editor_options
markdown wrap 72

Pickstone-development

This repository contains R code for processing, analysing and visualising the data presented in *"Estimating Canopy Height in Tropical Forests: Integrating Airborne LiDAR and Multi-Spectral Optical Data with Machine Learning"*by Brianna Pickstone under the supervision of Dr Andrew Cunliffe and Dr Hugh Graham for her MSc in Applied Data Science (Environment and Sustainability) at the University of Exeter in partnership with Permian Global.

Data Source: This study uses a combination of data:

1. The Airborne LiDAR data from Permian Global, supplied by Dr Hugh Graham
2. The PlanetScope optical data from Planet Labs, supplied by Dr Hugh Graham
3. Sentinel-2 Data, downloaded from the Sentinel-2-Cloud Optimised GeoTIFFs (COG)
4. Digital Terrain Model from Copernicus, downloaded from the Microsoft Planetary Computer STAC

Contact: [email protected]{.email}

Scripts

This repository contains the following scripts within the "Script" folder:

Preprocessing_S2.R This script contains code for downloading the Sentinel-2 Data from the COG. This also includes the generation of the canopy height model from LiDAR, accessing the digital terrain model, calculation of topographic metrics, calculation of vegetation spectral indices, and the creation of the 10 m S2 resolution data cube.

PlanetScope_and_LiDAR_processing.R This script contains code for pre-processing both the LiDAR data and optical Planet Labs data. This includes generation of the canopy height model, calculation of topographic metrics and calculations for the vegetation spectral indices from the optical data.

Combining_Data_S2_PS.R This script contains the code for combining the Sentinel-2 Data at 10 m resolution with the PlanetScope data at 10 m resolution

Intial_Plots.R This script contains code for generating a true colour image of the PlanetLabs data and Sentinel-2 data as well as the Canopy Height Model from LiDAR.

Map_Kat_Project.R This script contains the code for generating a map of the study area

Spatial Autocorrelation.R This script contains code for assessing spatial autocorrelation within the data, this includes the calculation of Moran's I metric and visualisation through a variogram. The code also provides the generation of the plots to visualise the resampling plan for spatial cross-validation.

MLR.R main script for undertaking multiple linear regression for prediction of the canopy height model

RF.R main script for undertaking Random Forest for prediction of the canopy height model

resample_2DCNN.R main script for undertaking a Convolutional Neural Network for prediction of the canopy height model

Performance_models.R This script creates the bar chart of comparing the performance of models using RMSE, MAE, R2 and time (mins)

Scatterplots.R This script creates the visualisation of comparing the predicted canopy height vs. the LiDAR derived Canopy Height of each data source and all three machine learning algorithms.

CHM.plots.R This script creates the visualisation of the predicted Canopy Height Model and the difference in the Predicted Canopy Height and the LiDAR derived Canopy Height of the best performing PlanetScope 3m, PlanetScope 10 m, Sentinel-2 10 m and combined data set at 10 m resolution.

Planet_Spectral_Indices.R This script is written by Dr Hugh Graham as an example of how to calculate vegetation spectral indices

Archive folder contains outdated and/or code that was used as testing different models/ feature selection

Data

This repository contains the data within the "Data" folder:

sample_P3.csv and sample_PS3cube.tif contain the PlanetScope data at 3 m resolution, which includes x and y coordinates, spectral bands, vegetation indices, topographic metrics and LiDAR-derived CHM

sample_P10.csv and sample_PS10cube.tif contain the PlanetScope data at 10 m resolution, which includes x and y coordinates, spectral bands, vegetation indices, topographic metrics and LiDAR- derived CHM

sample_S2*.csv and sample_S2cube.tif* contain the Sentinel-2 data at 10 m resolution, which includes x and y coordinates, spectral bands, vegetation indices, topographic metrics and LiDAR- derived CHM

sample_comb.csv and sample_combcube.tif contain the combined data of Sentinel-2 and PlanetScope data at 10 m resolution, which includes x and y coordinates, spectral bands, vegetation indices, topographic metrics and LiDAR- derived CHM

This data is just a subsample of the data. The actual size of the data is noted below:

DataSize:

• PlanetScope 3 m data table: 17,405,911 rows x 29 columns

• PlanetScope 10 m data table: 1,566,545 x 29 columns

• Sentinel-2 data table: 1,564,685 rows x 32 columns

• Combined data table: 1,563,954 rows x 53 columns