Skip to main content

Table 5 Summary of accuracy assessment of vulnerability-risk models MaxEnt and DNN, based on forest fire events in 2017, 2018, and 2019. FALSE represents the number of incorrectly detected fire events. TRUE represents the number of truly detected forest fire events. The probability of detection was computed using the values in FALSE and TRUE

From: Forest fire pattern and vulnerability mapping using deep learning in Nepal

Method

Fire year

Forest fire-vulnerability risk class (n)

FALSE (n)

TRUE (n)

Probability of detection

Very low

Low

High

Very high

MaxEnt

2019

295

155

432

354

450

786

0.64

2018

251

156

432

357

407

789

0.66

2017

325

172

515

329

497

844

0.63

Overall

871

483

1379

1040

1354

2419

0.64

DNN

2019

206

138

538

354

344

892

0.72

2018

171

134

525

366

305

891

0.74

2017

252

175

551

363

427

914

0.68

Overall

629

447

1614

1083

1076

2697

0.71