. For oceans/coast stations, TA assigned for water temperature sub-groups (reported by 16 stations) reduced misclassified precipitation by 17%. Models applying a daily TA threshold, had precipitation phase uncertainty reduced 10% with RH methods. Changing to an hourly timestep reduced error by more than 29%. Therefore, decreasing temporal resolution to 1-hour was more beneficial than adding parameters to the 24-hour model.