The aim in this study was to develop a model for the probability of finding active golden eagle nests during their breeding season. It was done by using environmental variables derived from expert models which were tested against empirical data. This resulted in a habitat suitability index (HSI), which in this case is assumed to indicate the probability of active nests of golden eagles. The study was conducted together with the County Administrative Board of Västernorrland with the purpose to improve golden eagle’s ecological status.To develop the model, different combinations of several explanatory variables were tested in a model selection process, where the most optimal and parsimonious model was chosen. The tested variables have earlier been shown to affect golden eagles, as slope, aspect, forest age, foraging habitat, suitable flight routes, human population density, roads, railways, power lines, wind power plants, hiking trails and clear cuts. The variables where applied in in ArcMAP at three different scales: nest scale (25 x 25 meter), proximate scale (a circle with the radius of 500 meter) and home range scale (a circle with the radius of 8253 meter). A preliminary test of the variables showed that all golden eagle nests were found in slopes with at least 5֯ degreesas well as in home ranges with human population density not more than 8 people/km2. Due to that a stratified analysis wasperformed. The variables where analysed by multiple logistic regression in R, where the occurrence of golden eagles’ nestswas compared towards random points in the landscape. All variables were also tested one by one by logistic regression. Afterperforming the multiple logistic regression, it was possible to apply its equation into ArcMap to obtain suitability maps withHSI values over Västernorrland’s county.The comparisons of different models show that it is better to combine different spatial scales in the model than only using one spatial scale. The result indicate that three different models might be the best, which all had different combinations of slope and aspect at nest scale and power lines at the proximate scale. Two of these models also include hiking trails and human population density, both at home range scale, in their equation. Since it was some unclarity about the causality between hiking trails and human population density, the conclusion was not to choose any of these as the final model. The final model was more parsimonious and had an additive effect from slope and southern aspect at the nest scale and an antagonistic effect from power lines at the proximate scale.This study clarifies that golden eagles’ habitat preferences for nesting sites during their breeding period is steep slopes (at minimum 5֯ degrees) in more southern aspects with few power lines in the proximate area surrounding the nest. Their homeranges are also situated in areas with less than 8 people/km2. The study also pinpoints a potential conflict between golden eagleand wind power planning, as golden eagles prefer steep slopes and remote areas, which also are valuable areas for wind powerplants. Golden eagles’ preference of remote areas also indicate that they might be affected by human persecution, why certainconservation effort should be focused into this issue. Out from the final model, you can find cluster in the landscape where youcan focus conservation management and restrict exploitation. Due to low number of wind power plants in the landscape, nothingcould be concluded about their effect on golden eagle in this study. An advice from the golden eagle’s perspective is to use theprecautionary principle and further plan wind power plants in areas which already have high disturbance, as for example closeto power lines or roads. The result also indicates that forest age from SLU Forest Map is not suitable for telling where to findgolden eagle nests. GIS-data over forest age would facilitate conservation management for plenty of species connected to theforest.Although good statistical results for the final model, cautions need to be taken in general, since neither population viability analysis have been included, nor changes over time in the landscape. Another issue is the low sample size, where a larger sample size would make it possible to perform profound calibration and validation of the data. To develop a more robust model, the advice is to include these into the model and use a larger sample size.