Date: May 4th, 2016
THE USBurnedBuildings PACKAGE ACCOMPANIES:
Alexandre, P.M., Susan I. Stewart, Nicholas S. Keuler, Murray K. Clayton, Miranda H. Mockrin, Avi Bar-Massada, Alexandra D. Syphard, Volker C. Radeloff, 2016, Factors related to building loss due to wildfires in the conterminous United States, Ecological Applications, in press.
It contains a copy of the main data files exactly as used for this paper, as well as other accompanying files (see below).
Please, contact me if you have suggestions, find errors, inconsistencies, or any other bug in the file. As well, please let me know about your uses of this data and send manuscripts and reprints when available. I�ll be happy to help you in any case, as far as I can. email - patmalico@fulbrightmail.org
Please, use these data files for peaceful purposes, enjoy doing science with them as I have enjoyed writing the paper quoted above, and learn as much as you can with them.
******************************** CONTENTS ***********************************
The data package associated with the publication cited above contains five files:
1. README.txt - This file. Including a description of the variables.
2. digitized_buildings_2000_2010.csv � Contains all the digitized burned and survived buildings that were within fire perimeters that occurred between 2000 and 2010 in the conterminous U.S.
3. all_clusters.csv � Contains all the clusters that were obtained from the digitized buildings and respective attributes as described below.
4. linear_regression_data.csv � Data used for the final analysis, where the clusters contain at least 8 buildings and at least one burned building.Not all variables in this file were included in the regression analysis as they are merely informative. The variables that were removed before running the regression were all the columns that appear before the "resp".
5. logistic_regression_data.csv - Data used for the final analysis, where the clusters contain at least 8 buildings. Not all variables in this file were included in the regression analysis as they are merely informative. The variables that were removed before running the regression were all the columns that appear before the "resp".
*****************************************************************************
Variables in �digitized_buildings_2000_2010.csv�:
cluster_id Each row in this file represents a building (point). However, the unit of analysis in this study was the �cluster�. We defined clusters by setting up a radius of 100 m around each building and if the circles touched or overlapped, then they would form a cluster. Each cluster can have several buildings inside, meaning points with the same cluster ID belong to the same cluster.
Year Year of the fire event.
type_b What the building was classified as (UB= unaffected building and BB = burned building)
fire_id This ID comes directly from the MTBS dataset. More information at http://www.mtbs.gov/ (assessed on May 4th, 2016). All the buildings, and therefore the clusters that were within a fire perimeter will have the same fire ID. We did not change this code.
state We conducted this analysis for the conterminous US only. Each cluster was assigned the state in which the majority of buildings was in.
NA_L2KEY The column corresponds to the Omernik Ecoregions Level II key that we kept unchanged. For more information see: https://archive.epa.gov/wed/ecoregions/web/html/na_eco.html#Level II
NA_L1KEY The column corresponds to the Omernik Ecoregions Level II key that we kept unchanged. For more information see: https://archive.epa.gov/wed/ecoregions/web/html/na_eco.html#Level I (assessed on May 4th, 2016)
elev Elevation (meters) of the location of the building. Data obtained from the LADNFIRE dataset. More info at: http://www.landfire.gov/topographic.php (assessed on May 4th, 2016)
sloped Slope in degrees at the building location. Obtained in a GIS environment from elevation.
sw Southwestnes is a cosine-transformation of aspect to create an index of �southwestness,� which could account for the influence of solar radiation and aspect on fuel properties and fire behavior (Syphard et al. 2012).
tpi It stands for Topographic Position Index, and we calculated it from the LANDFIRE elevation data using an algorithm that defines standardized threshold values for the difference between a cell elevation value and the average elevation of the cells around that cell measured in standard deviations from the mean (Jenness 2006).
near_bld It stands for �Nearest building� and it represents the distance in meters to the nearest building. We used the tool �Near� in ArcMap 10.x to calculate this distance.
bw40m Number of buildings within 40 m.
nlcd2006 It stands for National Land Cover Dataset. Each point was assigned the land cover code in which it falls into. More information at http://www.mrlc.gov/finddata.php (assessed on May 4th, 2016)
X and Y coordinates:
Projected Coordinate System: NAD_1983_Albers
Projection: Albers
False_Easting: 0.00000000
False_Northing: 0.00000000
Central_Meridian: -96.00000000
Standard_Parallel_1: 29.50000000
Standard_Parallel_2: 45.50000000
Latitude_Of_Origin: 23.00000000
Linear Unit: Meter
Geographic Coordinate System: GCS_North_American_1983
Datum: D_North_American_1983
Prime Meridian: Greenwich
Angular Unit: Degree
*****************************************************************************
Variables in �all_clusters.csv�:
X and Y coordinates of the cluster centroid:
Projected Coordinate System: NAD_1983_Albers
Projection: Albers
False_Easting: 0.00000000
False_Northing: 0.00000000
Central_Meridian: -96.00000000
Standard_Parallel_1: 29.50000000
Standard_Parallel_2: 45.50000000
Latitude_Of_Origin: 23.00000000
Linear Unit: Meter
Geographic Coordinate System: GCS_North_American_1983
Datum: D_North_American_1983
Prime Meridian: Greenwich
Angular Unit: Degree
cluster_id Each row in this file represents a cluster of buildings (points). We set up a radius of 100 m around each building and if the circles touched or overlapped, then they would form a cluster. Each cluster is given an ID where the first four number represent the year of the fire event. Example; cluster_ID 200014 represent a cluster of buildings that were inside a fire that occurred in the year 2000. The number 14 represents cluster #14 and it has no special meaning. These ID are one of the attributes of the spatial features in a GIS file.
