Land cover monitoring using remotely sensed data requires robust classification methods which allow for the accurate mapping of complex land cover and land use categories. Random forest (RF) is a powerful machine learning classifier that is relatively unknown in land remote sensing and has not been evaluated thoroughly by the remote sensing community compared to more conventional pattern recognition techniques. Key advantages of RF include: their non-parametric nature; high classification accuracy; and capability to determine variable importance. However, the split rules for classification are unknown, therefore RF can be considered to be black box type classifier. RF provides an algorithm for estimating missing values; and flexibility to perform several types of data analysis, including regression, classification, survival analysis, and unsupervised learning.

In this paper, the performance of the RF classifier for land cover classification of a complex area is explored. Evaluation was based on several criteria: mapping accuracy, sensitivity to data set size and noise. Landsat-5 Thematic Mapper data captured in European spring and summer were used with auxiliary variables derived from a digital terrain model to classify 14 different land categories in the south of Spain. Results show that the RF algorithm yields accurate land cover classifications, with 92% overall accuracy and a Kappa index of 0.92. RF is robust to training data reduction and noise because significant differences in kappa values were only observed for data reduction and noise addition values greater than 50 and 20%, respectively. Additionally, variables that RF identified as most important for classifying land cover coincided with expectations. A McNemar test indicates an overall better performance of the random forest model over a single decision tree at the 0.00001 significance level.

An assessment of the effectiveness of a random forest classifier for land-cover classification

The random forest algorithm, proposed by L. Breiman in 2001, has been extremely successful as a general-purpose classification and regression method. The approach, which combines several randomized decision trees and aggregates their predictions by averaging, has shown excellent performance in settings where the number of variables is much larger than the number of observations. Moreover, it is versatile enough to be applied to large-scale problems, is easily adapted to various ad hoc learning tasks, and returns measures of variable importance. The present article reviews the most recent theoretical and methodological developments for random forests. Emphasis is placed on the mathematical forces driving the algorithm, with special attention given to the selection of parameters, the resampling mechanism, and variable importance measures. This review is intended to provide non-experts easy access to the main ideas.

/pdf/a-random-forest-guided-tour-3vnpr6ncpy.pdf

A random forest guided tour

The random forest algorithm, proposed by L. Breiman in 2001, has been extremely successful as a general-purpose classification and regression method. The approach, which combines several randomized decision trees and aggregates their predictions by averaging, has shown excellent performance in settings where the number of variables is much larger than the number of observations. Moreover, it is versatile enough to be applied to large-scale problems, is easily adapted to various ad-hoc learning tasks, and returns measures of variable importance. The present article reviews the most recent theoretical and methodological developments for random forests. Emphasis is placed on the mathematical forces driving the algorithm, with special attention given to the selection of parameters, the resampling mechanism, and variable importance measures. This review is intended to provide non-experts easy access to the main ideas.

A Random Forest Guided Tour

Aim Spatial modelling techniques are increasingly used in species distribution modelling. However, the implemented techniques differ in their modelling performance, and some consensus methods are needed to reduce the uncertainty of predictions. In this study, we tested the predictive accuracies of five consensus methods, namely Weighted Average (WA), Mean(All), Median(All), Median(PCA), and Best, for 28 threatened plant species.

Location  North-eastern Finland, Europe.

Methods  The spatial distributions of the plant species were forecasted using eight state-of-the-art single-modelling techniques providing an ensemble of predictions. The probability values of occurrence were then combined using five consensus algorithms. The predictive accuracies of the single-model and consensus methods were assessed by computing the area under the curve (AUC) of the receiver-operating characteristic plot.

Results  The mean AUC values varied between 0.697 (classification tree analysis) and 0.813 (random forest) for the single-models, and from 0.757 to 0.850 for the consensus methods. WA and Mean(All) consensus methods provided significantly more robust predictions than all the single-models and the other consensus methods.

Main conclusions  Consensus methods based on average function algorithms may increase significantly the accuracy of species distribution forecasts, and thus they show considerable promise for different conservation biological and biogeographical applications.

https://onlinelibrary.wiley.com/doi/pdf/10.1111/j.1472-4642.2008.00491.x

Evaluation of consensus methods in predictive species distribution modelling

With the growing body of literature assessing the impact of invasive alien plants on resident species and ecosystems, a comprehensive assessment of the relationship between invasive species traits and environmental settings of invasion on the characteristics of impacts is needed. Based on 287 publications with 1551 individual cases that addressed the impact of 167 invasive plant species belonging to 49 families, we present the first global overview of frequencies of significant and non-significant ecological impacts and their directions on 15 outcomes related to the responses of resident populations, species, communities and ecosystems. Species and community outcomes tend to decline following invasions, especially those for plants, but the abundance and richness of the soil biota, as well as concentrations of soil nutrients and water, more often increase than decrease following invasion. Data mining tools revealed that invasive plants exert consistent significant impacts on some outcomes (survival of resident biota, activity of resident animals, resident community productivity, mineral and nutrient content in plant tissues, and fire frequency and intensity), whereas for outcomes at the community level, such as species richness, diversity and soil resources, the significance of impacts is determined by interactions between species traits and the biome invaded. The latter outcomes are most likely to be impacted by annual grasses, and by wind pollinated trees invading mediterranean or tropical biomes. One of the clearest signals in this analysis is that invasive plants are far more likely to cause significant impacts on resident plant and animal richness on islands rather than mainland. This study shows that there is no universal measure of impact and the pattern observed depends on the ecological measure examined. Although impact is strongly context dependent, some species traits, especially life form, stature and pollination syndrome, may provide a means to predict impact, regardless of the particular habitat and geographical region invaded.

