Volume IV-4/W4
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., IV-4/W4, 393-397, 2017
https://doi.org/10.5194/isprs-annals-IV-4-W4-393-2017
© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 4.0 License.
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., IV-4/W4, 393-397, 2017
https://doi.org/10.5194/isprs-annals-IV-4-W4-393-2017
© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 4.0 License.

  13 Nov 2017

13 Nov 2017

GLOBALLY-APPLICABLE PREDICTIVE WILDFIRE MODEL   A TEMPORAL–SPATIAL GIS BASED RISK ANALYSIS USING DATA DRIVEN FUZZY LOGIC FUNCTIONS

G. van den Dool1,2 G. van den Dool
  • 1Faculty of Engineering, Science, and Mathematics, School of Geography, University of Southampton, UK
  • 2CoreLogic, 7 rue Drouot, 75009, Paris, France

Keywords: Wildfire, GIS, Fuzzy Logic, Data Driven

Abstract. This study (van den Dool, 2017) is a proof of concept for a global predictive wildfire model, in which the temporal–spatial characteristics of wildfires are placed in a Geographical Information System (GIS), and the risk analysis is based on data-driven fuzzy logic functions. The data sources used in this model are available as global datasets, but subdivided into three pilot areas: North America (California/Nevada), Europe (Spain), and Asia (Mongolia), and are downscaled to the highest resolution (3-arc second).

The GIS is constructed around three themes: topography, fuel availability and climate. From the topographical data, six derived sub-themes are created and converted to a fuzzy membership based on the catchment area statistics. The fuel availability score is a composite of four data layers: land cover, wood loads, biomass, biovolumes. As input for the climatological sub-model reanalysed daily averaged, weather-related data is used, which is accumulated to a global weekly time-window (to account for the uncertainty within the climatological model) and forms the temporal component of the model. The final product is a wildfire risk score (from 0 to 1) by week, representing the average wildfire risk in an area. To compute the potential wildfire risk the sub-models are combined usinga Multi-Criteria Approach, and the model results are validated against the area under the Receiver Operating Characteristic curve.