Rivers in karst (caves, limestone, sinkholes) areas often lose water to sinkholes or conduits in the bed of the river. Near Valdosta, Georgia, for example, the Withlacoochee River loses as much as 100 cfs to a sinkhole area. Consequently, karst hydrologic modeling is difficult for several reasons:
The watershed cannot be determined by examination of a topographic map as non-karst areas can The area of the watershed can change as a function of water level because conduits at different levels act as relief valves Many of the features of interest lie underground and are inaccessible or accessible only with great difficulty The discharge point or points of a drop of water falling in a karst watershed can only be assigned a probability Discharge points in karst can change periodically as plugging and unplugging of conduits occur with storms of different intensities Karst water tables are discontinuous
Excerpted from the final report:
The researchers' work "presents a model for calculating surface stream losses to karst aquifers. It assumes that the loss conduits are phreatic (submerged) passages. We showed that for laminar flow, a single conduit friction factor is equal to a function of the geometry of the cross section multiplied by 64/Re, the friction factor for a circular conduit. Calibration accounts for the difference between a circular conduit and a non-circular one."
We showed that a complex network of conduits taking water from a surface stream can be replaced by a single equivalent conduit as long as the network has only one input and one output. If this stipulation is untrue, a more complicated two-port network may be used.
We developed the CLG model that can be implemented as an operation in the NWSRFS. We developed the CLG model in Visual Basic that can easily be converted to Fortran or any other language.
Observations of plugging in Saturday Cave point to the importance of debris to the morphology of a karst aquifer. We also developed a model of flow in Saturday Cave that showed good agreement with observations.
The Withlacoochee data were not adequate for proper testing of the current model. Other data sets may be used to calibrate and verify the methods developed here. Additional work is needed to calibrate and verify the method with an adequate data set."