Country to global prediction of soil organic carbon and soil moisture using digital soil mapping
Date
2020
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
The largest carbon pool in terrestrial ecosystems is contained in soils and it plays a key role regulating hydrological processes, such as the spatial variability of soil moisture dynamics. Specifically, soil moisture and soil organic carbon are variables directly linked to ecosystem services such as food production and water storage. However, there are important knowledge gaps in the spatial representation (e.g., maps) of soil moisture and soil organic information from the country specific to the global scales. There is a pressing need to update the spatial detail of soil moisture estimates and the accuracy of digital soil carbon maps for improved land management, improved Earth system modeling and improved strategies (i.e., public policy) to combat land degradation. From the country specific to the global scale, the overreaching goal of this PhD research is to develop a reproducible digital soil mapping framework to increase the statistical accuracy of spatially continuous information on soil moisture and soil organic carbon across different scales of data availability (e.g., country-specific, regional, global). Chapter 1 provides a general introduction. Chapters 2 and 3 are focused on up-scaling soil organic carbon from the country-specific scale to the continental scale. Chapter 2 provides a country-specific and multi-modeling approach for soil organic carbon mapping across Latin America, where I identify key predictors and conclude that there is no best modeling method in a quantifiable basis across all the analyzed countries. In Chapter 3, I compare and test different methods and combinations of prediction factors to model the variability of soil organic carbon across Mexico and conterminous United States (CONUS). I describe soil organic carbon stocks across different land covers across the region, quantify the model uncertainty and discuss estimates derived from previous studies. Chapters 4 and 5 are devoted to improving the statistical detail and accuracy of satellite soil moisture from the country to the global scale. Chapter 4 describes how the machine learning fusion of satellite soil moisture with Geomorphometry increase the statistical accuracy and spatial detail of current soil moisture estimates across CONUS. Chapter 5 extends the previous chapter to the global scale and identifies global soil moisture trends. I provide a novel (gap-free) soil moisture global estimate that could be potentially used to predict the global feedback between primary productivity and long-term soil moisture trends. Chapters 4 and 5 reveal evidence of soil moisture decline across large areas of the world. Finally, chapter 6 summarizes the main findings of this research, the key conclusions, emergent challenges and future steps. The results of this research were useful to generate benchmarks against which to assess the impact of climate and land cover changes on soil organic carbon stocks and soil moisture trends. This research provides a framework (including high quality data and novel methodologies) to generate environmentally relevant science that can be used for the formulation of public policy around soil and water conservation efforts.
Description
Keywords
Geo-Statistics, Machine Learning, Pedometrics, Soil Data Science, Soil Moisture, Soil Organic Carbon