Year Year of the fire event.
fire_id This ID comes directly from the MTBS dataset. More information at http://www.mtbs.gov/ (assessed on May 4th, 2016). All the buildings, and therefore the clusters that were within a fire perimeter will have the same fire ID. We did not change this code.
State We conducted this analysis for the conterminous US only. Each cluster was assigned the state in which the majority of buildings was in.
NA_L2KEY The column corresponds to the Omernik Ecoregions Level II key that we kept unchanged. For more information see: https://archive.epa.gov/wed/ecoregions/web/html/na_eco.html#Level II
NA_L1KEY The column corresponds to the Omernik Ecoregions Level II key that we kept unchanged. For more information see: https://archive.epa.gov/wed/ecoregions/web/html/na_eco.html#Level I (assessed on May 4th, 2016)
BB Stands for �Burned Building�. It represents the number of burned buildings inside the cluster.
UB Stands for �Unaffected Building�. It represents the number of surviving buildings inside the cluster.
bld_clust Total number of buildings inside the clusters. We used the following equation: =BB+UB
area_clust Cluster area in m2.
bld_den It stands for �Building density� and we used the following equation: =bld_clust/area_clust (number of buildings.m-2)
near_clust It stands for �Nearest Cluster� and it represents the distance in meters to the nearest cluster. We used the tool �Near� in ArcMap 10.x to calculate this distance.
rdens It stands for �road density� and we calculated using the total road length (meters) inside the cluster divided by the cluster area (m2).
elev Elevation (meters) of the location of the building. Data obtained from the LADNFIRE dataset. More info at: http://www.landfire.gov/topographic.php (assessed on May 4th, 2016)
sloped Slope in degrees at the building location. Obtained in a GIS environment from elevation.
sw Southwestnes is a cosine-transformation of aspect to create an index of �southwestness,� which could account for the influence of solar radiation and aspect on fuel properties and fire behavior (Syphard et al. 2012).
tpi It stands for Topographic Position Index, and we calculated it from the LANDFIRE elevation data using an algorithm that defines standardized threshold values for the difference between a cell elevation value and the average elevation of the cells around that cell measured in standard deviations from the mean (Jenness 2006).
near_bld It stands for �Nearest building� and it represents the average of the distances in meters of each building inside the cluster to the nearest building.
near_edge It stands for �Nearest distance to edge of the cluster� and it represents average of the distances, in meters, of each building to the nearest edge of the cluster.
bld_disp It stands for �Building dispersion� and we used the following formula: Building dispersion=(st dev. of dist. among buildings within a cluster)/(mean distance among buildings within a cluster)
nlcd2006 It stands for National Land Cover Dataset. Each cluster was giving the dominant land cover class, using a majority rule. More information about the original dataset at: http://www.mrlc.gov/finddata.php (assessed on May 4th, 2016)
CONTAG and CONNECT
Landscape metrics �contagion� and �connectance index� calculated using the software FRAGSTATS (http://www.umass.edu/landeco/research/fragstats/fragstats.html - assessed on May 4th, 2016). For more information about the meaning of the variables, see: http://www.umass.edu/landeco/research/fragstats/documents/Metrics/Contagion%20-%20Interspersion%20Metrics/Metrics/L115%20-%20CONTAG.htm (assessed on May 4th, 2016)
http://www.umass.edu/landeco/research/fragstats/documents/Metrics/Connectivity%20Metrics/Metrics/L122%20-%20CONNECT.htm (assessed on May 4th, 2016)
PLAND1, CONNECT1, PLAND3, CONNECT3
Class metrics �Percentage of Land of class 1 and 3� and �connectance index for classes 1 and 3�. Class 1 represents land cover classes that were considered Highly-flammable, and class 3 represents land cover classes that were considered Non-flammable. We removed class 2 because it was highly correlated to classes 1 and 3 and it was never part of the models in the preliminary analysis. More information about PLAND and CONNECT at:
http://www.umass.edu/landeco/research/fragstats/documents/Metrics/Area%20-%20Density%20-%20Edge%20Metrics/Metrics/C4%20-%20PLAND.htm (assessed on May 4th, 2016)
http://www.umass.edu/landeco/research/fragstats/documents/Metrics/Connectivity%20Metrics/Metrics/C122%20-%20CONNECT.htm (assessed on May 4th, 2016)
*****************************************************************************
Variables in �linear_regression_data.csv� and �logistic_regression_data.csv�:
Meaning and units of the variables are the same as explained for the file �all_clusters.csv�, except for the resp variable.
For the linear regression data = Stands for �response variable� and it represents the percentage of burned buildings inside the cluster. We used the following equation: =(BB/UB)*100.
For the logistic regression data = Stands for �response variable� and it is a binary variable where zero represent a cluster where all the buildings survived the fire and one represents a cluster where at least one building burned. We used the following equation: =IF(BB=0,�0�,�1�)