No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

<p dir="ltr">Canopy height is a fundamental metric for extracting valuable information about forested areas. Over the past decade, Lidar technology has provided a straightforward approach to measuring canopy height using various platforms such as terrestrial, unmanned aerial vehicle (UAV), airborne, and satellite sensors. However, satellite Lidar data, even with its global coverage, has a sparse sampling pattern that doesn’t provide continuous coverage over the globe. In contrast, satellites like LANDSAT offer seamless and widespread coverage of the Earth's surface through spectral data. Can we exploit the abundant spectral information from satellites like LANDSAT and ECOSTRESS to infer structural information obtained from Lidar satellites like Global Ecosystem Dynamic Investigation (GEDI)? This study aims to develop a deep learning model that can infer canopy height derived from sparsely observed Lidar waveforms using multi-sensor spectral data from spaceborne platforms. Specifically designed for localized site, the model focuses on county-level canopy height estimation, taking advantage of the relationship between canopy height and spectral reflectance that can be established in a local setting – something which might not exist universally. The study hopes to achieve a framework that can be easily replicable as height is a dynamic metric which changes with time and thus requires repeated computation for different time periods.</p><p dir="ltr">The thesis presents a series of experiments designed to comprehensively understand the influence of different spectral datasets on the model’s performance and its effectiveness in different types of test sites. Experiment 1 and 2 utilize Landsat spectral band values to extrapolate canopy height, while Experiment 3 and 4 incorporate ECOSTRESS land surface temperature and emissivity band values in addition to Landsat data. Tippecanoe County, predominantly composed of cropland, serves as the test site for Experiment 1 and 3, while Monroe County, primarily covered by forests, serves as the test site for Experiment 2 and 4. When compared to the Airborne Lidar dataset from the United States Geological Survey (USGS) – 3D Elevation Program (3DEP), the model achieves a Root Mean Square Error (RMSE) of 4.604m for Tippecanoe County using Landsat features while 5.479m for Monroe County. After integrating Landsat and ECOSTRESS features, the RMSE improves to 4.582m for Tippecanoe County but deteriorates to 5.860m for Monroe County. Overall, the study demonstrates comparable results to previous research without requiring feature engineering or extensive pre-processing. Furthermore, it successfully introduces a novel methodology for integrating multiple sources of satellite data to address this problem.</p>