Modelling temporal variation of fire-occurrence towards the dynamic prediction of human wildfire ignition danger in northeast Spain
Models of human-caused ignition probability are typically developed from static or structural points of view. This research analyzes the intra-annual dimension of fire occurrence and fire-triggering factors in NE Spain and moves forward towards more accurate predictions. Applying the Maximum Entropy...
| Autores: | , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2018 |
| País: | España |
| Institución: | Universitat de Lleida (UdL) |
| Repositorio: | Repositori Obert UdL |
| OAI Identifier: | oai:repositori.udl.cat:10459.1/65970 |
| Acceso en línea: | https://doi.org/10.1080/19475705.2018.1526219 http://hdl.handle.net/10459.1/65970 |
| Access Level: | acceso abierto |
| Palabra clave: | Wildfire Ignition danger Human drivers Temporal dimension |
| Sumario: | Models of human-caused ignition probability are typically developed from static or structural points of view. This research analyzes the intra-annual dimension of fire occurrence and fire-triggering factors in NE Spain and moves forward towards more accurate predictions. Applying the Maximum Entropy algorithm (MaxEnt) and using wildfire data (2008–2011) and GIS and remote sensing data for the explanatory variables, we construct eight occurrence data scenarios by splitting wildfire records into the four seasons and then separating each season into working and non-working days. We assess model accuracy using a cross-validation k-fold procedure and an operational validation with 2012 data. Results report a substantial contribution of accessibility across models, often coupled with Land Surface Temperature. In addition, we observe great temporal variability, with WAI strongly influencing winter models, whereas distance to roads stands out during working days. Model performances stand consistently above 0.8 AUC in all temporal scenarios, with outstanding predictive effectiveness during summer months. The comparison among static-to-dynamic approaches reveals superior performance of simulations considering temporal scenarios, with AUC values from 0.7 to 0.85. Overall, we believe our approach is reliable enough to derive dynamic predictions of human-caused fire occurrence. |
|---|