As part of an earlier pilot project (SMED Report No. 52 2011) and based on data from the national forest survey (RIS) and satellite scenes, the statistical method Probabilistic classifier was used for classifying the forest status of forest land, peat land and mountain areas. Riparian forests were defined based on the virtual streamwater network (VIVAN). Based on data from 200 randomly selected head-waters of the rivers Dalälven, Viskan, Ätran, Nissan and Lagan and the forest sta-tus in riparian and more remote forests, models were created to estimate summer type concentrations of N and P in stream water. The explanation power of the Tot-N and Tot-P models was significantly higher than those used in connection with PLC5 and the modelled type concentrations, representing summer conditions, were significantly higher than those used in PLC5.
Within this new project, water samples were collected and analyzed in spring 2011 from the 200 randomly selected headwaters of river Dalälven and in western Sweden. The measured summer and spring concentration levels were compared and, based on the same methodology as in the previous pilot project, models were created to estimate the springtime Tot-N and Tot-P type concentrations in stream water. Additionally, water samples were collected in September and November 2011 (in progress), but funding for the chemical analyses are lacking.
The spring survey showed lower Tot-N and Tot-P concentrations than during the summer inventories, especially in the Dalälven area. At spring, the median values were just over half the summer values. In western Sweden, the differences between seasons were much smaller and based on median values the spring concentrations were 14% and 25% lower for Tot-N and Tot-P, respectively. Compared with the PLC5 type concentrations, the spring values in the Dalälven area were about 35% and 45% lower for Tot-N and Tot-P, respectively, while on the West Coast they were 24-51% higher for Tot-N and 62% higher for Tot-P. Similarly to the summer survey, the spring inventory indicates that the PLC5 type concentrations underes-timated forest leaching of both Tot-N and Tot-P in Western Sweden, while the results from the Dalälven area instead indicates an overestimate. Type concentra-tions produced by models with data only from spring and summer cannot be uncrit-ically used for source apportionment modelling and as a basis for estimating N and P retention from the headwaters to the sea. This requires models based on water chemical data from all seasons.
The spring survey also showed that the explanation power of the models is signifi-cantly higher than for those used in PLC5, suggesting that there are relations between catchment characteristics and N and P concentrations and that Probabilistic classifier is a useful method for estimating these properties. During both summer and spring, the models explained approximately 60% and 31% of the variation for Tot-N and Tot-P, respectively. The models indicate that the nutrient losses are controlled primarily by the amount of biomass in spring, while growth is ruling during summer.
Given the typically high water discharge in spring and autumn and the resulting impact on the nutrient transport calculations (= concentration x discharge), there are good reasons to take into account the spatial concentration variations during fall in the PLC6 source apportionments by improving the N and P type concentrations estimates. The chemical analysis of the project collected autumn samples should therefore be funded. In addition, similar surveys should be conducted in northern Sweden in order to improve the estimates of N and P losses from forest land, peat land and mountain areas also in this part of the country.