 Naveen Aachi Project Manager RMSI Pvt Ltd India naveen.aachi@rmsi.com MVK Satya Prasad Project Lead RMSI Pvt Ltd India Paper Overview During the hurricane and high wind seasons, there is usually extensive damage to residential and commercial establishments which result in losses which are usually very difficult to assess. Strong winds not only damage the structures but also uproot trees there by disturbing the power supply lines and damages caused by uprooting trees too is high where density of tree canopy is higher. For one such research project done by RMSI for an Insurance Company based in U.S., it was observed that a lot of claims made during 2006-2007 season were for damages to properties due to tree falls( As shown in Figure 1). However these damages were significantly higher for single family residential dwellings. In-order to assess the risks from the tree falls, one should have the information on the tree coverage within residential areas and mapping this kind of a data through conventional methods is an expensive approach. RMSI has worked out a cost effective solution and developed a county wise mapping of percentage tree cover within single family residential dwellings for the assessment of exposure due to this hazard. The objective of this paper is to provide an insight into the methodology and the concept which can be made global using similar techniques for evaluation of losses for the insurance industry.  Figure 1: Tree Fall during a Hurricane damaging a house in Riva, MD (Source: emhglobalimages.com) Inputs and Process Flow For the county wise mapping of percentage tree cover within single family residential buildings, the inputs used were National Land Cover Database (NLCD) land cover grid, NLCD tree canopy grid of 2001 vintage and US counties shape file. The NLCD land cover and tree canopy grids were obtained from The Multi-Resolution Land Characteristics (MRLC) Consortium (www.mrlc.gov). MRLC consortium is a group of federal agencies which have joined together to coordinate the production of a comprehensive land cover database for the entire US. The NLCD database consists of several independent data layers of which the ones required were 21 classes of land-cover data derived from the imagery and ancillary data, and per pixel percentage tree canopy. The process flow chart is shown in Figure 2.  Figure 2: Tree Fall during a Hurricane damaging a house in Riva, MD (Source: emhglobalimages.com) Approach 3.1. Data Extraction and Overlay Analysis The tree canopy information available in img format was converted to ESRI compatible grids in order to facilitate the spatial analysis with the land cover information containing the classification cells of residential classes. All the residential classes having information of single family residential dwellings were reclassified to a single class before the overlay analysis with Tree Canopy grids. Figure3 shows Reclassified land cover grid  Figure 3:Reclassified land cover grid on the left and the tree canopy grid on the right for one of the United States zones The overlaying method required both the canopy grids and land cover grids to be of the same resolution and grids snapped for accurate results. Tree canopy grids were overlaid on the land cover details and only those pixels corresponding to single family residential dwellings were selected for the analysis. US counties information was projected to Albers Equal Area meters projection to suit NLCD projection details. The tree canopy area calculations were done after ascertaining the areas of each county followed by surface area percentage of urban classes. 3.2. Tree Canopy Area Calculation The NLCD tree canopy grid is at a resolution of 30m and area of each cell is 30m x 30m which is 900 sq m. Each class in the tree canopy grid which has values from 1 to 100 represents the percentage area distribution of tree canopy. Example: If a canopy class has value 1, it represents 1% tree cover in a single pixel of 30m resolution. The tree canopy coverage area is calculated as 1% of 900 sq m which would be 9 sq m. Similarly for class 50, the tree canopy coverage area would be 450 sq m and for class 100 it would be 900 sq m. The class wise area was then summed to get the total tree coverage area within each county. Figure below shows tree canopy area calculation for some counties in New Jersey(Table 1).  Table 1: Tree Canopy area calculation for some counties of New Jersey Results For each county once the single family residential areas and tree canopy areas were derived, percentage tree canopy within single family residential areas was obtained. This was performed for all the counties of US. The results were tabulated as shown Table 2:  Table 2: Percentage tree canopy within single family residential areas for each county A thematic view of the Northeastern states below shows the percentage tree cover for the Single family residential dwellings shown in figure 4  Figure 4: Thematic of percentage tree cover within single family residential areas |