Variance, Uncertainty, & Data Products
Focused on three main topics
Disturbance
Variability arising from ecosystem and human causes
Interannual variability and anomalies
Variability arising from climate forcing
Uncertainty
Characterization and quantification of errors
Data
What are we missing
Disturbance
Disturbance not well represented in models
Define disturbance as natural component of ecosystems instead of stochastic events
Networks, and remote sensing do not capture longer time scale processes (e.g., processes with 30-50 year return periods)
Need flux tower chronosequences in much larger set ecosystem types
Character/severity of disturbance matters
Much effort in remote sensing towards developing data sets
Fire, insect outbreaks – lots of disagreement across products
Mismatch between what remote sensing can/is providing and what models need
To most effectively link carbon/water consequences of disturbances fluxnet/modeling community needs to define and communicate needs to remote sensing community
----
Interannual Variation and Anomalies
Interannual variability is hard to capture
Low variance hard to explain
Opportunity to evaluate capacity of models and remote sensing to capture dynamics by focusing on big events/anomalies
Low hanging fruit to help understand what we can and cannot detect and model
Interannual variability/Event-based analysis
Hydrologic anomaly in 2011; European heat wave 2003/2010; Amazonian droughts
Need better information/understanding of drivers, particularly lagged or cumulative effects
Remote sensing of phenological anomalies and changes in seasonality
Exploit information at site level from webcams (Andrew, Lisa)
----
Uncertainty (in models and data)
Characterization of uncertainty is difficult
Multiple sources of error propagate through model results
Structural error, calibration error, forcing error, representation error (sampling uncertainty) – Enting et al.
Errors in met drivers and remote sensing inputs used for large scale models need more attention
Gaps in sampling
Geography matters (but what can you do about it?)
How can we think about more effective sampling
Biogeographic stratification; clustering of remote sensing/ecoregions, etc
Stratification based on spatio-temporal variation in model outputs
Need community discussion of how to characterize and quantify sources and magnitudes of uncertainty; uncertainty products
----
Data Products
MODIS subsetting tool at ORNL DAAC provides model to apply other remote sensing data sets at fluxnet sites
One stop shopping is good (DAAC, NEX)
Preliminary wish-list includes
Field measured ecophysiological variables used by models (e.g., Vmax)
Medium spatial resolution (10-50 m) maps of model PFTs at fluxnet sites
Landsat archive (time series in support of for e.g., disturbance histories)
Hi-Res Data (<1m) for upscaling (Quickbird, etc)
LIDAR: DESDynI is on hold, but still opportunies from airborne
Hyperspectral (AVIRIS,Hyperion)
Need to lower barriers to access
Probably many others
Remotely sensed PAR