Carbon Storage & Tree Preservation at River Ridge Ranch

This study relied on a variety of data sources. Available preliminary data includes:

• UAV images were collected by California State University - Long Beach professors and students from different majors and environmental projects. The imagery was collected at different periods between 2018 and 2021.

• The nadir and off-nadir imagery were analyzed with ArcGIS Pro (Version 2.8), ERDAS Imagine, and ENVI Software to build and georeference an orthomosaic image, create tree polygons, and calculate areas.
• The 5-band images were utilized to build a Normalized Difference Vegetation Index (NDVI), a Normalized Difference Red Edge (NDRE), a Soil-Adjusted Vegetation Index (SAVI), and a Canopy Height Model (CHM), which allowed us to distinguish and classify trees among other vegetation.

• In-field data was taken from tree samples on a 53-acre portion of the ranch, named Gary’s Swath (or Block). Four types of trees were mainly found: Blue Oak (Quercus douglasii), Coastal Live Oak, herein mentioned as Live Oak (Quercus agrifolia), Valley Oak (Quercus lobata), and California Buckeye (Aesculus californica).

• In-field data collection included tree heights, diameter at breast height (DBH), and canopy sizes, as well as some altitude, incline, and plant health assessment information.
• Heights were measured using a clinometer/tangent height gauge and trigonometric calculations as well as a TruPulse 360 laser rangefinder
• DBH was measured with a girthing tape.
• Canopy size and location were recorded by using a real-time kinematic (RTK) GPS device (model: Arrow Gold RTK GNSS receiver, firmware 6.0Aa01)
• Collected info was logged into a Trimble Juno 3B device with TerraSync software.

Climate change has been recognized as one of the greatest challenges faced by present and future generations. Studies have shown that carbon dioxide (CO ₂ ), a powerful greenhouse gas (GHG), is one of the main precursors of climate change, greatly due to fuel combustion along with increasing urbanization and deforestation (Pidwimy & Jones 2010).

Trees are an essential carbon sink as they remove CO ₂  from the atmosphere and can be of great aid for combating and mitigating the effects of global warming.

Previous studies, attempting to measure carbon sequestration by trees, have used a diversity of methods (Garbulsky et al. 2008). Commonly used methods usually rely on measurements of gross-primary production (GPP) (Piao et al. 2006), tree diameter at breast height (DBH), canopy height, and canopy size (Getzin & Schöning 2012). According to the data collected at River Ridge Ranch, we decided to use vegetation indices, spectral signatures, and segmentation methods to classify trees species and determine canopy size, height, and canopy area for the entire ranch in order to calculate total carbon uptake at this site and build a methodology that can be used by future studies at similar environments. 

• Georeferencing: Collected images were georeferenced for geographic accuracy accordingly to a reference image for which GPS coordinates were verified by an Arrow Gold real-time kinematic GNSS receiver.

• RGB Maps: Multispectral images taken at 5 different bands (Red, Green, Blue, Near-Infrared, and RedEdge) are layer-stacked and re-arranged to create a true-color image using the visible spectrum bands (Red, Green, Blue).

• MULTISPECTRAL ANALYSES:
• Spectral Profiles display the pattern of wavelength reflected by an object. Multiple spectral profile charts were created for each tree species studied on River Ridge Ranch. The bands used to create the spectral profiles included red, green, blue, Red Edge, and near-infrared (NIR) spectral bands. The mean spectral signatures were determined to help discern differences between various tree species.

• Normalized Difference Vegetation Index (NDVI) is a way to measure healthy vegetation using red and near-infrared bands. Healthy vegetation would be shown with a high NDVI value, while low NDVI will have a smaller value representing lower vegetation. Pictorially this can be seen on the orthomosaic as brighter areas on the image for high NDVI values. With these brighter areas the pixel value can be determined from each individual tree canopy.

• Normalized Difference Red Edge (NDRE) uses longer wavelengths from 710-720 nm, which are closer to the near-infrared wavelengths and further away from the green wavelengths, which are shorter. Because the longer wavelength is being used for the Red Edge band, there will be a more defined and sharper distinction showing the amount of red light the chlorophyll absorbs. This would allow for better accuracy of pixel values for each distinct canopy.

