Disturbance is a major force regulating nutrient cycling, carbon retention, and hydrologic fluxes in terrestrial ecosystems. Evaluating future carbon balance in disturbance prone systems requires understanding the underlying mechanisms driving biogeochemical 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 distribution of carbon and nitrogen stores when ecosystems are not in steady state. Remote sensing has potential to improve projections of biogeochemical and hydrologic fluxes in non-steady state watersheds that are subject to regular disturbance. Methods for incorporating remote sensing into model simulations often rely on empirical, allometric relationships between a single vegetation parameter (typically leaf area index; LAI) and the various carbon and nitrogen stores being modeled. Allometric relationships are problematic however, because they do not account for resource variability, which can influence carbon allocation (to leaves, roots, etc.), and when used to initialize or update model carbon and nitrogen stores, can result in distributions that are unrealistic and/or unstable. We developed a new approach for incorporating remote sensing into carbon cycling models using the spatially distributed ecohydrologic model RHESSys. In this new approach, one or more vegetation parameters derived from remote sensing (e.g. LAI from LANDSAT or MODIS, stem wood from LiDAR, etc.) are used to define targets for the various patches across a watershed. The model then tracks carbon and nitrogen stores for each patch separately until the chosen parameters reach their target. Once all targets have been met, the carbon and nitrogen state variables can be used to initialize recovery. This approach allows carbon and nitrogen pools to develop mechanistically over time, accounting for the effects of resource and climate variability. Also, unlike traditional spin up, which assumes steady state conditions, the new approach uses remote sensing to spatially constrain pool sizes. Thus, the new approach supports non-homogeneous stand ages that are characteristic of disturbance prone systems.
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 used LANDSAT and MODIS data to update carbon stores following fire, and compared simulations using allometric relationships, with estimates using the new target-driven carbon and nitrogen allocation approach. Model estimates of carbon, nitrogen, and water fluxes varied depending on which approach was used and the target-driven approach provided the best LAI estimates after 10 years of simulation. This method shows promise for improving projections of carbon, nitrogen, and water fluxes in disturbance prone watersheds.