mmlCHICOTla, Source code for construction and implementation of multiple machine-learning models of water levels in the Chicot aquifer system, southwestern Louisiana

The mmlCHICOTla repository contains R, Mermaid, and Perl language source code that can be used for construction multiple methods of machine learning (MML) of water levels in the Chicot aquifer system, southwestern Louisiana. The source code is written in R (primary and extensive), Mermaid (workflow diagrams), and Perl (minor utility for simple implementation of parallel processing on certain operating systems for two workflows within the repository). Extensive Markdown documentation is provided in many README and similar files. The mmlCHICOTla software requires input files stemming from the companion software titled covCHICOTla presenting observed monthly aquifer water levels (response variable) and associated covariates (predictor variables). The MML generation workflow (the genMML/ subdirectory) creates, using R, large quantities of statistical output, intermediate diagnostics, extensive visualization of overall MML prediction performance and detailed visualization of MML results for selected groundwater wells within the study area. The mmlCHICOTla also has extensive demonstration of workflow capacity (1) to predict aquifer water-level time-series for specific locations on the 1-kilometer National Hydrogeologic Grid (NHG) and (2) to predict gridded surfaces of aquifer water-levels in time on the NHG for the the study area. These two capacities represent two independent implementation phases of mmlCHICOTla. The mmlCHICOTla software for the implementation phase again requires input files of covariates on the NHG that come from the covCHICOTla companion software. MML implementation workflow (the useMMLsu/ and useMMLts/ subdirectories) creates, using R, large quantities of statistical output, geospatial layers, and visualization of status and trends of aquifer water levels. In total, the source code is expansive and technically demanding with advanced statistical methods involved along with novice-level geospatial operations. Sophisticated understanding of the R language itself and numerous external open-source libraries also are needed. The repository organization is such that the greater workflow is designed to nearly run-out-of-the box so that interested readers can readily evaluate and otherwise test the veracity of the algorithms herein.