Preprocess Ground Truth Data
Description
This tool preprocesses ground truth features and ground truth extents to create the ground truth raster input to the WIM workflow.
Usage
Use this tool as a preliminary step to using WIM to ensure your ground truth raster input is in the proper format. Preprocess Ground Truth Data will return a raster with correct snap raster and cell size settings. The output raster will render known landcover classes (e.g., wetland or nonwetland) as integer raster values. Areas within the ground truth extents that are not known (i.e., unknown background areas) will be rendered as NoData and ignored in subsequent analyses. Users must take note of the integer value that represents specific classification targets. Integer values in the output raster must begin at 0 and increase by 1. The output raster will subsequently be used to create training and/or testing data. For that reason, the output raster dictates the cells that will be used for Random Trees training and accuracy assessments.
Parameters
| Parameter Name | Type | Direction | Data Type | Dialog Reference |
|---|---|---|---|---|
| Input Ground Truth Features | Required | Input | Feature Layer | Feature class containing the set of known ground truth features. Features can be polygons or points. Input point features will be buffered by a single linear unit (according to the spatial reference of the Input Cell Size Constraint) before rasterizing. Features can delineate all or some of the known target features within the ground truth extents. |
| Input Ground Truth Extents | Required | Input | Feature Layer | Feature class that delineates the ground truth limits and dictates size of output raster. If Input Ground Truth Features represent all known wetlands in the area, the Input Ground Truth Extents should encompass all known nonwetland area. If Input Ground Truth Features represent some known wetlands and/or nonwetland classes, choose extents that mimic a minimum bounding polygon around the known features. |
| Input Cell Size Constraint | Required | Input | Raster Layer | Raster that dictates the output cell size and snap raster of the output raster. This should be the DEM from which topographic predictor variables will be calculated. If other predictor variables are going to be calculated in place of topographic variables (e.g., imagery-based indices), the source raster for those variables should be used instead. |
| Features represent more than one target class | Required | Input | Boolean | True if the Input Ground Truth Features represent more than one target class (e.g., multiple wetland classes; a single wetland class, agricultural areas, developed areas). If True, specify the field that distinguishes the classes. The output raster will represent each class as a unique integer value, beginning from 0. False if features in the Input Ground Truth Features belong to the same target class (e.g., all features are a single, generic "wetland" class). The output raster will represent all ground truth features with a value of zero. |
| Class of areas outside of ground truth features is known | Required | Input | Boolean | True if Input Ground Truth Features encompass all known wetlands, or other target class(es), within the Input Ground Truth Extents. If True, areas outside of the ground truth features and within the ground truth extents are assumed to be true nonwetland/background area. The output raster will represent these areas by a unique integer value.False if Input Ground Truth Features encompass only some known wetlands, or other target class(es), within the Input Ground Truth Extents. If False, areas outside of the ground truth features and within the ground truth extents are unknown and cannot be assumed to represent the presence or absence of a target class. The output raster will represent these areas as NoData. |
| Class Identifier Field | Optional | Input | Field | The field that distinguishes feature types in the input ground truth feature class. |
| Output Ground Truth Raster (*.tif) | Required | Output | Raster Layer | Output raster representation of ground truth data. Cells in this raster with data are assumed to be known occurrences of specific target classes. This raster is used for training and testing WIM models. |