https://onlinelibrary.wiley.com/doi/pdf/10.1111/j.1365-2486.2011.02636.x

A global assessment of invasive plant impacts on resident species, communities and ecosystems: the interaction of impact measures, invading species' traits and environment

Classification procedures are some of the most widely used statistical methods in ecology. Random forests (RF) is a new and powerful statistical classifier that is well established in other disciplines but is relatively unknown in ecology. Advantages of RF compared to other statistical classifiers include (1) very high classification accuracy; (2) a novel method of determining variable importance; (3) ability to model complex interactions among predictor variables; (4) flexibility to perform several types of statistical data analysis, including regression, classification, survival analysis, and unsupervised learning; and (5) an algorithm for imputing missing values. We compared the accuracies of RF and four other commonly used statistical classifiers using data on invasive plant species presence in Lava Beds National Monument, California, USA, rare lichen species presence in the Pacific Northwest, USA, and nest sites for cavity nesting birds in the Uinta Mountains, Utah, USA. We observed high classification accuracy in all applications as measured by cross-validation and, in the case of the lichen data, by independent test data, when comparing RF to other common classification methods. We also observed that the variables that RF identified as most important for classifying invasive plant species coincided with expectations based on the literature.

/pdf/random-forests-for-classification-in-ecology-4nfz47gm9v.pdf

Random Forests for Classification in Ecology

Phragmites australis (Cav.) Trin. ex Steud. (common reed), already one of the world's most widespread plant species, has realized rapid range expansion in coastal wetlands of North America in the past century, but little is known about P. australis range expansion in inland wetland systems. We used genetic analyses, aerial photographs, field surveys, and a greenhouse experiment to study the extent and mechanism of nonnative P. australis invasion of Utah wet- lands. We collected and genetically analyzed 39 herbarium samples across the state and 225 present-day samples from northern Utah's major wetland complexes. All samples collected before 1993 and all samples collected outside the major wetlands of northern Utah, including some as recent as 2001, were identified as native (haplotypes A, B, D, and H). Only 10 (4%) of the present-day samples were native, each from small, discrete, low-density stands; the remaining sam- ples were nonnative (haplotype M). Our earliest nonnative sample was collected near the Great Salt Lake in 1993. Around the Great Salt Lake, which contains 40% of Utah's wetlands, P. australis cover has increased from 20% to 56% over the past 27 years—an increase that appears attributable to the nonnative strain. In a 3-month-long greenhouse experiment, the nonnative haplotype grew taller, had more aboveground biomass, and had a greater above- to below- ground biomass ratio than the native haplotypes regardless of nitrogen, phosphorus, or water availability. Nonnative P. australis is rapidly invading the wetlands of northern Utah. Areas in Utah where the native P. australis remains should be identified and protected.

/pdf/nonnative-phragmites-australis-invasion-into-utah-wetlands-266bjcyc2s.pdf

Nonnative Phragmites australis Invasion into Utah Wetlands

Species distribution models (SDMs) were built with US Forest Inventory and Analysis (FIA) publicly available plot coordinates, which are altered for plot security purposes, and compared with SDMs built with true plot coordinates. Six species endemic to the western US, including four junipers (Juniperus deppeana var. deppeana, J. monosperma, J. occidentalis, J. osteosperma) and two pinons (Pinus edulis, P. monophylla), were analyzed. The presence–absence models based on current climatic variables were generated over a series of species-specific modeling extents using Random Forests and applied to forecast climatic conditions. The distributions of predictor variables sampled with public coordinates were compared to those sampled with true coordinates using t tests with a Bonferroni adjustment for multiple comparisons. Public- and true-based models were compared using metrics of classification accuracy. The modeled current and forecast distributions were compared in terms of their overall areal agreement and their geographic mean centroids. Comparison of the underlying distributions of predictor variables sampled with true versus public coordinates did not indicate a significant difference for any species at any extent. Both the public- and true-based models had comparable classification accuracies across extent for each species, with the exception of one species, J. occidentalis. True-based models produced geographic distributions with smaller areas under current and future scenarios. The greatest areal difference occurred in the species with the lowest modeled accuracies (J. occidentalis), and had a forecast distribution which diverged severely. The other species had forecast distributions with similar magnitudes of modeled distribution shifts.

https://www.fs.fed.us/rm/pubs_journals/2013/rmrs_2013_gibson_j001.pdf

Using Publicly Available Forest Inventory Data in Climate-Based Models of Tree Species Distribution: Examining Effects of True Versus Altered Location Coordinates

Populations of pinons and junipers across the interior west have been highly dynamic over the last two centuries, undergoing an overall expansion but punctuated with regional mortality. Accumulating demographic studies across the interior west indicate the drivers of expansion and contraction of populations are compounded by regional land use legacies, but have an underlying climatic component. The key implication is that pinon and juniper distributions respond to climatic variation across spatial and temporal scales by expansion from, and contraction to, areas of long-term persistence relative to their centurial (pinons) and millennial (junipers) life spans. Individualistic climatic tolerances among pinon and juniper species lead to temporally dynamic assemblages in the landscape that vary regionally.

https://www.nrs.fs.fed.us/pubs/gtr/gtr-nrs-p-105papers/18gibson-p-105.pdf

Jacob Gibson

Papers

Random Forests for Classification in Ecology

Nonnative Phragmites australis Invasion into Utah Wetlands

Using Publicly Available Forest Inventory Data in Climate-Based Models of Tree Species Distribution: Examining Effects of True Versus Altered Location Coordinates

Expansion and contraction tension zones in western pinon-juniper woodlands under projected climate change