Autors: Nikova H., Deliyski, R. T., Tashev, T. A.
Title: ANALYSIS OF WILDFIRE EARLY PREDICTION MODELS
Keywords: wildfire, prediction, statistical analysis, uncertainty

Abstract: This paper presents statistical comparative analysis of five forest fire prediction methods. The methods used in the study are normalized Burning index (BI), normalized Energy Release Component (ERC), Severe Fire Danger Index (SFDI) as well as binary regression forest fire prediction model and output of the two-layer artificial neural network. The results of the analysis show that BI’, ERC’ and SFDI’ in both states – fire and no fire as well as regression model and the neuron network in case of fire occurrence has to be presented with standard uncertainty. Based on the defined coefficient of mutually exclusive states determination, the regression model and ANN provide the best results for both states. Additionally the RMSE is determined for all studied models and compared.

References

    Issue

    , pp. 78-82, 2023, Bulgaria, KING, ISSN 2683-0337

    Вид: публикация в международен форум, публикация в реферирано издание