Extreme flooding in southeast Texas in light of paleoflood and hydroclimate data

Abstract

Flooding in and around Houston, Texas that followed Hurricane Harvey in 2017 caused $125 billion in damages and serves as an example of the vulnerability of communities in the United States to extreme flood events. The extreme Hurricane Harvey floods raise the question of whether they were amplified by anthropogenic climate change and urbanization, and if extreme floods will occur more often in the near future. Studying these questions, however, is difficult due to short instrumental discharge records that may not capture the most extreme and infrequently occurring flood events. This dissertation presents research that aims to improve the statistical analysis of flood hazard in southeast Texas by incorporating paleoflood data and hydroclimatic data. First, I examine the use of paleoflood archives in alluvial settings to improve flood hazard assessments in southeast Texas. Floodplain lakes form depositional environments where sediments entrained by rivers accumulate during overbank floods. Laminated sequences of sediment stored in these lakes are used as paleoflood archives in which coarse-grained sediments represent major flood events. Hydraulic models, such as HEC-RAS, can simulate flow velocity patterns through floodplain lakes that match observed flood sediment depositions and support theoretical models of lake infilling. The use of hydraulic modeling in these alluvial settings provides an avenue to improve paleoflood magnitude estimates that can extend the length of instrumental records and increase the accuracy and precision of flood frequency analyses when instrumental records are short. However, when uncertainties of the paleoflood magnitude are large (>20%), or when instrumental records are long (>60 years), the integration of paleoflood events into the flood frequency analysis can be detrimental. Next, I evaluate how flood frequency analyses can be improved by considering the hydroclimatic conditions at a stream gauge. Traditionally, flood frequency analyses derive extreme flood probabilities by fitting annual maxima discharge data to a parametric probability function a priori. However, the hydroclimatic conditions of a region affect flood generating mechanisms (e.g., precipitation type, precipitation seasonality, land cover) and thus also influence the flood frequency distribution of a river reach. Here I examine stream gage records from across the United States to show that annual maxima from continental climates (Köppen group D) are best described by a General Extreme Value distribution, whereas records from arid regions (Köppen group B) are best described by a Log-Normal 3 distribution. Finally, I examine large-scale hydroclimatic patterns using climate reanalysis and gridded observations that mediate flood occurrence in southeast Texas. These analyses show that a western extended North Atlantic Subtropical High (NASH) occur in the days and weeks preceding the largest spring floods in southeast Texas, and that high antecedent soil moisture in this region mediated by the El Niño-Southern Oscillation serves as a seasonal precursor to floods in this region. --Author's abstract