Modeling the Occurrence of Springs and Seeps Along the Blanco River, Texas Using Logistic Regression




Frodge, Jonathan Blaire

Journal Title

Journal ISSN

Volume Title



Springs and groundwater seepage are essential contributors to continued streamflow in times of little or no precipitation. They also support unique wetland plant communities. These facts make them vital to the ecological health of the stream system and knowing where springs and seeps are located is essential to protecting the ecological integrity of a watershed. The objective of this research is to quantify and understand the relationships between springs in the riparian corridor of the Blanco River and the proximate geomorphic and vegetation characteristics. This research evaluated a logistic regression to model the occurrences of these springs and seeps. The locations of these riparian springs and the wetlands they support were mapped m the field by visual observation of groundwater emerging at the surface and/or identification of wetland plant species. Geomorphic and vegetative variables were measured at both spring locations and locations where springs are known not to occur. The variables regressed against spring presence/absence are: bank slope, bankfull width, average bankfull depth, the width to depth ratio, 100-year and 500-year model floodplain widths, coincidence with underlying geologic faults, and values of vegetation reflectance derived from color infrared imagery using the Normalized Difference Vegetation Index (NDVI). These variables were selected based on hypothesized processes to explain their relevance in terms of their contribution to the predictive power of the regression model. Not all variables tested were known to have direct, causal links to the presence of these riparian springs, thus this research uses regression m an exploratory manner to gain insights into riparian landscape patterns and complex processes that may go undetected using conventional process-response based modeling because they operate across wide ranging spatial and temporal scales. Results show that steep slopes are significant predictors of spring presence, perhaps because slopes expose more geologic strata increasing the likelihood of springs emerging from a saturated layer. Increased hydraulic gradients due to such steep slopes also increase pore pressures m the substrate; encouraging emergence of springs or increased flows from a spring or seep. Vegetative reflectance values reveal that available moisture provides contrast between the spring and non-spring sites, but are not strong predictors m the multivariate model. Bankfull widths and depths proved to be poor predictors of springs suggesting that the channel cross section geometry is not directly linked to the presence of springs and seeps. Floodplain widths were tested as surrogates for valley geometry at these sites and show the observed association of springs and steep-walled valley segments (narrow floodplains). Geologic faults are known to serve as conduits for groundwater that facilitate springs, but data precluded accurate analysis of this as a variable in the regression. Two models were created using significant predictor variables one of which uses field data; the other uses digitally available data. Either can be used to predict emergence of springs and seeps in similar watersheds using: 1) bank slope as measured in the field; or 2) 100-year floodplain data and NDVI values aggregated at an appropriate level (m this study a 2 x 2 m neighborhood was used to aggregate 1 m NDVI pixels).



riparian areas, logisitic regression analysis


Frodge, J. B. (2006). Modeling the occurrence of springs and seeps along the Blanco River, Texas using logistic regression (Unpublished thesis). Texas State University-San Marcos, San Marcos, Texas.


Rights Holder

Rights License

Rights URI