Welcome to CGD

expand   collapse

 

Changes in Ecosystems

Change in climate, land use and disturbance regimes are driving increasingly evident changes to ecosystems. The impacts of these three related forces are complex and non-linear and current state of the art coupled carbon-climate models show divergent results. Reducing this uncertainty requires improved parameterizations of key ecosystem processes and better estimates of pertinent rate constants. The regulation of many ecosystem physiological processes is, to first order, understood. However, model comparisons show that different simulations, broadly including similar understanding, diverge widely when examined in climate change scenarios. The subtle and nonlinear interactions amongst physiological and substrate level control of both photosynthesis and respiration are becoming clearer, and reveal that even small errors in simulated environmental regulation of fluxes can cause the net flux to be in error or even have the wrong sign. Developing and parameterizing models to capture these interactions and forecast them accurately has forced the development of a new family of modeling techniques, known in the literature as "data assimilation" and "model data fusion", drawing on techniques pioneered in control theory, statistics and weather forecasting.

Figure 1. Results from flux data assimilation: Observed (a) and modeled (b) NEE, and modeled GEE (c) and RE (d) for the years 1999-2004. By separating NEE into GEE and RE, we are better able to diagnose process controls of observed interannual variability. These analyses, using eddy covariance-observations and an assimilation model, show that warmer years trigger an earlier start to carbon uptake, but lower mid-summer uptake. Earlier spring results in fast snowmelt and hence less available water for the later season. A consequence is that longer growing seasons cause reduced carbon uptake.

Most of the extant papers in this genre have focused on understanding climate controls over ecosystem carbon fluxes, and their application to this problem is becoming relatively paralleling the increase in process level understanding. While climate controls over terrestrial carbon fluxes are critical, changes to land use and disturbance regimes are at least as important and may interact with climate in complex ways. Our group of NCAR and University scientists has pioneered developments in ecosystem data assimilation in studying land use effects on biogeochemistry and has begun to actively address the role of disturbance.

We are now integrating the relatively continuous, albeit nonlinear, effects of climate with the often abrupt or even discontinuous effects of land use and disturbance in an assimilation or model-data fusion approach. The multiscaled, multiprocess (biology, biophysics, wildfire and other dynamics, all responding to similar environmental drivers) nature of the earth system interactions poses a unique opportunity to interface novel and state-of-the art numerical and computational approaches within the terrestrial ecosystem domain.

Fig. 2. The NCAR C-130 used in airborne measurements over the Rockies.

The achievement of these objectives will result in the rapid improvements in the realism and predictability of terrestrial biosphere simulation system. While the underlying process-level science and numerical techniques associated with climate and ecosystem simulations are relatively mature, our capacity to efficiently apply these data assimilation and error analysis techniques to complex models and to applications where significant non-linear behaviors occur is currently not well developed. Students of the land surface and terrestrial carbon problem have known for decades that this family of problems requires coupling "fast" and "slow" processes from turbulence to soil carbon dynamics, seconds to millennia) and more recently the importance of abrupt events has become evident (fires, severe storms). The mathematical legacy from control theory and meteorological data assimilation does not emphasize such multiscaled processes. We are developing an assimilation system designed for the multiscaled problem.

Preliminary results computing fluxes from aircraft (over 100s of km) show interesting comparisons to local (1 km) flux observations, and suggest that process information from local observations can be combined with spatially integrated observations from aircraft or concentration observing networks.

The results suggest that seasonal cycle patterns can be observed using integrative regional patterns and interpreted using intensive local observations, combined within a consistent modeling framework. Preliminary results from this project show that carbon uptake can be estimated at multiple spatial scales even in mountainous landscapes, that soil moisture is perhaps the critical driver of variability in carbon fluxes, and that the principal sign of potential climate change in the Rockies, warmer springs and longer growing seasons, reduces carbon uptake by intensifying the mid-summer drought. These results are being used in the evaluation of process models and coupled models and should lead to an improved generation of simulations of the control of terrestrial carbon exchange by temperature, precipitation and soil moisture.

For additional information, visit Terrestrial Sciences (TSS).