Open AccessApplying Multivariate Statistical Methods for Predicting Pinus Forest Fire Danger at Bidoup-Nui Ba National Park
Author(s) -
Ле Ван Хыонг,
Нгуен Нгок Киенг,
Нгуен Данг Хой,
Данг Хунг Куонг
Publication year - 2021
Publication title -
trudy karadagskoj naučnoj stancii im. t.i. vâzemskogo - prirodnogo zapovednika ran
Language(s) - English
Resource type - Journals
ISSN - 2712-9586
DOI - 10.21072/eco.2021.13.05
Subject(s) - discriminant function analysis , statistics , multivariate statistics , mathematics , linear discriminant analysis , canonical correlation , correlation coefficient , canonical analysis , data set , pearson product moment correlation coefficient , mahalanobis distance
The paper presents results of applying multivariate statistical methods (CCA: canonical correlation analysis and DFA: discriminant function analysis) for determining canonical correlation between a set of variables {T, H, m1, K} and a set of variables {Pc, Tc} (T: temperature, H: relative humidity, m1: mass of dry fuels, K: burning coefficient, K = m1/M, with M: total mass of fire fuels, Pc: % burned fuels and Tc: burningtime) as well as through results of discriminant function analysis DFA to set up models of predicting forest fire danger at Bidoup - Nui Ba National Park. From research data in November, December, January, February and March in the period of 2015-2017 from 340 sampling plots (each 2mx2m), at Bidoup - Nui Ba National Park, we carry on data processing on Excel (calculating) and Statgraphics (multivariate statistical methods: CCA&DFA). Three results were revealed from our analysis: (i) Canonical correlation between a set of variables {T, H, m1, K} and a set of variables {Pc, Tc} is highly significant (R = 0.675581 & P = 3.17*10-58<< 0.05); therefore, we can use a set of variables {T, H, m1, K} in models of predicting forest fire danger, (ii) Coefficients of standardized & unstandardized canonical discriminant functions (SCDF &UCDF) and Fisher classification function (FCF) are determined, (iii) Setting up two models of predicting forest fire danger (Mahalanobis distance model & Fisher classification function model).