Land Cover Classification in an Ecuadorian Mountain Geosystem Using a Random Forest Classifier, Spectral Vegetation Indices, and Ancillary Geographic Data

Ayala-Izurieta, Johanna E.; Márquez, Carmen O.; García, Vega; Recalde-Moreno, Celso; Rodríguez-Llerena, Marcos V.; Damián-Carrión, Diego A.

doi:10.3390/geosciences7020034

Cited by 37 publications

(31 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The ML approach is one of the classical parametric statistical classifiers and is widely used for LULC classification [53]. Several studies have also evaluated alternative, recently-developed machine-learning algorithms for land-cover classification, such as support vector machine (SVM) methods [54], decision trees (DT) [55], and random-forest models (RF) [56,57]. Schneider [55] observed that the DT and SVM classifiers outperformed the ML classifier in the context of highly dynamic land-cover change and 'fuzzy' multi-signal classes around the Chinese cities of Chengdu, Xi'An, and Kunming.…”

Section: Data Processing and Analysismentioning

confidence: 99%

Urban Expansion Occurred at the Expense of Agricultural Lands in the Tarai Region of Nepal from 1989 to 2016

et al. 2018

View full text Add to dashboard Cite

Recent rapid urbanization in developing countries presents challenges for sustainable environmental planning and peri-urban cropland management. An improved understanding of the timing and pattern of urbanization is needed to determine how to better plan urbanization for the near future. Here, we describe the spatio-temporal patterns of urbanization and related land-use/land-cover (LULC) changes in the Tarai region of Nepal, as well as discuss the factors underlying its rapid urban expansion. Analyses are based on regional time-series Landsat 5, 7 and 8 image classifications for six years between 1989 and 2016, representing the first long-term observations of their kind for Nepal. During this 27-year period, gains in urban cover and losses of cultivated lands occurred widely. Urban cover occupied 221.1 km 2 in 1989 and increased 320% by 2016 to a total 930.22 km 2 . Cultivated land was the primary source of new urban cover. Of the new urban cover added since 1989, 93% was formerly cultivated. Urban expansion occurred at moderately exponential rates over consecutive observation periods, with nearly half of all urban expansion occurring during 2006-2011 (305 km 2 ). The annual rate of urban growth during 1989-1996 averaged 3.3% but reached as high as 8. 09% and 12.61% during 1996-2001 and 2011-2016, respectively. At the district level, the rate of urban growth and, by extension, agricultural loss, were weakly related to total population growth. Variability in this relationship suggests that concerted urban-growth management may reduce losses of agricultural lands relative to historic trends despite further population growth and urbanization. Urbanization and LULC change in the Tarai region are attributable to significant inter-regional migration in a context of poor urban planning and lax policies controlling the conversion and fragmentation of peri-urban cultivated lands. Urban expansion and farmland loss are expected to continue in the future.

show abstract

Section: Data Processing and Analysismentioning

confidence: 99%

Urban Expansion Occurred at the Expense of Agricultural Lands in the Tarai Region of Nepal from 1989 to 2016

et al. 2018

View full text Add to dashboard Cite

show abstract

“…We also computed the correlation matrix between quantitative input variables. Even though the impact of correlated variables on RF models is not yet fully demonstrated [25,75], we removed highly correlated variables (r > 0.8) to avoid redundancy. To select the most explicative attributes, a RF model with a high number of trees (N = 1500) was used.…”

Section: Attribute Selectionmentioning

confidence: 99%

Developing a Random Forest Algorithm for MODIS Global Burned Area Classification

Ramo

Chuvieco

2017

Remote Sensing

View full text Add to dashboard Cite

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.

show abstract

“…The parks (plots 7 and 9) had the lowest NDVI values (0.156 and 0.228). Values close to zero indicate sparse vegetation on bare soil [24]. The NDVI is the remote sensing product most widely used worldwide to analyse and map differences in vegetation types and plant phenology [25] including to estimate the diversity of trees over large areas when the vegetation is at the maximum growing season [26][27][28][29].…”

Section: Discussionmentioning

confidence: 99%

Ecological Structure of a Tropical Urban Forest in the Bang Kachao Peninsula, Bangkok

Sommeechai

Wachrinrat

Dell

et al. 2018

Forests

View full text Add to dashboard Cite

Abstract:Rapid urbanization has changed the structure and function of natural ecosystems, especially floodplain ecosystems in SE Asia. The ecological structure of vegetation stands and the usefulness of satellite images was investigated to characterize a disturbed tropical urban forest located in the Chao Phraya River lower floodplain, Thailand. Nine sample plots were established on the Bang Kachao Peninsula (BKP) within 4 tropical forest types in an urban area: rehabilitation forest, home-garden agroforestry, mangrove and park. The tree habitats were beach forest, swamp forest, moist evergreen forest, dry evergreen forest, mangrove forest and abandoned orchard or home-garden. Normalized difference vegetation index (NDVI) values obtained from Landsat 7 satellite images were correlated with plant structure from field surveys. NDVI had the highest relationship with stand factors for number of families, number of species, Shannon-Weiner index and total basal area. Linear regression predicted well the correlation between NDVI and stand factors for families and basal area. NDVI trends reflected urban tropical forest typing and biodiversity, being high in rehabilitation and mangrove forests, moderate in home-gardens and low in parks. We suggest that the application of NDVI for assessments can be useful for future planning, monitoring and management of the BKP and hence may contribute for increasing biodiversity and complexity of these urban forests.

show abstract

Land Cover Classification in an Ecuadorian Mountain Geosystem Using a Random Forest Classifier, Spectral Vegetation Indices, and Ancillary Geographic Data

Cited by 37 publications

References 36 publications

Urban Expansion Occurred at the Expense of Agricultural Lands in the Tarai Region of Nepal from 1989 to 2016

Urban Expansion Occurred at the Expense of Agricultural Lands in the Tarai Region of Nepal from 1989 to 2016

Developing a Random Forest Algorithm for MODIS Global Burned Area Classification

Ecological Structure of a Tropical Urban Forest in the Bang Kachao Peninsula, Bangkok

Contact Info

Product

Resources

About