Data2HSM User Manual

The Data2HSM User Manual provides a quick‑start guide, setup and execution steps, tool descriptions, input and output options, and key assumptions and limitations. Data2HSM is a tool that allows users to generate hydrostratigraphic models and aquifer surfaces from AEM data, lithology logs, and geophysical logs. Possible outputs include: + Groundwater model layering prototypes + Hydrogeologic conceptual models + 3D lithologic models + Surfaces such as aquifer/aquitard bases, geologic formations, the water table, base of freshwater, or top of bedrock Data2HSM consists of one interface with three main tools: + ###Gaussian Mixture Model (GMM) GMM converts AEM data into 2D and 3D lithologic or stratigraphic models. Users can also include existing stratigraphic or texture models to guide the tool. GMM uses unsupervised machine learning to assign data points to user‑defined clusters that represent either lithologic groups (e.g., sand, clay, gravel) or stratigraphic units. When configured for stratigraphy, the tool uses prior geologic knowledge and a coordinate transformation to help maintain geologic consistency. + ###Smart Interpretation (SI) Smart Interpretation helps users rapidly delineate hydrogeologic surfaces such as aquifer/aquitard bases, the water table, the base of freshwater, or the top of bedrock. Users provide a small set of interpretations—either from past work or from other datasets—and SI uses polynomial regression informed by AEM data to complete the surface. An interactive 3D viewer allows additional interpretations using AEM sections, boring logs, and geologic cross‑sections. Multiple surfaces can be interpreted in one session. + ###Geological Pseudolabel Deep Neural Network (GeoPDNN) GeoPDNN is a deep‑learning tool that generates continuous 3D geological models from interval‑based datasets such as boring logs or AEM soundings. It takes in location coordinates and interval data and outputs a 3D stratigraphic model and surfaces marking the base of each unit. ###Key References: This manual draws on foundational research supporting Data2HSM’s methods, including: + [GeoPDNN 1.0: A semi‑supervised deep learning framework for 3D geologic modeling using pseudo‑labels](https://doi.org/10.5281/zenodo.10604091 "GeoPDNN 1.0: A semi‑supervised deep learning framework for 3D geologic modeling using pseudo‑labels") (Guo & X, 2023). + [Smart Interpretation – supervised statistical models for automated geological interpretations](https://doi.org/10.1007/s10596-017-9621-8 "Smart Interpretation – supervised statistical models for automated geological interpretations")(Gulbrandsen et al., 2017). This file contains Appendix B: Data2HSM User Manual from the [BCTools User Manual](https://data.cnra.ca.gov/dataset/dwr-basin-characterization-bctools/resource/d9d4792a-6746-4ea0-8e8b-694bdb545fc7 "BCTools User Manual"). File updated on 3/27/2026. ###Questions and comments may be directed to Basin.Characterization@water.ca.gov.