STREAM FLOW ACROSS HYDROLOGIC BOUNDARIES: IMPLICATIONS FOR FLOOD FREQUENCY ANALYSIS IN KARST TERRAIN

The complexity of karst landscapes present unique challenges in hydrological analysis and watershed management. Karst terrain is characterized by solutional depressions which capture runoff and function as discrete recharge points to underlying highly transmissible aquifers. A high degree of interaction between river and groundwater can result in river water head exceeding groundwater head due to storms which causes transient storage to occur in connected aquifers. These characteristics impact process such as runoff, storage and infiltration which ultimately affect a streams response to storm events. This response is used in flood frequency analysis (FFA) to estimate flood quantiles used to inform water resource management. FFA requires fitting probability distributions to annual maximum flow (AMF) series for gaged sites or, where insufficient data is available, using regionalization techniques which require an estimate of regional stream flow statistics or physical basin characteristics. Most regional analyses typically don’t distinguish karst terrain as a disparate area, though these rivers display distinctive responses to storm events which may require these regions to be treated separately. In the Suwannee River Basin, river flow abruptly transitions from impermeable siliciclastics with a high degree of runoff (Region 1) to karst terrain with high aquifer matrix permeability, little to no runoff and large volumes of aquifer storage during floods (Region 2). We used L-Moment diagrams to select the best distribution fit for each region and compare that to best fit for the basin as a whole. Additionally, we tested the goodness of fit for each site using the probability plot correlation test. Results indicate the best fit distribution for stations on Region 1 was the Log Pearson Type III (LP3), but the best fit distribution for stations on Region 2 was the Generalized Extreme Value (GEV). Using the LP3 on Region 2, which is the recommended and most frequently used distribution, upper flood quantiles were > 10% higher than those estimated by the GEV. This could possibly reflect transient aquifer storage during high flow events, which we address further by analyzing historical flood flows in Region 2 and assessing the impact on AMF series and regionalization techniques.