Optimizing Remote Sensing Approaches for Dryland Carbon Flux Estimation
Abstract
Dryland ecosystems account for 40% of the global land surface area and play a vital role in the global carbon cycle. Gross Primary Productivity (GPP) is crucial for carbon exchange between ecosystems and the atmosphere, serving as a fundamental determinant of the carbon balance. However, precise modeling of GPP in terrestrial ecosystems, especially drylands, remains a complex challenge. This research utilizes a decade long spectral reflectance dataset acquired with a robotic tram system to estimate GPP using Random Forest model. The research exceled in capturing the complex spatio-temporal dynamics of the study site. Notably, the RF model exhibited best performance at extended temporal scales (Monthly vs weekly), boasting an R square of 0.69 and a RMSE of 0.13 g C m−2 in a 10-fold cross-validation. Variable importance highlights the likely pivotal role of pigment-focused indices. Furthermore, the research explores temporal and spatial changes as well as seasonal and interannual variations in different vegetation types. This dissertation also addresses the challenge of modeling soil organic carbon (SOC) dynamics, a critical factor influencing carbon-climate feedbacks. Machine learning models were employed to identify dominant predictors and their functional relations to SOC stocks using 54,000 SOC field observations and 46 geospatial environmental factors. The findings highlight the significance of diurnal temperature, drought index, cation exchange capacity and precipitation as influential observed predictors of SOC stocks. The RF prediction of global scale SOC stocks exhibit an R2 of 0.61 and a RMSE of 0.46 kg m-2. In contrast, precipitation, temperature, and net primary productivity (NPP) explained > 96% of Earth System Models (ESMs) of SOC stock variability, revealing different functional relationships between predictors and SOC stocks in observations and ESMs. These insights contribute to a more accurate understanding of carbon exchange processes and their impacts on climate feedback within the Earth System.
Subject Area
Remote sensing|Environmental engineering|Environmental science
Recommended Citation
Nyaupane, Kamal, "Optimizing Remote Sensing Approaches for Dryland Carbon Flux Estimation" (2023). ETD Collection for University of Texas, El Paso. AAI30819600.
https://scholarworks.utep.edu/dissertations/AAI30819600