• Using a combination of a digital surface model (DSM) and a digital terrain model (DTM), a canopy height model (CHM) can be determined. With this information, trees can be visualized, and tree species can be determined based on height.  

The spectral profile charts seen here represent the spectral signature of each tree species studied for the summer of 2021. These charts include the red, green, blue, Red Edge, and near-infrared bands. These charts are created from field collected data from block 1 of River Ridge Ranch. The top left chart displays the spectral signature of Buckeye, the bottom left displays Blue Oak, the top left displays Live Oak, and the bottom right display the median line from previously shown spectral profiles.

The Normalized Difference Vegetation Index is meant to determine the health of the vegetation. High values represent green density while low values may represent moisture-stressed vegetation.

Reference:  https://www.sciencedirect.com/topics/earth-and-planetary-sciences/normalized-difference-vegetation-index#:~:text=The%20NDVI%20is%20a%20dimensionless,Weier%20and%20Herring%2C%202000). 

In areas of low vegetation cover, the Soil Adjusted Vegetation Index (SAVI) can be used to correct Normalized Difference Vegetation Index (NDVI) for the influence of soil brightness.

SAVI was calculated using the "layer arithmetic" algorithm on eCognition software and used as a condition for tree classification. For the purpose of this study, an object was considered a tree if it had a height greater than or equal to 0.5 meters, and a SAVI greater than or equal to 0.2.

Normalized Difference Red Edge uses higher red wavelength bands to give a more distinct appearance between the tree canopies and the soil. The best time to use NDRE is mid to late growing season. The NDRE image was taken in June 2021 using eBee X Red Edge camera.

Reference:  https://eos.com/industries/agriculture/ndre/ 

Vegetation Indices are not always sufficient to determine whether an object is a tree or a shrub but the canopy height model can help distinguish the difference. One way of doing so is to drape the NDVI image over the canopy height model in ArcGIS pro for visualization. Another way to achieve this is retrieving pixel values from the canopy height model and overlay it with tree species polygons which should give a good enough sample to see a relation between tree and height species.

• This was done in ENVI software using supervised classification and majority/minority analysis.
• Regions of Interest was used for each individual Tree Species and other land covers.
• Canopy Height Model (CHM) was used at 3 meters for a mask.
• When the supervised classification was complete, the tree species were exported into ArcGIS.
• The classification was created by using data from the SODA 3D sensor imagery from May 2021.

• Using eCognition, we first ran a pixel level multi-resolution segmentation. Based on the CHM layer, we identified all objects that were above 0.5 meters next. Within the elevated objects, the ones with a SAVI greater than or equal to 0.2 were classified as trees.

• Mean spectral signatures, NDVI, and SAVI values were added as conditional parameters in order to classify individual species accordingly. Blue and Live Oaks demonstrated the greatest amount of overlap.

After the trees were classified and canopies obtained, an accuracy assessment was performed from field data taken at River Ridge Ranch. The accuracy assessment was based off the average height, perimeter, and area that was gathered from the field data. See tables below for more information

The objective of this study,  to calculate carbon sequestration at River Ridge Ranch, was achieved within a 10-month period between October 2021 and August 2022, while the data used had been previously collected within a 3-year period (2019-2021). After analyzing the UAV imagery, co-registering them to a georeferenced image, and running a variety of vegetation analyses, we were able to identify parameters that helped us identify and classify the different trees at the ranch with a total above-ground storage volume of 31,913,062 kg, without the contribution from Valley Oaks.

For this study, and based on previous studies already performed on this site by other CSULB students and staff, we decided to perform a multiresolution segmentation and classification in eCognition software. This allowed us to use height, vegetation analyses, and reflectance values (spectral signatures) as conditions for classifying trees and species of interest. Future studies may also consider other software and tools for performing a similar type of analysis. Within eCognition software, there are also many other algorithms that could have been explored for reaching similar results. These may include a multi-threshold segmentation in which different variables could have been selected for filtering out unwanted information (such as shape, texture, area, etc.), and the "multiresolution region grow" tool, depending on the data available.