Landuse landcover using ArcGIS Machine Learning Tools

Select the best satellite image by considering crop stages and NDVI to avoid misclassifying land use, choosing the mature crop stage around February–March or August–September.

Explore how a support vector machine classifies images in GIS by finding a hyperplane that separates desert and agricultural land with maximal margin using labeled training data.

Explain how random forest uses many decision trees to classify satellite image pixels into land cover types such as forest, water, and urban areas, and majority voting determines the label.

Download Landsat 30 meter and Sentinel two 10 meter imagery via Earth Explorer, then select cloud-free images for land use land cover classification and organize raw data for ArcGIS.

Stack Landsat bands in ArcGIS with composite bands to create a multi-band TIFF for land use and land cover classification using an SVM model, and view infrared bands.

Download a sentinel-2 10m data set from the Copernicus portal by creating an account, logging in, defining study area, and selecting March imagery with red, green, blue, and near-infrared bands.

Stack sentinel-2 ten-meter bands 2, 3, 4, and 8 in ArcGIS to generate RGB and infrared imagery for land use and land cover classification.

Identify possible land use and land cover classes from satellite images and build a reference table including water, settlements, agriculture, forest, hills shade, roads, river beds, wasteland, and sandy areas.

Identify land use classes on a Landsat image and collect precise, nonmixed training samples for agriculture, forest, settlements, water, and riverbed, then save them for supervised ArcGIS classification.

train a support vector machine classifier using the inbuilt trainer in the segmentation and classification tools, loading the input raster and training samples, and saving the trained file for classification.

Apply SVM classification in ArcGIS to classify a raster using a support vector machine. Learn how training data, input classifier definitions, and post-classification corrections contribute to accurate land-use labeling.

Compare random forest and support vector machine classifications on Landsat imagery using a random tree classifier, training samples, and raster classification; reveals urban area delineation and limitations in mountainous terrain.

Train a Sentinel-2 classifier in ArcGIS using high-resolution imagery and SVM, with field survey points and samples to map crops, forests, urban areas, and water.

Train an SVM classifier in ArcGIS with sentinel-2 data using color mean and standard deviation to map land use—urban, agriculture, cropland, vegetation, wetlands—with post-classification corrections.

Use the ArcGIS reclassify tool to correct misclassified land-use values by mapping old class values to new, consistent categories, and assign colors for clear output visualization.

Apply post classification pixel corrections in ArcGIS by drawing error polygons, converting them to raster, assigning a single value, and mosaicking with the original Sentinel image to fix urban misclassifications.

Generate an accuracy report for a landuse landcover classification in ArcGIS using accuracy assessment points, stratified or equalize sampling, and a confusion matrix to report user and producer accuracy.

Learn to cut the study area from a land use image in ArcGIS by exporting a shapefile for the area and using Extraction by Mask to crop the classified image.

Calculate land use area in square kilometers by counting pixels from ArcGIS imagery, using 30 by 30 meter resolution in utm projection and a field calculator.

Learn to compute land use change between 2001 and 2011 using ArcGIS, producing a confusion matrix with a tabulate area tool, reclassification, and conversion to square kilometers for interpretation.

Send your trained model file (such as SVM or random forest) to a friend, who can classify a similar satellite image using nearby dates and matching reflectance.

Create a final land use map layout by adjusting the image, adding a title, north arrow, scale bar, grid, legend, and exporting as a jpg.