Statistical error analysis of a nitrate deposition

Sammanfattning: ECDS, Environment Climate Data Sweden, is a commitment that SMHI rendered the Science Council of Sweden. Their purpose is to assist with search, documentation and publication of data regarding environment and climate. ECDS oer services for data stored in their database. One such service that very useful is THREDDS, Thematic Realtilme Enviromental Distributed Data Service, which is a tool to make sections from large data sets and visualize data. The purpose of this Master's Thesis is to combine data from dierent sources in ECDS portal and create added value by co-evaluation. The essay is then used as a showcase for ECDS. SMHI, the Swedish Meteorological and Hydrological Institute, monitors the deposition of many dierent pollutants using direct measurements and model calculations. Data from measurement stations are often very accurate but there are too few stations to provide knowledge about the geographic distribution of pollutants. Therefore an important compliment to measurements are model calculations, although these are less accurate. SMHI uses a deterministic hierarchical model to predict deposition. The most important component of this model is the atmospheric transport of substances model, called MATCH, Multi-scale Atmospheric Transport and Chemistry. To use data from a model it is important to know how well the predictions reproduce reality. The aim of this Master's Thesis is to perform an error analysis of the deposition model in wet-deposition of nitrate, NO3, by comparing model calculations with real measurements. The error is constructed as the ratio between the model predictions and measurement data. To be able to explain the model error a linear regression model, using a subset of the input-parameters in MATCH as covariates, is constructed. At each observation site the same set of covariates is used but the regression parameters are re estimated. Building a second linear model in the spatial variation in the regression coecient makes it possible to predict and analyze the error everywhere in Europe. Combining the two models results in a mixed-eect model. Measurements data, model calculations and explanatory variables were mostly found in the ECDS's database. The error analysis is developed based on 149 dierent measurement stations and the 8595 model grid of MATCH over Europe. The Mixed-eect model is tested at 9 randomly selected station sites and the results are promising. Keywords Nitrate, Mixed Eect Model.