Modeling landscape forest productivity and stocking potential across the Pacific Northwest

Size: px
Start display at page:

Download "Modeling landscape forest productivity and stocking potential across the Pacific Northwest"

Transcription

1 Modeling landscape forest productivity and stocking potential across the Pacific Northwest INGY Winter Meeting 2018 Jarred Saralecos University of Montana, Graduate Research Assistant Advised by David Affleck and Zack Holden

2 Objectives Demand for site productivity estimates, landscape scale approaches Methods Data distribution, downscaled climate product, modeling approach Results Evaluating site index projections from multiple perspectives Next Steps Exploring management impacts of fine-scale productivity measures

3 Measuring site productivity Empirical: Site Index Based on Ht:Age relationships Time consuming, not always available Species specific Resolution based on sample density Process-based: Net Primary Productivity (NPP) gc/m 2 /yr Conversion of atmospheric carbon into organic compounds Mostly used at continental or global scales Output not in an actionable form for land managers

4 Geographic distribution

5 Climate data Monthly, gridded 250m climate data, Solar Radiation (W/m) Minimum temperature ( C) Maximum temperature ( C) Vapor Pressure Deficit Frost Days Precipitation (mm) Srad_April Fdays_April VPD_April

6 3-PG model formatting Process based model that uses soils, climate, and species data to grows biomass on monthly timestep 3-PG requires two sets of input parameters Site Latitude, Soil Class, StemNo., ASW Monthy Temp min, Temp max, Srad, Fdays, PPT, VPD Species Optimum temp. for growth, maximum canopy conductance, production lost to frost days, light use efficiency, growth allocation to roots, foliage, stem,

7 3-PG model formatting 6 model variations were run using 2 climate and 3 soil fertility measures Monthly climate averages (12 rasters) Continual monthly climate (360 rasters) Fertility Ratings (Soil productivity index; Schaetzl et al., 2012).5 fixed 50 year cumulative gross primary productivity (cgpp) calculated Soil class input was not used in final model

8 Comparing topography, climate, and productivity

9 Douglas-fir Site Index Models PSME Site Index 50 3-PG cgpp 50

10 Douglas-fir Site Index Models Compared Salas and Stage, 2007, RF_Topography, RF_Climate, GAM_NDVI, GAM_3-PG Random Forest Model Improvement Ratio

11 Douglas-fir Site Index Models Salas and Stage and NDVI models spatially constrained Addition of smaller region variables greatly improved accuracy NDVI model estimates frequently maxed Topographic position index and topographic wetness index not significant Model Inputs R 2 RMSE Salas and Stage (2007) Aspect(cos, sin), slope, elevation, ecoregion Topography_RF Aspect(cos, sin), slope, elevation, ecoregion Climate_RF PPTbal, GSRbal, avpd, VPDbal, GSPbal, MAR, MAT NDVI_GAM NDVI_Augustmax, andvi, ecoregion PG_GAM cgpp_50, (ecoregion)

12 Douglas-fir Site Index Models GAM_3-PG RF_Climate

13 Predicted Observed Aspect effect: Salas and Stage Sin(Aspect)

14 Aspect effect: 3-PG North Aspect North Aspect North Aspect North Aspect

15 Predicted Observed Stand density effects BAH explained addition 4.2% of variation in SI(3-PG) model Greatest influence occurred at <27 and >80 BAH TPH and QMD explained 2.2% and <1.0% Basal area per hectare

16

17 Model to model variability 3-PG vs Topography High Moderate Low

18 Study results Compared 5 increasingly mechanistic models for predicting PSME site index SI(3-PG) model outperformed topography and NDVI models Comparable results to random forest model using climate data Random forest is a black box Additional SI variation explained by BAH especially at very low and very high levels Model evaluation and prediction work continues

19 PIPO TSHE ALRU Site Index Models PIPO PSME TSHE

20 PIPO TSHE ALRU Site Index Models PIPO PSME TSHE Species R 2 RMSE PSME TSHE PIPO ALRU

21 Effects of site productivity on stocking density What effect does site quality (NPP, SI, LAI) have on size-density relationships? Can it predict maximum stocking capacity through determination of available resources? How sensitive is the maximum stocking capacity to mixed and pure stands? What trends occur along the gradients of species composition?

22 Stand data 25,597 stands PSME site index, QMD, TPH, %BA for PSME, TSHE, PIPO, ALRU Separate stands by SI into 4 equal groups Used quantile regression (q=.9) to assess the effect of site productivity on size:density relationships Selected stands with SI 22-28m Subset stands to include continuous gradient between two species Evaluated stand SDI along gradient Species Stands Basal area % (range) PSME-PIPO PSME-TSHE PSME-ALRU

23

24 Mixed-species effect on SDI PIPO PSME ALRU PSME TSHE PSME All stands: PSME SI 22-28m and >90% combined basal area of selected species

25 Study Results Increased site quality results in increased maximum stocking potential Site Index, Leaf Area Index, Net Primary Productivity Higher y-intercept Comparable slope Initial increases to basal area (PIPO and ALRU) results in higher SDI before decreasing steadily Transition to TSHE does not show significant decline at any point

26 Next steps Finalize PSME model comparisons and prediction across Washington Apply 3-PG outcome to ALRU, PIPO, TSHE How much does model tuning help? Predicted SI across habitat range for each Construct model for estimating SDI max from site index Evaluate PIPO and TSHE SDI max

27 Questions?