Land-Cover Classification Using MaxEnt: Can We Trust in Model Quality Metrics for Estimating Classification Accuracy?

Abstract

MaxEnt is a popular maximum entropy-based algorithm originally developed for modelling species distribution, but increasingly used for land-cover classification. In this article, we used MaxEnt as a single-class land-cover classification and explored if recommended procedures for generating high-quality species distribution models also apply for generating high-accuracy land-cover classification. We used remote sensing imagery and randomly selected ground-true points for four types of land covers (built, grass, deciduous, evergreen) to generate 1980 classification maps using MaxEnt. We calculated different accuracy discrimination and quality model metrics to determine if these metrics were suitable proxies for estimating the accuracy of land-cover classification outcomes. Correlation analysis between model quality metrics showed consistent patterns for the relationships between metrics, but not for all land-covers. Relationship between model quality metrics and land-cover classification accuracy were land-cover-dependent. While for built cover there was no consistent patterns of correlations for any quality metrics; for grass, evergreen and deciduous, there was a consistent association between quality metrics and classification accuracy. We recommend evaluating the accuracy of land-cover classification results by using proper discrimination accuracy coefficients (e.g., Kappa, Overall Accuracy), and not placing all the confidence in model’s quality metrics as a reliable indicator of land-cover classification results.

An Improvised Machine Learning Model Based on Mutual Information Feature Selection Approach for Microbes Classification

“…The accuracy metric considers the values of true positives and true negatives while the F1-score [ 76 ] helps to make crucial facts that arise due to the values of false negatives and false positives. Since, the F1-score is the harmonic mean of recall and precision, it gives a better idea about the incorrectly classified cases.…”

One-Class Classification of Natural Vegetation Using Remote Sensing: A Review

“…The literature review confirmed that thresholding is the parameter that influences the accuracy of OCCs the most [42,73] and that a threshold remains challenging to set [51,53]. Although Chignell et al [46] noted that a continuous map is more informative than a categorized map for end-users, few of the studies reviewed avoided thresholding (Table 8).…”

“…Although Sentinel-1 and Sentinel-2 data have been freely available since 2014 and 2016, respectively, their use for mapping vegetation in OCCs remains low (n = 6). Indeed, most studies used multi temporal Sentinel-2 images [45,[71][72][73][74] whereas one study used Sentinel-1 images [75].…”

“…Specifically, several studies showed the utility of variables that characterize vegetation phenology derived from multispectral satellite data to map natural vegetation, such as MODIS [82,83], Landsat [54], ASTER [84], RapidEye [51], or hyperspectral airborne data [85]. Although Sentinel-2 data are increasingly relevant for OCCs, their use remained limited to a few acquisitions [45,[71][72][73]. Notably, Delalay et al [74] reported the utility of Sentinel-2 time-series in OCCs for mapping vegetation classes in Nepal.…”

“…Fernandes et al [147] reported that the Kappa index is more sensitive to degradation of the quality of reference data than the AUC or TSS. Morales and Fernández [73] suggested not placing all confidence in statistical indices of classifier quality (e.g., AUC, AIC, TSS), but rather in conventional Kappa or global accuracy indices derived from independent reference datasets.…”