top of page

The Historical Trends of GEDI Biomass Predictions Across the Emerald Edge Rainforest application predicts yearly biomass observations (Mg/ha) between 2000 – 2021 within Jefferson and Clallam counties in Washington State. This model was developed by  implementing a regression model to determine whether remote sensing observations and various geospatial datasets, can be used to predict the biomass levels as determinized by NASA's Global Ecosystem Dynamic Investigation (GEDI) product. If a robust model can be developed, then confident biomass predictions can be made across different time periods to assess how levels have changed and identify opportunities for improvement. This effort was achieved by performing extensive data preprocessing on over 1,400 satellite images and implementing various machine learning algorithms such as, support vector machine, random forest, and xgboost, within the Python and R Studio ecosystems. Preliminary results show the model to produce accuracies over 78% with a root mean squared error ranging between 60-70 Mg/ha.

Link to Google Earth Engine Application: https://kolodzieds13.users.earthengine.app/view/gedi-biomass-predictions-across-the-emerald-edge-rainforest#lat=47.66;lon=-123.6537;zoom=9;

Deploying Machine Learning Models to Assess the Temporal Change of Biomass Across the Emerald Edge Rainforest
 

Normalized difference vegetation index (NDVI) is used for many purposes when analyzing remotely sensed data. The vegetative index NDVI can provide valuable information on vegetation growth based on the calculated NDVI value of a given area. Real time crop growth and biomass information can be obtained through remote sensing by using reflectance values in the red band and Near-infrared (NIR) regions to obtain NDVI values (Sishodia et al., 2020). The formula to calculate NDVI is shown in Eq.1 NDVI values range from −1 to 1 where the positive values indicate greenness or increasing vegetation areas and negative values indicate non-vegetated surfaces such as urban areas, bare landscapes, and bodies of water.

​

This project aims to generate a versatile script in Google Earth Engine (GEE) that can compute NDVI over a region of interest and a graph of how those observations are changing over time. Additionally, this tool was designed so that it does not confine users to a specific area but instead allows users to navigate any area they are interested in. Users will be able to seamlessly alter date ranges of images and create alternative vegetation indices with minor alterations to the code. Furthermore, the script will introduce users on how to create visualizations and will specifically display the specific vegetative indices within the assigned time frame.

Using Google Earth Engine to Assess Temporal Changes of NDVI Anywhere throughout the World
 

Applying Unsupervised Classification to Map Guatemala City.

GC_FalseColor_Original.JPG
GC_SVM.jpg
GC_SVM_legend.JPG

Guatemala City is a mountainous city in Central America surrounded by forests, grasslands, and cropland. This city is the capital of Guatemala and is home to approximately 995,130 according to macrotrends.net. On November 9th, 2008 the ASTER Satellite captured an image of Guatemala City with a 15-meter resolution and is composed of the green band, red band, Near Infrared (NIR) band, SWIR 1 band, SWIR 2 band, and thermal band. For this analysis, a land classification map was created through supervised classification methods through the ENVI environment. The user identified the various landcover classifications by delineating Regions of Interest (ROI) to supplement the supervised classification. Two separate classification maps were generated using the Maximum Likelihood Classification method and the Supported Vector Machine algorithm. Following the creation of these maps, an accuracy assessment was conducted by classifying randomized test sites (ground truth pixels) and cross-validating those classifications with our final product to create a confusion matrix highlighting user accuracy and producer accuracy.

​

Spectral Transforms: Simple Ratio, NDVI, and Tassel Cap Transformation on Madison Wisconsin

Spectral_transform_v2.png

This analysis compares how Simple Ratio, NDVI, and Tassel Cap transformations throughout the Madison Wisconsin.

bottom of page