Developing a Random Forest Algorithm for MODIS Global Burned Area Classification

Abstract

This paper aims to develop a global burned area (BA) algorithm for MODIS BRDF-corrected images based on the Random Forest (RF) classifier. Two RF models were generated, including: (1) all MODIS reflective bands; and (2) only the red (R) and near infrared (NIR) bands. Active fire information, vegetation indices and auxiliary variables were taken into account as well. Both RF models were trained using a statistically designed sample of 130 reference sites, which took into account the global diversity of fire conditions. For each site, fire perimeters were obtained from multitemporal pairs of Landsat TM/ETM+ images acquired in 2008. Those fire perimeters were used to extract burned and unburned areas to train the RF models. Using the standard MD43A4 resolution (500 × 500 m), the training dataset included 48,365 burned pixels and 6,293,205 unburned pixels. Different combinations of number of trees and number of parameters were tested. The final RF models included 600 trees and 5 attributes. The RF full model (considering all bands) provided a balanced accuracy of 0.94, while the RF RNIR model had 0.93. As a first assessment of these RF models, they were used to classify daily MCD43A4 images in three test sites for three consecutive years (2006)(2007)(2008). The selected sites included different ecosystems: Australia (Tropical), Boreal (Canada) and Temperate (California), and extended coverage (totaling more than 2,500,000 km 2 ). Results from both RF models for those sites were compared with national fire perimeters, as well as with two existing BA MODIS products; the MCD45 and MCD64. Considering all three years and three sites, commission error for the RF Full model was 0.16, with an omission error of 0.23. For the RF RNIR model, these errors were 0.19 and 0.21, respectively. The existing MODIS BA products had lower commission errors, but higher omission errors (0.09 and 0.33 for the MCD45 and 0.10 and 0.29 for the MCD64) than those obtained with the RF models, and therefore they showed less balanced accuracies. The RF models developed here should be applicable to other biomes and years, as they were trained with a global set of reference BA sites.

Object based burned area mapping with random forest algorithm

“…20 different random forest classifier models which were a combination of the ten numbers of trees (100-1000 increase by 100) and two numbers of attributes (m = 4, 5 which indicate integer value the lower and upper values close to √m) was built for determining the optimum parameter. The result was evaluated Balanced Accuracy (Kuhn & Johnson, 2013;Ramo & Chuvieco, 2017) defined in Table 2.…”

“…There is only one study is available in the literature for the mapping of burned areas with the random forest based classifier. This classifier was developed to extract the burned areas on the global scale from the MODIS images (Ramo & Chuvieco, 2017).…”

An Exploration of Some Pitfalls of Thematic Map Assessment Using the New Map Tools Resource

“…Although papers detailing good practices for working with map accuracy are available and widely cited [34,35], some of these issues have not, to our knowledge, been noted previously. Others have been discussed in the literature, but remain sources of confusion in published studies [24][25][26]. For this reason, the examples that are presented here are a useful clarification of commonly held misconceptions.…”

“…However, it is unclear what meaning that these averages have when all cover classes, from very rare to the most common, are weighted equally. Even so, it is not hard to find published examples of averaged user's or producer's accuracies ( [24][25][26]). Note that in some cases, the unweighted mean producer's accuracy of a binary matrix is sometimes called the 'balanced accuracy rate' ( [25]).…”

Hyperspectral Classification Based on Texture Feature Enhancement and Deep Belief Networks

“…Hyperspectral classification is a key technique employed in aforementioned applications. A majority of classification methods have been promoted in the last several decades to distinguish physical objects and classify each pixel into a unique land-cover label, such as maximum likelihood [5], minimum distance [6], K-nearest neighbors [7,8], random forests [9], Bayesian models [10,11], neural networks, etc., and their improvements [12][13][14][15]. Among these supervised classifiers, one of the most important classifiers is kernel-based support vector machine (SVM), which can also be considered as a kind of neural network.…”