Evaluating effectiveness of down-sampling for stratified designs and unbalanced prevalence in Random Forest models of tree species distributions in Nevada

Freeman, Elizabeth A.; Moisen, Gretchen G.; Frescino, Tracey S.

doi:10.1016/j.ecolmodel.2012.03.007

Cited by 61 publications

(39 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Climatic variables, their interactions and the Heat Load Index were used as predictors. RFs have multiple applications in ecological studies (Cutler et al, 2007), being widely used particularly for the prediction of species distributions (Iverson, 2008;Attorre et al, 2011;Freeman et al, 2012). This method is generally based on fewer assumptions than classical parametrical methods (e.g.…”

Section: Predictive Vegetation Modellingmentioning

confidence: 99%

Modelling the Last Glacial Maximum environments for a refugium of Pleistocene biota in the Russian Altai Mountains, Siberia

Hais

Komprdová

Ermakov

et al. 2015

Palaeogeography, Palaeoclimatology, Palaeoecology

View full text Add to dashboard Cite

Section: Predictive Vegetation Modellingmentioning

confidence: 99%

Modelling the Last Glacial Maximum environments for a refugium of Pleistocene biota in the Russian Altai Mountains, Siberia

Hais

Komprdová

Ermakov

et al. 2015

Palaeogeography, Palaeoclimatology, Palaeoecology

View full text Add to dashboard Cite

“…In addition, according to Kernes et al [88] tree models had a better understanding of the relationship and the boundaries, than logistic regression models for predicting the shrub cover spatial shifting. Also, according to [89], RFs are often used in very large geographical areas and when the number of samples in classes is unbalanced, the RF algorithm can be used with an acceptable level of accuracy for classification in such instances and it can be one of the factors for superiority over the SVM and k-NN. Naidoo et al [90] studied the possibility of modelling savanna tree species in the Kruger national park in South Africa, using integrated hyperspectral, light detection and ranging (LiDAR) data and the RF algorithm and their results showed that the RF model produced 87% accuracy.…”

Section: Discussionmentioning

confidence: 99%

Prediction of Dominant Forest Tree Species Using QuickBird and Environmental Data

Abdollahnejad

Panagiotidis

Shataee

et al. 2017

Forests

View full text Add to dashboard Cite

Abstract:Modelling the spatial distribution of plants is one of the indirect methods for predicting the properties of plants and can be defined based on the relationship between the spatial distribution of vegetation and environmental variables. In this article, we introduce a new method for the spatial prediction of the dominant trees and species, through a combination of environmental and satellite data. Based on the basal area factor (BAF) frequency for each tree species in a total of 518 sample plots, the dominant tree species were determined for each plot. Also, topographical maps of primary and secondary properties were prepared using the digital elevation model (DEM). Categories of soil and the climate maps database of the Doctor Bahramnia Forestry Plan were extracted as well. After pre-processing and processing of spectral data, the pixel values at the sample locations in all the independent factors such as spectral and non-spectral data, were extracted. The modelling rates of tree and shrub species diversity using data mining algorithms of 80% of the sampling plots were taken. Assessment of model accuracy was conducted using 20% of samples and evaluation criteria. Random forest (RF), support vector machine (SVM) and k-nearest neighbor (k-NN) algorithms were used for spatial distribution modelling of dominant species groups using environmental and spectral variables from 80% of the sample plots. Results showed physiographic factors, especially altitude in combination with soil and climate factors as the most important variables in the distribution of species, while the best model was created by the integration of physiographic factors (in combination with soil and climate) with an overall accuracy of 63.85%. In addition, the results of the comparison between the algorithms, showed that the RF algorithm was the most accurate in modelling the diversity.

show abstract

“…Sample balance is also important in 130 training data selection, as unbalanced training data (i.e., dramatically more or less training data for one or 131 multiple classes) may result in rare land cover types being under-represented relative to more abundant 132 classes, which may degrade the overall classification accuracy (Weiss & Provost, 2003;Estabrooks et al, 133 2004;Mellor et al, 2015). Techniques such as down-sampling of majority classes (Freeman et al, 2012), 134 over-sampling of minority classes (Ling and Li, 1998), and a combination of over-sampling and down-135 sampling training classes (Chawla et al, 2002) have been explored to alleviate the problem of unbalanced 136 training data. Outliers in the training data may also impact the training process (Radoux et al, 2014) and 137 can be eliminated based on spectral or spatial distances.…”

Section: Introduction 41mentioning

confidence: 99%

Optimizing selection of training and auxiliary data for operational land cover classification for the LCMAP initiative

Zhu

Gallant

Woodcock

et al. 2016

ISPRS Journal of Photogrammetry and Remote Sensing

147

103

View full text Add to dashboard Cite

15The U.S. Geological Survey's Land Change Monitoring, Assessment, and Projection (LCMAP) initiative 16 is a new end-to-end capability to continuously track and characterize changes in land cover, use, and 17 condition to better support research and applications relevant to resource management and environmental 18 change. Among the LCMAP product suite, there are annual land cover maps that will be available to the 19 public. This paper describes an approach to optimize the selection of training and auxiliary data for 20 deriving the thematic land cover maps based on all available clear observations from Landsats 4-8. land cover classification, Landsat 40 community, data needs are changing and innovative products and mapping approaches to characterize 47 land cover are required to meet the community's appetite for science-quality, geospatial land cover data. 48The emergence of land change science has had a particularly significant influence on land cover mapping 49 and monitoring requirements. Driven by concerns about climate change and resource sustainability, 50 Rindfuss et al. (2004) and Turner et al. (2007) defined land change science as a framework to understand 51the interactions between people and nature that lead to changes in the type, magnitude, condition, and 52 location of land use and cover. Implicit is the need to use remote sensing, modeling, and synthesis to 53 understand and explain land change rates, causes, and consequences. Meeting the diverse needs of land 54 cover data users means that the data producers must transform traditional approaches into new strategies 55 that generate higher quality products with additional land cover variables, more detailed legends, and 56 more frequent geospatial information and statistics on change. 57The USGS response to the growing requirements for land use, cover, and condition data, information, and 58 knowledge is the Land Change Monitoring, Assessment, and Projection (LCMAP) initiative. LCMAP is 59 an end-to-end capability to continuously track and characterize changes in land cover, use, and condition 60 and translate such information into assessments of current and historical processes of cover and change 61 that can serve as the science foundation for research and applications relevant to resource management 62 and environmental change. As implied in the name, key elements include mapping and monitoring 63 change, analyzing and assessing the rates, causes, and consequences of change, and providing scenario-64 based projections of future land change. LCMAP continues the long-standing USGS commitment to land 65 cover research and development and represents the foundation for a Landsat-based Federal land 66 monitoring system. Furthermore, LCMAP will stimulate geographic research on understanding the 67 connections between human activity and natural systems, and ultimately represents an essential step 68 towards better understanding the combined impacts of climate and land use and land cover change. 69

show abstract

Evaluating effectiveness of down-sampling for stratified designs and unbalanced prevalence in Random Forest models of tree species distributions in Nevada

Cited by 61 publications

References 28 publications

Modelling the Last Glacial Maximum environments for a refugium of Pleistocene biota in the Russian Altai Mountains, Siberia

Modelling the Last Glacial Maximum environments for a refugium of Pleistocene biota in the Russian Altai Mountains, Siberia

Prediction of Dominant Forest Tree Species Using QuickBird and Environmental Data

Optimizing selection of training and auxiliary data for operational land cover classification for the LCMAP initiative

Contact Info

Product

Resources

About