Browsing by Subject "Random forest"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Data supporting "Predicting total phosphorus levels as indicators for shallow lake management"
    (2018-07-18) Vitense, Kelsey; Hanson, Mark A; Herwig, Brian R; Zimmer, Kyle D; Fieberg, John R; kelsey.vitense@gmail.com; Vitense, Kelsey
    This repository contains data supporting "Predicting total phosphorus levels as indicators for shallow lake management" in Ecological Indicators.
  • No Thumbnail Available
    Item
    Duluth 1-Meter Land Cover Classification (Impervious Surface Focused)
    (2016-08-01) Host, Trevor K; Rampi, Lian P; Knight, Joe F; jknight@umn.edu; Knight, Joe F
    A high-resolution (1-meter) land cover classification raster dataset was completed for three different geographic areas in Minnesota: Duluth, Rochester, and the seven-county Twin Cities Metropolitan area. This classification was created using high-resolution multispectral National Agriculture Imagery Program (NAIP) leaf-on imagery (2015), spring leaf-off imagery (2011- 2014), Multispectral derived indices, LiDAR data, LiDAR derived products, and other thematic ancillary data including the updated National Wetlands Inventory, LiDAR building footprints, airport, OpenStreetMap roads and railroads centerlines. These data sets were integrated using an Object-Based Image Analysis (OBIA) approach to classify 12 land cover classes: Deciduous Tree Canopy, Coniferous Tree Canopy, Buildings, Bare Soil, other Paved surface, Extraction, Row Crop, Grass/Shrub, Lakes, Rivers, Emergent Wetland, Forest and Shrub Wetland. We mapped the 12 classes by using an OBIA approach through the creation of customized rule sets for each area. We used the Cognition Network Language (CNL) within the software eCognition Developer to develop the customized rule sets. The eCognition Server was used to execute a batch and parallel processing which greatly reduced the amount of time to produce the classification. The classification results were evaluated for each area using independent stratified randomly generated points. Accuracy assessment estimators included overall accuracies, producers accuracy, users accuracy, and kappa coefficient. The combination of spectral data and LiDAR through an OBIA method helped to improve the overall accuracy results providing more aesthetically pleasing maps of land cover classes with highly accurate results.
  • Thumbnail Image
    Item
    Predicting forest carbon content and tree canopy cover on Forest Inventory and Analysis plots along a forest-prairie gradient
    (2021-12) Nolan, Jennifer
    Trees outside forests (TOF) are increasingly recognized as an important resource for carbon storage. TOF include trees used in agricultural windbreaks, urban ornamentals, and trees in non-forest ecosystems like prairie. This study used tree data from the US Forest Service’s Forest Inventory and Analysis (FIA) database to detect variation in forest carbon stocks and tree canopy cover between trees in forest ecosystems and trees in prairie ecosystems. A longitudinal transect was established, extending from Bismarck, ND to Duluth, MN to capture a gradient of mixed temperate forest in the east, prairie in the west, and a transitional zone in the middle. A number of potential transect sizes were evaluated using a series of power analyses (alpha = 0.05, power = 0.88, 0.83) and these determined that a transect radius of 50 km (total transect height 100 km) and transect length of 660 km was sufficient to capture the gradient and provide statistically significant results if differences existed. All FIA plots within the transect which had tree measurements taken from 2012 – 2018 (n = 4,155) were then used in a series of random forest analyses. The response variables of interest were carbon content (in megagrams per hectare) and percent live canopy cover, both at the plot level. Twenty-seven predictor variables were assessed: a few plot condition indicators from the FIA data, but mostly climate variables (30-year climate normals). Six random forest analyses were run: three examining canopy cover as the response variable using all plots, forest plots, and non-forest plots, and three examining carbon content within the same groupings. The power analysis lent confidence to the establishment of an effective study area and transect size; however, the random forest analyses were ultimately unable to consistently predict tree canopy cover or carbon content at the plot level. Although the random forest analyses did not provide statistically significant evidence for variation at the plot level, the patterns they revealed between which climate predictors performed best under forested and non-forested conditions are intriguing and may invite further investigation.