Measuring the Carbon Cycle to Navigate Hydraulic Uncertainty

Abstract:

Imperata cylindrica, or cogongrass, is a grass that is invading the Southeastern United States, with a range reaching from Texas to Virginia. Observational studies and experimental manipulations have established that cogongrass reduces the biodiversity of pine forests and reduces the survival of pine seedlings. However, most potential mechanisms of invasion remain untested and we have no estimates of how cogongrass mediates the carbon and water cycles in mature pine stands. In addition, much of the work informing future cogongrass behavior is based on bioclimatic envelope modeling - a technique that excludes vegetation competition.

However, models with a mechanistic representation of vegetation competition (Dynamic Vegetation Model -DVM’s), have hydraulic schemes with assumptions that do not capture grass hydraulics. Cogongrass hydraulics specifically is rarely studied and has many structural unknowns, making it difficult to modify existing schemes. One possible solution is to indirectly constrain hydraulic parameters using carbon-cycle data. However, it’s unclear how much can actually be inferred from carbon-cycle data and what techniques capture that inference.

We used three different constraint techniques: a Bayesian meta-analysis of the literature, targeted trait measurements after a sensitivity and uncertainty analysis, and single-sample parameter data assimilation (PDA). PDA was done by simulating biomass-levels at a precipitation exclusion experiment. We then measured the reduced uncertainty associated with each technique. We found that PDA offered the most uncertainty reduction alone, but worked best with all three forms of constraint.