Blog‎ > ‎

Challenges to Forest Landscape Modeling Under Climate Change: The Road Forward: Goals

posted Aug 4, 2014, 11:11 AM by Robert Scheller   [ updated Aug 4, 2014, 11:14 AM ]

How do we produce the next generation of forest simulation models?  Do we need to?  


Before we consider the means, we should carefully articulate our goals.  These are the goals as I see them:  


1) We need to improve model accuracy across all scales.  In the not-too-distant past, the goal was to avoid for excessively detailed models by emphasizing the ‘focal scale’: the scale for which a hypothesis or problem was formulated.  And, in general, this remains an overarching mantra among modelers of all stripes.  The increasingly non-stationary nature of the world has, however, thrown us a curve ball.  It is no longer satisfactory to assume that succession will be driven by generally known processes.  Early successional process in particular have become complex (in the mathematical sense): high sensitivity to initial conditions that generates large uncertainty.  Therefore we cannot ignore critical fine-scale processes:  seed rain density, competition for light, water, nutrients, seedling herbivory.  Doing so doesn't require that we consider how many stomates per leaf, the proverbial hobgoblin of forest modeling, rather that our models are sensitive to emerging data (see below) and questions.  Achieving accuracy across a wider range of scales (sensitive to local influences, applicable to broad landscapes) will require sophisticated computational approaches to estimate local conditions across broad areas.  An excellent example is iLand which uses simulation annealling methods to estimate local light availability across landscape scales.


2) We need an increased capacity to more quickly use more data sources.  Data assimilation (DA) is an ever-expanding challenge.  Bayesian DA approaches are becoming more common but are not necessarily congruent with the discrete time series data that are more characteristic of ‘Big Data’.  Another approach is exemplified by ongoing research within our lab that is creating a seamless conduit to re-scaled climate projection data (a new Climate Library for the LANDIS-II model).  In similar and numerous areas, more and more data are becoming available on-line in a standardized format.  Gone (or quickly fading at the least) are the days of struggling with netCDF files to extract a little bit of data.  But serious investments are necessary to avoid being trapped by our existing processes.  Often times, there is an initial investment of funding during initial model design and construction.  Acquiring ongoing funding to maintain and continue upgrading existing models is a huge challenge.  Many funding vehicles, e.g., NSF, do not generally support model development.  This is the reason that we launched The LANDIS-II Foundation:  to build a solid financial foundation for both maintenance and to allow the model to continue to grow: including building new connections to new data sources.


Next is the final entry in this series: Assuming that forest ecologists agree on the goals for future models (we don’t), how do we achieve them?  How do we maximize the power of our achievements to inform policy?


Rob Scheller


Comments