Assessing Canopy Shade in Tropical Headwater Streams
Description
Forests play a vital role in maintaining healthy riparian ecosystems, influencing hydrological processes and water quality. This study aims to characterize the forest canopy structure using drone-based LiDAR technology to model shading in headwater streams. Specifically, we evaluate the effectiveness of fractional canopy cover metrics derived from high-resolution 3D point clouds in predicting solar radiation interception (fPAR) by riparian forests. Ecological Context: By linking forest fractional cover derived from LiDAR with light interception data, the aim was to better understand the relation of canopy cover with light interception. This approach would enhance studies that consider light interception in ecological processes, such as gas exchange and biomass productivity in tropical riparian forests. This project involves two interconnected analyses: (1) Forest Cover Extraction from LiDAR Data. (2) Correlate to the fractions of light intercepted by riparian canopy, using the fractional canopy cover (fcover) across study sites, providing spatially explicit forest structure data. Modeling and spatial prediction of light interception (fPAR): Using environmental and forest variables, a predictive model for fPAR was developed based on field measurements and rasterized predictor data, thereby enabling the spatial mapping of the potential light interception. High-resolution drone LiDAR data were collected from five riparian zones of low-order headwater streams. Simultaneous measurements of solar radiation in open areas and under forest canopy allowed us to quantify the fraction of light intercepted by vegetation (fPAR). Fractional canopy cover was calculated following the methodology by Hopkinson and Chasmer (2009). A Random Forest model calibrated with field measurements assessed the relationship between fractional cover and solar radiation interception, with model performance evaluated using leave-one-out cross-validation, showing strong predictive capacity (R² = 0.79, RMSE ≈ 0.061, MAE ≈ 0.045). This approach demonstrates the potential of integrating high-resolution LiDAR and ground-based data for ecological research and forest management. In the particular case for understanding canopy effects on aquatic ecosystems in tropical regions.
Files
Steps to reproduce
Data Collection: Acquire point cloud data from riparian zones using a drone-based LiDAR sensor. Collect simultaneous solar radiation measurements: In open areas (PAR_background) Under forest canopy (PAR_canopy) Record georeferenced locations of sampling plots using an SP-60 GNSS receiver. Data Processing: Estimate light interception: 𝑓PAR = 1 - PAR_canopy/PAR_background Convert raw LiDAR point clouds into high-resolution 3D models (e.g., using lidR in R). Calculate fractional canopy cover (FC) for each plot following Hopkinson and Chasmer (2009). Buffer each plot center by 2 meters. Derive the fraction of LiDAR points representing canopy (above 2 m height) within this buffer. Organize solar radiation measurements and corresponding FC values into a dataset. Model Calibration and Evaluation: Fit a Random Forest model with FC as predictor and fPAR (interception fraction) as response. Use leave-one-out cross-validation (LOOCV) to evaluate model predictions. Analyze model performance metrics (R², RMSE, MAE). Application: Apply the calibrated model to LiDAR-derived FC maps to generate spatial predictions of canopy shading across the study area, and to use these predictions to inform ecological assessments related to stream light availability and ecosystem productivity. Note: To ensure proper reproducibility, it is essential to update any changes in the work directory, file list, and file locations accordingly. Additionally, the RStudio environment and package versions used should be considered and maintained consistent with those during analysis. Users are encouraged to document package versions and session details using sessionInfo() in R before running the scripts.
Institutions
- Universidade de Sao Paulo