COS 89-2 - Accounting for disturbance history in watershed-scale models: Using remote sensing to constrain carbon and nitrogen pool spin-up

Wednesday, August 9, 2017: 8:20 AM
B113, Oregon Convention Center
Erin J. Hanan1, Christina L. Tague2, Janet S. Choate2, Mingliang Liu1, Crystal Kolden3 and Jennifer C. Adam1, (1)Civil and Environmental Engineering, Washington State University, Pullman, WA, (2)Bren School of Environmental Science and Management, University of Calfornia, Santa Barbara, Santa Barbara, CA, (3)Department of Forest, Rangeland, and Fire Sciences, University of Idaho, Moscow, ID
Background/Question/Methods

Evaluating future carbon (C) balance in disturbance-prone systems requires understanding the underlying mechanisms that drive ecosystem processes over multiple scales of space and time. Simulation modeling is a powerful tool for bridging these scales, however, model projections are limited by large uncertainties in the initial state of vegetation C and nitrogen (N) stores. Watershed models typically use one of two methods to initialize these stores. Spin-up involves running the model until vegetation reach steady state based on climate. The steady state approach however assumes the vegetation across the entire watershed has reached maturity and is the same age. Alternatively, remote sensing of a single vegetation parameter (typically Leaf Area Index; LAI) can be combined with allometric relationships to allocate C and N to various stores in a watershed model, accounting for variation in stand ages. However, allometric relationships are species and region specific and do not account for the effects of resource variation on C allocation. To address this problem, we developed a new approach for initializing C and N stores using the watershed-scale model RHESSys. The new approach merges the mechanistic stability of spin-up with the spatial fidelity of direct measurements. In this approach, one or more real vegetation parameters, derived from a spatial data layer are used to define spatially explicit targets for each patch across a watershed. The model then runs in a spin-up mode that tracks C and N stores for each patch separately until it reaches its target. Once all patches have reached their targets, the model outputs a set of state variables that can be used to initialize subsequent simulations. Unlike traditional spin-up, this approach supports non-homogeneous stand ages. We tested our target-driven approach in a pine-dominated watershed in central Idaho and a chaparral-dominated watershed in southern California, both of which have experienced recurrent wildfire. We compared simulations using the three initialization approaches (spin-up to steady-state, direct measurements coupled with allometric relationships, and the new target-driven approach).

Results/Conclusions

Model estimates of C, N, and water fluxes varied depending on which approach was used and the target-driven approach provided the best estimates of LAI and streamflow after ten years of simulation. This method shows promise for improving projections of biogeochemical and hydrologic dynamics in disturbance prone watersheds.