Dr. Mahmud, Kimbell School of Geosciences
Cave drip water response to surface meteorological conditions is complex due to the heterogeneous nature of water movement in the karst unsaturated zone. In this project, we aim to understand infiltration water hydrology in the Cretaceous karst formation of south-central Texas. The strata are characterized by Cretaceous limestone, marl, and shale, representing over 2000 ft of shelf and shelf margin depositions, and is comprised of shallow marine sedimentary rocks. Faults at this location are part of the Balcones Fault Zone and include horsts, grabens, anticlines, monoclines, and relay ramps. The faulting alignment was possibly influenced by lines of weakness in the Ouachita structural belt. This project will build on our current study of the Natural Bridge Caverns system and utilize the existing spatial survey of 20 automated cave drip loggers, installed in the two largest chambers (the hall of the mountain king and the castle of white giants) of this cavern. Stalagmite drip loggers were set up in approximate transects throughout the chambers from higher to lower elevations and have been monitored since May 2023. The sites were identified as actively dripping locations associated with different speleothem formations, including chandeliers, stalactites, soda straws, and flowstones. Data loggers were set to count drips continuously at 15-minute time intervals. We have collected the first 7 months of drip data in January 2024 and plan the next trip in May 2024 to obtain the remaining five months of data, completing the full hydrological year since the installation of the drip loggers. We hypothesize that drip logger time series are deemed similar if they are well correlated and only have a small offset with each other, and so these time series should cluster together in one group.
Initially, we plan to test the variability of drip discharge (i.e. coefficient of variance) with the sampling frequency to find the optimum sampling frequency that minimizes sampling artifacts while maximizing the capture of natural variability. The project will then use multidimensional scaling (MDS), which allows data dimensionality reduction by mapping complex multidimensional data on a low-dimensional manifold. First, we calculate a distance matrix that represents the similarity between all drip time series with the optimum sampling frequency. Specifically, a distance will be calculated between any two drip logger time series to characterize the similarity between those two loggers’ drip data. The technique operates on a distance used to measure the temporal similarity between two time series being compared. MDS will be used to translate these distances into a configuration of points defined in an n-dimensional Euclidean space. It results in a set of points arranged so that their corresponding Euclidean distances indicate the dissimilarities of the time series. Finally, the k means clustering algorithm will be used to divide these points into k clusters, which corresponds to a flow classification of the drip data time series. k means clustering is a method of vector quantization that is popular for cluster analysis in data mining. k means clustering aims to partition n observations into k clusters in which each observation belongs to the cluster with the nearest mean, serving as a prototype of the cluster.
We plan to analyze drip time series for an entire year from multiple sites taking advantage of ensemble of drip loggers to extract common properties by clustering. We believe considering multiple simultaneous drip time series, one can make better inferences about infiltration water flow and unsaturated karst zone properties. Overall, the findings of this work will provide a better understanding of processes that control water flow and transport processes within karst formation.