Tuesday, March 27, 2012

Uncertainty in precipitation inputs to ecosystems

Measuring the amount and chemistry of rainfall at a precipitation station is relatively straightforward.  However, estimating the input of rain water and solutes to ecosystems requires interpolation between the precipitation stations.  Various methods of interpolation are used in precipitation and atmospheric deposition studies (Garcia et al. 2008, Weathers et al. 2006), but the uncertainty in the interpolation is rarely reported or used in estimating uncertainty in deposition estimates.  

John Campbell conducted a preliminary analysis of spatial uncertainty in rainfall amounts using data from the Hubbard Brook Experimental Forest in New Hampshire. Hubbard Brook uses eleven precipitation gauges to estimate annual precipitation for six adjacent experimental watersheds. Thiessen polygons (Viessman and Lewis 1996) are used to define the area characterized by each of these eleven precipitation estimates, and precipitation to each watershed is calculated as the sum of the areas contributed by each polygon.  He compared this to several other interpolation methods (spline, inverse distance weighting, kriging, and regression modeling) and found differences of less than 1% across the methods for annual precipitation of the nine watersheds at Hubbard Brook (Figure 1).  The error associated with model selection is thus likely to be small.  The error within the models describing precipitation amounts (e.g. model or parameter error in the regression) has yet to be estimated.  Accounting for uncertainty in solute deposition is further complicated by the spatial and temporal mismatch between volume and chemistry samples, with fewer samples typically collected for solute chemistry than for rainfall volume.  Additional challenges to be addressed in estimates of atmospheric inputs are associated with the difficulty of monitoring dry deposition and cloud deposition and their interaction with vegetation structure.

Shannon LaDeau will develop a hierarchical regression model that can accommodate the spatial and temporal mismatches in precipitation volume and chemistry observations, using weekly data sampled from five watersheds at Hubbard Brook.  The regression will estimate monthly and annual wet deposition of solutes for each watershed and probability distributions for inference on predictive covariates. The model will also partition uncertainty due to measurement error, missing data, and poor model fit and will provide estimates of environmental stochasticity.  Methods for spatial interpolation of regressions will be explored and impacts on annual budgets compared, including spatial analyses packages in ArcView, R (e.g, ModelMap package) and Bayesian kriging methods in OpenBUGS (i.e., geoBugs). 

We also have support from the LTER Network Office for a Synthesis Working Group to further develop approaches to estimating uncertainty in precipitation fluxes.  We will extend our model protocol and results to similar data from other LTER sites.

Friday, March 23, 2012

Summer Course: Assimilating Long-Term Data into Ecosystem Models

Offered by: PaleoEcological Observatory Network (PalEON)

Dates: August 12-18, 2012

Course description: Estimating the impact of global change processes like land-use and climate on terrestrial ecosystems requires an integration of long-term data and ecosystem models. This course will provide 20 graduate students and postdocs with intensive training in the emerging tools that allow us to:

-estimate the signal and uncertainty in historical and paleoecological data
-assimilate both signal and uncertainty into the current suite of terrestrial ecosystem models

The course has a hands-on, integrated curriculum emphasizing the data/model process from design through data collection, analysis and back to design. We will collect tree-rings, historical survey data and sedimentary data (e.g., pollen, charcoal, and macrofossils). Analysis of these data will take place in a Bayesian mode of inference addressing uncertainty in age-models, calibration of proxy data, and integration of diverse historical data. After an introduction to inference from ecosystem models in traditional "forward" mode, participants will integrate ecological parameters estimated from their data sets into these ecosystem models using formal Bayesian data assimilation.

Participating faculty: Mike Dietze (University of Illinois); Steve Jackson (University of Wyoming); Jason McLachlan (University of Notre Dame); Chris Paciorek (University of California Berkeley); Jack Williams (University of Wisconsin)

Location: University of Notre Dame Environmental Research Center, Land O'Lakes, WI, USA.

Fees: This workshop is funded by a grant from the National Science Foundation and is free to participants. You must provide your own means of transportation to Chicago, Illinois, or Madison, Wisconsin.
There are a limited number of travel grants available to applicants from NEON, Inc., member institutions (see www.neoninc.org/content/paleon-data-assimilation-course).

Application: We are seeking students with interests and backgrounds in paleoecology, terrestrial ecosystem modeling, and/or statistics.  Send a CV, a statement detailing why you want to take the course and how you anticipate it helping your research, and arrange to have a letter sent from your major advisor supporting your application.

Apply to: Jason McLachlan at jmclachl@nd.edu

Deadline: March 30, 2012. Selections announced by April 15, 2012

Thursday, March 22, 2012

QUEST presentation at the Northeast Soil Monitoring Cooperative annual meeting

Carrie Rose Levine gave a presentation on QUEST and sources of uncertainty in soils at the annual meeting of the Northeast Soil Monitoring Cooperative on March 13, 2012.

Uncertainty in soils is difficult to quantify due to the natural spatial and temporal variability of soils.
Soil depth, rock content, bulk density, and nutrient concentration are all highly variable in soils. We may also expect high temporal variation of nutrient concentration and mass of the forest floor depending on changes in litterfall inputs over time.

Along with natural variation in soils, researchers have difficulty quantifying the uncertainty in soils as a result of what we call "knowledge uncertainty." In soils, knowledge uncertainty includes r
ejection criteria (obstructions), problems with excavation, representative subsampling, and uncertainty in analytical methods.

A link to the full presentation will be available shortly through the NESMC website:

Friday, March 16, 2012

LTER Working Group Funded: Uncertainty in Atmospheric Deposition

The estimation of wet atmospheric deposition requires the interpolation between point measurements of precipitation and solute chemisty across a landscape.  There are many approaches to interpolation, resulting in different estimates, usually reported without uncertainty.  QUEST has received funding from the LTER Network Office to conduct a Synthesis Working Group to quantify the uncertainties in these interpolations.
A workshop will be held May 21-22, 2012, at HJ Andrews Experimental Forest, involving representatives from four LTER sites: Hubbard Brook, Sevilleta, HJ Andrews, and Coweeta.  Sites were selected for their long-term precipitation records and multiple locations for precipitation solute chemistry.  On the first day, attendees will compare the various models for scaling up point measurements to the landscape, by applying all the models to all the sites.  The second day will be devoted to a more general discussion of uncertainty in deposition estimates, and will include a presentation by Chris Daly of the PRISM group.

Thursday, March 8, 2012

QUEST Workshop at LTER All Scientists Meeting

Members of QUEST held a two hour workshop entitled “Quantifying Uncertainty in Ecosystem Studies” at the 2012 LTER All Scientists Meeting.  The two hour discussion-based workshop started with a round of introductions in which the 40 participants identified their system of study, and the biggest source of uncertainty they face.  The rest of the time was devoted by ten “lightening round” presentations on uncertainty in different ecosystem components (see presenters below).  Each five minute talk was followed up by a group discussion.   We received positive feedback from many of the workshop participants, and a proposal has been submitted for an organized oral session at next years ESA conference in Minneapolis. Full meeting notes are available for download below

List of presenters:
John Campbell:  Intro to QUEST, Participant introductions (what's your biggest uncertainty?)
Harmon:  Intro to sources of uncertainty
Adam Skibbe: Precipitation Uncertainty
Xuesong Zhang: Precipitation Uncertainty
John Battles: Biomass Uncertainty
Mark Green: Streamflow Uncertainty
Ruth Yanai: Uncertainty in Soils
Jeff Taylor: NEON products and their uncertainty
Craig See: Uncertainty in Data Gaps
Carrie Rose Levine: Monitoring Uncertainty