Search

Highlights • CORINE LC and CCI-LC maps were analyzed and compared in detail in Europe. • LUCAS dataset was used for validation purposes. • Dissimilarities were mainly found in grassland and shrubland covers, important for wildfire analysis. • Wildfire estimation at European and at local scale was done using LC interfaces derived from the two maps. • Differences between models were more meaningful at the local scale and the performance improved despite the LC map used. Abstract Updated and harmonized land cover (LC) data is essential for wildfire estimation in fire-prone areas as is the case in southern Europe. CORINE Land cover (CLC) and ESA Climate Change Initiative Land Cover (CCI-LC) maps have been analyzed and compared their performance in the estimation of wildfire occurrence in Europe at regional and local scales for the period 2010–2014. LC maps legends were harmonized and similarities and discrepancies were compared. Overall agreement between the two maps for the whole Europe was ˜75%. Forest and agriculture showed the largest agreement, while shrubland and grassland the lowest. Quantity and allocation disagreements were calculated including exchange and shift components (Pontius and Santacruz, 2014) which provided detailed information about the contribution of each class to the overall disagreement. Spatial discrepancies were found in areas where grassland and shrubland were the dominant classes as in United Kingdom or East Turkey. Land Use and Coverage Area frame Survey (LUCAS) was used as ground truth for validation purposes. The agreement with LUCAS was slightly higher for CCI-LC (59%) than for CLC (56%). Generalized Linear Models (GLM), based on presence-absence of wildfires, were used to estimate wildfire occurrence at 3 × 3 km grid cell resolution from both LC maps at the European scale. LC interfaces and climatic variables (temperature and precipitation) where used as explicative variables while fires from European Forest Fire Information System EFFIS (2010–2014 period) were used as response variable. Wildfire occurrence was also estimated with the two maps at local scale in a test region (Zamora, Spain) using a more precise location of the response variable (x, y fire ignition points). At the European scale models obtained within the two maps showed similar results. CCI-LC model sensitivity was 77.26%, specificity 25.89% and omission error 22.74% while CLC model sensitivity was 75.68%, specificity 29.99% and omission error 24.32%. However, CLC performed slightly better in terms of the percent correct classification (62%). In the test region the models achieved better results in terms of specificity (66.07% and 68.93% for CCI-LC and CLC models respectively) and percent correct classification (˜68% for CLC model). At local scale CLC model performed better than CCI-LC model. Wildfire occurrence estimation was more accurate at local scale because of the differences in the spatial accuracy of the response variable used. [ABSTRACT FROM AUTHOR]

The country reports were presented in one of the working groups at the 11th BOBCATSSS 2003 symposium in Torun. The reports show how European Union programmes influenced the development of national information policies. Experiences from three countries were compared: Lithuania, Slovenia and the already full EU member United Kingdom. The first part of the presentation explains what national information policies really mean for these countries, then followed an account of the EU programmes in which they participated. Finally the authors assess the impact of these programmes on the development of the national information policy, in order to reveal the differences in impact between the three countries and predict likely future developments. [ABSTRACT FROM AUTHOR]