Back to search

Michigan State University

Investigating landscape-stream water quality relationships and stream water quality preservation strategies in the Texas Gulf Region using a hybrid of machine learning and hydrological modeling approach

Abstract

dc:description

This research investigates how land use, urban development pattern, topography, soil, climate, and population influence the stream nitrate (NO3--N), ammonium (NH4+-N), orthophosphate (PO43--P), total phosphate (TP), and Escherichia coli (E.coli) concentrations in the Texas Gulf Region. Specifically, the study focuses on how the land-stream water relationship varies by different sample sites, basins, ecoregions, and different years between 1991 and 2011. It also examines the benefits of compact urban development and verifies the management strategies to place best management practices (BMP) in hydrologically sensitive areas (HSAs). The 2011 cross-sectional study in the Texas Gulf Region indicates that the connectedness of developed areas and the adjacencies between developed areas and other land covers were more significant than the percentage of developed areas in their effect on stream water quality. The relationships between landscape factors and stream water quality varied by season, location, and pollutant category, with these associations generally stronger in dry seasons and in coastal suburban watersheds. Using a random forest machine learning algorithm, a predictive model demonstrated that high density aggregated urban development is the most effective in protecting stream water quality. The predicted average dry season NO3-N and TP concentrations were 0.17 mg/l and 0.09 mg/l in high density aggregated scenarios, compared to 1.2 mg/l and 0.28 mg/l in the current sprawled development scenario. The longitudinal study from 1991-2011 confirms the effects of controlling developed areas and agricultural areas in improving stream water quality. With the derived annual land cover composition and longitudinal nutrient and E.coli concentration data, it was found that adding 1 percent of developed area led to a 6.31% increase of NO3--N concentration and a 3.52% increase of PO43--P concentration in the Texas Gulf Region. Some unobserved characteristics led to high nutrient concentrations in the Middle Colorado-Concho and the Lower Trinity basins, and high E.coli concentration in the San Jacinto basin. The relationships between land cover and stream water quality varied more at the local scale than basin and region scales; they did not change significantly in the 20 years between 1991 and 2011. In the BMP siting strategy study, the effectiveness of placing BMP in HSAs was verified using a Soil & Water Assessment Tool (SWAT). The hydrological sensitivity of subbasins had a significantly nonlinear positive association with NO3-N concentrations. Defining HSAs as areas with the highest 2% hydrological sensitivity and designating them to be preserved as green space was the most effective in reducing NO3-N output. Generally, it was suggested that evidence-based ecological planning should incorporate performance evaluation with valid data-driven methods. Overall, this research was one of the first empirical studies to demonstrate the water quality degradation consequence of urban sprawl and the advantage of compact urban development. Machine learning and big data approaches were proven to be powerful tools for scenario prediction in land use planning to forecast environmental impacts of different urban development patterns. This study also established a robust Texas regional scale longitudinal water quality modeling approach depending upon efficient data fusion techniques, which can guide multiscale land use planning and watershed management.

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Wang, Runzi (Graduate of Michigan State University)
Contributors dc:contributor
  • Li, Ming-Han
  • Kim, Jun-Hyun
  • Wilson, Mark
  • Loveridge, Scott
  • Michigan State University

Subjects

dc:subject × 5

Rights

dc:rights
Statement dc:rights
  • In Copyright
Language dc:language
English

Identifiers

dc:identifier.*
Identifier
etd:48743
isbn:9798664734041
oclc:1201189990
umi:28085805
local:Wang_grad.msu_0128D_17773

Chain of custody

source
Harvested from
Michigan State University
Base URL
d.lib.msu.edu/oai
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Wang, Runzi (Graduate of Michigan State University). Investigating landscape-stream water quality relationships and stream water quality preservation strategies in the Texas Gulf Region using a hybrid of machine learning and hydrological modeling approach. 2020. https://doi.org/doi:10.25335/gn0t-1067