Innovative Methodology for Geospatial Insights

Biomass

The estimation of Above Ground Biomass (AGB) is based on data acquired through the Global Ecosystem Dynamics Investigation (GEDI) mission, which was operational from 2019 to 2023. GEDI's primary objective was to scan the Earth's surface using Light Detection and Ranging (LiDAR) technology. As a result of this mission, one of the valuable products generated is the AGB data, which is represented as point-based measurements in units of Mg/ha.

The points derived from the GEDI LiDAR measurements serve as the training dataset for our machine-learning model. To ensure that the training data only includes points from forested areas, we utilized the woodland delimitation dataset obtained from the UK government for the year 2020. Additionally, we acquired Sentinel 2 Optical Imagery and Sentinel 1 Synthetic Aperture Radar (SAR) imagery for the same year. To focus solely on forested regions, we applied a masking process to the Sentinel 2 and Sentinel 1 imagery using the woodland delimitation data. Subsequently, we identified the corresponding points from the GEDI dataset that align with these specific forested areas. As a result, we now possess a set of points that are spatially associated with forests, and their respective pixels from the Sentinel imagery, which will be utilized for training the AGB prediction model.

The training process involves employing a random forest Machine Learning model, utilizing the aforementioned points as the training dataset. Once the model is trained, it becomes capable of predicting the AGB for a given area that possesses Sentinel 2 and Sentinel 1 imagery for different years. This predictive capability enables us to estimate the AGB of forested regions over time using the provided remote sensing data.

In summary, our scientific approach leverages GEDI LiDAR point data, Sentinel 2 and Sentinel 1 imagery, and a random forest Machine Learning model to estimate AGB in forested areas. This integrated methodology holds promise for ecological and environmental studies, aiding in forest monitoring, carbon accounting, and other related applications.

Forest/Non Forest

The Forest/NonForest model is built upon data collected during the operational period of the Global Ecosystem Dynamics Investigation (GEDI) mission, which spanned from 2019 to 2023. GEDI's primary objective was to utilize Light Detection and Ranging (LiDAR) technology to scan the Earth's surface. One of the mission's valuable outputs is the Height data, which comprises point-based measurements representing tree heights in meters.

To achieve temporal sampling, we adopt the following procedure: Upon selecting an area of interest, we create a buffer around it and establish a boundary box. Subsequently, we divide this box into four quadrants, with each quadrant being associated with a different year. GEDI and Sentinel data are then collected for each year within these quadrants. The training dataset for our machine-learning model consists of points derived from the GEDI LiDAR measurements. To ensure that the training data exclusively encompasses points from forested regions, we apply a threshold. Points with a height greater than 15 meters are categorized as belonging to the forest class, while points with a height below 5 meters are considered non-forest.

For further analysis, we acquire Sentinel 2 Optical Imagery and Sentinel 1 Synthetic Aperture Radar (SAR) imagery, all corresponding to the same year as the GEDI data. The training process entails utilizing a random forest Machine Learning model with the aforementioned points as the input dataset.

In conclusion, our scientific approach combines GEDI LiDAR point data, Sentinel 2 and Sentinel 1 imagery, and a random forest Machine Learning model to estimate forest cover on the area of interest. This integrated methodology holds considerable promise for ecological and environmental studies, particularly in applications such as forest monitoring and carbon accounting. By leveraging remote sensing data, we can gain valuable insights into forest dynamics and contribute to various environmental management efforts.

Indicators