SSF PATTERNS and PROXIES
Project staff:
Prof. Dr. Markus Weiler
Dr. Ilja van Meerveld
Dr. Theresa Blume
Description:
The SSF Patterns and Proxies project aims to identify and characterize subsurface stormflow (SSF) response units across catchments by combining static landscape patterns, dynamic proxies, and an uncalibrated process-based modeling framework. SSF is a key hydrological process in headwater catchments, yet its spatial and temporal variability remains challenging to predict due to complex interactions between soil properties, topography, antecedent wetness, and rainfall dynamics. This project addresses the critical gap between detailed trench-scale observations and catchment-scale predictions by developing a data-driven, transferable approach to identify SSF-relevant landscape units and their response patterns. Using a three-pronged strategy, the project first identifies static structural response units through landscape similarity analysis and machine learning applied to GIS data and soil maps. Second, it detects dynamic response units by clustering groundwater and streamflow dynamics across distributed monitoring sites. Third, it uses these patterns to develop pragmatic proxies—measurable indicators that can substitute for direct SSF measurements—enabling SSF identification without trenches. The project leverages the unique, high-resolution dataset from Phase 1 of the DFG Research Unit, including over 600 groundwater wells, stream gauges, and trench flow records. These data are used to evaluate and refine the RoGeR model, a flexible, open-source hydrological model, to simulate SSF under different structural and process assumptions. The project tests alternative model structures for SSF initiation, connectivity, and preferential flow, comparing them against observed patterns. Finally, the approach is validated and transferred to other catchments worldwide, including Panola Mountain (USA), Krycklan (Sweden), and Alptal (Switzerland), to assess its generalizability. By integrating landscape evolution modeling, machine learning, and process-based simulation, Patterns and Proxies establishes a scalable framework for identifying SSF response units, advancing both experimental design and predictive hydrology.
Research Questions
1. Can process knowledge and machine learning be used to improve the mapping of relevant landscape characteristics, structural patterns, SSF-related properties and related model parameters? and what are the minimum data needs?
2. Can we detect patterns and similarities of groundwater table responses, SSF and discharge in time and space (collaboration with Transform) and can we use these patterns for model evaluation?
3. Are current model structures for SSF in distributed process-based models (e.g., RoGeR) already sufficient to predict the observed SSF patterns and responses (benchmark model) or can new mapping approaches of new SSF related model structures improve our ability to simulate SSF without intensive parameter calibration?
4. Are the approaches to map structural patterns, dynamic response mapping and the developed models transferable to other catchments?
Methods/ Approach
Patterns and Proxies employs a multi-method, integrative approach to identify SSF response units. Core methods include machine learning and landscape similarity analysis to derive SSF-relevant catchment properties (e.g., soil depth, bedrock permeability) from topography and geospatial data, using the OpenLEM landform evolution model. These are combined with cluster analysis and principal component analysis (PCA) to classify groundwater and streamflow responses into dynamic response units based on timing, lag, and intensity. Ensemble Rainfall-Runoff Analysis (ERRA) and recession curve analysis are used to extract SSF signatures from time series. Temperature and EC profile probes (from Connect) and geophysical data (from Subsurface Controls) are integrated to validate flow pathways. The RoGeR model is used as a flexible framework to simulate SSF at multiple spatial (2–10 m) and temporal scales, with benchmark simulations based on existing soil maps. Alternative model structures—testing different SSF initiation, connectivity, and preferential flow concepts—are evaluated against the benchmark using the derived signatures and proxies. A shared Postdoctoral Researcher ensures cross-project collaboration with Transform and Connect. The approach is tested in four German catchments and transferred to 10 international sites (e.g., USA, Sweden, Switzerland) using published datasets. Data are uploaded to the SSFDW, following FAIR principles. This integrated framework enables the identification of SSF response units and the development of low-data, high-impact proxies for catchment-scale hydrological assessment.