Assessing the occurrence, vulnerability, expansion and disturbance of forests of the Pacific Northwest in response to recent climate variation Nicholas Coops Canadian Research Chair in Remote Sensing Department of Forest Resource Management Forest Sciences Centre. 2424 Main Mall. University of British Columbia. And Richard H Waring Oregon State University

Context In the Pacific northwestern (PNW) region of North America, climatic conditions have significantly warmed since a predominantly cool phase of the Pacific North American circulation patterns between 1950-1975.

What are the implications of this shift in climate for the vulnerability of native tree species? In places where species are predicted to have been less well suited, is there evidence of increased disturbance as assessed by remote sensing ?

Hybrid Modeling Approach Use a physiological model (3-PG) to predict how photosynthesis and canopy leaf area on Douglas-fir respond to climate change (1950-1975) versus (1976-2006) Define competitive climatic niches of other tree species in reference to physiological thresholds expressed for Douglas-fir Use the MODIS Disturbance index to link predicted tree species stress to observed disturbance

Phase I: The 3PG Model Freely downloadable in Excel format (developed by Peter Sands) C++ version which reads spatial data files (raster coverages) also freely available Developed by Landsberg and Waring (1997)

3PG Landsberg and Waring 1997 GPP NPP Absorption Utilization Photosynthetically Active Radiation (PAR) GPP x Fractional Interception x Physiological modifiers (Avg Tree Age, ASW, VPD, temperature) x Frost modifier x Temperature modifier x Quantum Efficiency = ƒ (Nutrient modifier) x Ratio NPP to GPP Rainfall Available Soil Water Transpiration Evaporation Fraction of NPP Fraction of NPP to Stem Root matter Foliage Turnover NPP Standing Volume Stem x Foliage allometrics x Stem allometrics Absorption Photosynthesis Utilization Respiration The partitioning fraction depends on Tree Age, ASW, VDP and fertility Landsberg and Waring 1997

Use of Modifiers Like other models 3PG makes use of modifiers which are dimensionless values, varying between zero (shut down) and unity (optimum). They represent the degree to which photosynthesis is limited by (a) water, (b) atmospheric vapour pressure deficits (c) suboptimal temperatures, and (d) frost damage. 1: Full Growth 0: No Growth 7

Environmental constraints on photosynthesis for Douglas-fir vary seasonally in the Pacific NW, U.S.A. soil water evaporative demand suboptimal temperature Frost limitations Autumn Summer Winter Spring

The 3-PG Model has been well calibrated for Douglas-fir Waring RH, N McDowell 2002 Use of a physiological process model with forestry yield tables to set limits on annual carbon balances. Tree Physiology 22:179-188 Swenson JJ, RH Waring, W Fan, NC Coops 2005 Predicting site index with a physiologically based growth model across Oregon, USA. Canadian Journal of Forest Research 35:1697-1707 Coops NC, SB Coggins, WA Kurz 2007 Mapping the environmental limitations to growth of coastal Douglas-fir stands on Vancouver Island, British Columbia. Tree Physiology 27:805-815

Data for 3PG: CLIMATE WNA Easy to use interface. Suite of climate layers and climate change scenarios Can generate point files or GIS layers of climate http://www.genetics.forestry.ubc.ca/cfcg/ClimateWNA/ClimateWNA.html

Soil Data To limit the analysis to climatic effects, we set the available water holding capacity at 200 mm for a sandy loam soil throughout the region. We also assigned a constant soil fertility rank of 50% of maximum, which results in an even partitioning of growth above- and belowground

Species Databases For British Columbia, tree species were taken from polygons located in protected forested areas and from vegetation resource inventory plots Accuracy of plot coordinates is roughly ± 500 m. For the United States, tree species data were taken from USFS Forest Inventory and Analysis plots. FIA data are recorded on a permanent sampling grid established across the conterminous United States at a density of approximately one plot per 2,400 ha As actual FIA plot locations are confidential, we used the publically available coordinates which have similar spatial accuracy as the Canadian data (i.e., ± 500 m). The presence / absence data from both Canada and the US were combined into one database with a total of 22,771 plots.

Plot Locations

3-PG Predicted Modifiers

3-PG Predicted Modifiers

Phase II of the modelling Define competitive climatic niches of other tree species in reference to physiological modifers expressed for Douglas-fir

We applied the 3-PG model to predict stand growth / LAI @ 50 yr using 1950–75 average climate. The model was then stopped and at 22,771 plots, monthly climatically-restricting modifiers were extracted. Decision tree analysis applied to develop models. A confusion matrix was developed which provides an indication of the positive and negative predictive power was well as kappa statistic We also used existing USFS maps of species’ distributions (Little 1971, esp.cr.usgs.gov/data/atlas/little/)

Decision tree analysis for LPP [Values in reference to optimum for Douglas-fir] After Coops & Waring. 2010. Climatic Change.

Douglas-fir

Lodgepole Pine

Translating Distribution Models to Stress-Based Information Use the decision tree model developed for each species using the long term (1950-75) climate Apply the model each year to the monthly climate data from 1976 – 2006 Years which have climate which correspond to the long term climate will be well suited to the species and thus we assume no stress Years which have climate outside the modelled range for the species will be years the species is “stressed” We then “count” the number of stressed years

(1976 - 2006)

Predicted stressed (red) and improved areas (green) since 1950-75 period

Phase III: Comparison against MODIS Disturbance Index Disturbance index utilizes a simple ratio of Annual-maximum composite LST and Annual-maximum EVI (greenness) to identify inter-annual changes in surface energy partitioning.   The DI is calculated on a per-pixel basis, and in this example, it is calculated annually.    +/-1 standard deviations away from the DI defines the range of natural annual variability.   SD above or below this indicates significant change in surface energy partitioning.  Above can indicate ‘negative’ disturbance (e.g. wildfire). Scores below 1 SD indicates areas of recovery.

Indicator: Disturbance Mildrexter et al (2007)

Indicator: Disturbance The disturbance index for 2000, 2003, and 2006. Areas of grey are indicative no change, bright red indicates major disturbance in 2000, Green indicates disturbance in 2003 and blue in 2006.

Predicted versus observed disturbance in Pacific NW Ecoregions Satellite-observed Disturbance (MODIS) EPA defined ecoregions Predicted stressed area (1990-2000)

Correlation between relative number of tree species stressed within an ecoregion vs. observed disturbance =0.7

Conclusions Hybrid approach has promise.. Partial validation: predicts reasonable values of max LAI, recorded distributions of 15 conifers, Stressed areas in ecoregions that correlate with the fraction of forested areas observed as recently disturbance Models appear to be accurate based on plot data New project underway to test model predictions with regeneration surveys in regions identified as stressed and recently disturbed Results are published in J. App. Veg. Science, Ecological Modeling and Remote Sensing of Environment.

All model outputs are available at: http://www.pnwspecieschange.info/ Project partly funded by a NASA Biodiversity and Ecological Forecasting Grant Collaborators: Steve Running David Mildrexler Maosheng Zhao University of Montana Clayton Beier Raphael Roy-Jauvin Tongli Wang University of British Columbia