- Effects of deforestation and climate change on carbon & water cycling in Amazonia
- Improving the predictive ability of terrestrial biosphere models
- Mechanistic models of animal movement
- Enhancing vegetation structure for terrestrial biosphere modeling using Lidar and Radar techniques
- Development of the AIRMOSS Level 4 Net Ecosystem Exchange (NEE) products using the ED2 terrestrial biosphere model
- HyspIRI: Linking Terrestrial Biosphere Models with Imaging Spectrometry Measurements of Ecosystem Composition, Structure, and Function
- Integrated assessment of land-use and hydrology for sustainable development of the Amazon under changing climate (SSP)
- Previous research projects
Improving the predictive ability of terrestrial biosphere models
Recent analysis of terrestrial carbon dynamics in the Northeastern US have shown that the Ecosystem Demography model version 2 (ED2) can be jointly constrained against eddy-flux measurements and forest-inventory measurements to yield improvements in the accuracy of short-term and longer-term ecosystem dynamics. Building on this work, we are now incorporating novel constraints, re-optimizing the model, quantifying the value added, and developing new short and long term predictions.
In the previous work (mentioned above) the ED2 model was initialized with the observed canopy structure in the footprint of the Harvard Forest flux tower, then fitted simultaneously to the 1995 and 1996 hourly, monthly and yearly CO2 and ET flux data, as well as the observed rates of deciduous and coniferous tree basal area growth and mortality in these years. Prior to the optimization, the model significantly underestimated the seasonal cycle of Net Ecosystem Productivity and significantly over-estimated rates of tree growth and mortality. After fitting, the model accurately captured the observed CO2 fluxes, ET fluxes, and canopy growth and mortality dynamics over timescales spanning hours to decades. Changes in parameters most responsible for the improved goodness-of-fit include an increased maximum photosynthetic rate of hardwoods, a marked increase in the rate of fine root turnover, and a decrease in the carbon allocation to fine roots in conifer species. All of the new parameter values also fell within a priori acceptable ranges.
Following optimization, we evaluated the ED2 model at a different site in the Northeastern US, Howland forest. As before, the model was initialized with the observed canopy composition in the tower footprint, but the model parameters were not re-optimized. Despite the markedly different forest composition between the Howland and Harvard Forest sites (conifer-dominated as opposed to mixed-hardwood) there was a substantial improvement in the model's predictions of the 5-year CO2 flux record, as well as measured tree growth and mortality dynamics.
We are now building upon these results, specifically aiming to refine approximation to daytime respiration rates (in particular their partitioning from NEE), and to improve both model representation and basic scientific understanding of soil organic matter decomposition and autotroph-heterotroph-soil-efflux relationships. Towards this end we have increased the resolution and sophistication of ED's carbon accounting mechanisms, allowing us to track ecosystem pools and fluxes of the stable carbon isotope 13C. We will evaluate the potential for a similar accounting of the stable oxygen isotope, 18O as well, and these improvements will allow us to exploit recent advances in mass spectrometry and the additional data thus generated to yield further tests and constraints on the model’s predictive power. We will re-optimize the model, produce new predictions across the northeast, and quantify both the value of these isotopic data and any improvement in model fit. This will result in a more robust understanding of terrestrial carbon cycling, allow us to generate longer term predictions with more confidence and increased fidelity to the corpus of data on forests in the northeastern US, and guide the priorities and investigations of future research.