H101-09
Watershed Functional Zonation: Linking Watershed Patterns, Processes and Functions through Advanced Characterization Technologies
Thursday, 10 December 2020: 18:02
Virtual
Haruko M Wainwright1, Nicola Falco1, Baptiste Dafflon1, Sebastian Uhlemann2, Qina Yan3, Michelle E Newcomer4, Maya Franklin1, Bhavna Arora5, Nicholas Bouskill6, Kenneth Hurst Williams6, James Bentley Brown1 and Susan S. Hubbard1, (1)Lawrence Berkeley National Laboratory, Berkeley, CA, United States, (2)British Geological Survey Keyworth, Nottinghamshire, United Kingdom, (3)University of Illinois at Urbana Champaign, Urbana, IL, United States, (4)University of California Berkeley, Berkeley, CA, United States, (5)Lawrence Berkeley National Laboratory, Energy Geosciences Division, Berkeley, CA, United States, (6)Earth and Environment Sciences Area, Lawrence Berkeley National Laboratory, Berkeley, CA, United States
Abstract:
Predictive understanding of watershed functions is often hindered by the heterogeneous and multiscale fabric of watersheds. Heterogeneity exists within each of the watershed compartments, including aboveground compartments (i.e., plant species distribution and plant dynamics, topography) and belowground compartments (i.e., soil and bedrock structures/properties). Such watershed patterns influence ecohydrological and biogeochemical processes, which in turn affect functions and create emerging patterns as feedback. Watershed functions may include diverse signatures, including hydrological (i.e., partition, storage), ecological (e.g., species adaptation, productivity) and geochemical (e.g., solute export) signatures. Recently, there have been significant advances in characterization technologies across bedrock-to-canopy compartments, including airborne LiDAR, hyperspectral and electrical magnetic surveys. These technologies provide a critical opportunity to explore the relationship between patterns, processes and functions.
In this study, we propose the concept of watershed zonation to integrate multiple spatial data layers and to capture the heterogenous fabric of watersheds in a tractable manner. The zonation approach is based on unsupervised machine learning, which is rapidly evolving and has a great potential for supporting scientific discovery or for understanding complex systems. We hypothesize that (1) a suite of above/belowground compartments are correlated with each other through ecohydrological-biogeochemical interactions, (2) we can identify watershed subsystems or zones that have unique distributions of bedrock-through-canopy properties, (3) we can find such zones to distribute hard-to-scale parameters at the watershed scale, and (4) the identified zones – associated with key bedrock-to-canopy variabilities – can capture the variability of key watershed functions, such as ecosystem responses to droughts and nutrient exports. We demonstrate this concept, using the remote sensing, geophysics and point measurements as well as streamflow data collected in the East River Watershed, Crested Butte, Colorado, USA, particularly focused on the nitrogen export.
