climate change and biome shifts in alaska and western canada

25
Climate Change and Biome Shifts in Alaska and Western Canada Current Results and Modeling Options December 2010

Upload: owena

Post on 23-Feb-2016

46 views

Category:

Documents


0 download

DESCRIPTION

Climate Change and Biome Shifts in Alaska and Western Canada. Current Results and Modeling Options December 2010. Participants. Scenarios Network for Alaska Planning (SNAP), University of Alaska Fairbanks EWHALE lab, Institute of Arctic Biology, University of Alaska Fairbanks - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Climate Change and Biome Shifts  in Alaska and Western Canada

Climate Change and Biome Shifts in Alaska and Western Canada

Current Results and Modeling OptionsDecember 2010

Page 2: Climate Change and Biome Shifts  in Alaska and Western Canada

ParticipantsScenarios Network for Alaska Planning

(SNAP), University of Alaska FairbanksEWHALE lab, Institute of Arctic Biology,

University of Alaska FairbanksUS Fish and Wildlife ServiceThe Nature ConservancyDucks Unlimited CanadaGovernment of the Northwest TerritoriesGovernment of CanadaOther invited experts

Page 3: Climate Change and Biome Shifts  in Alaska and Western Canada

Goals of this meetingSummary of project backgroundExplanation of modeling methodsUpdate on progress thus farDiscussion and decisions from group:

◦ Confirm clustering inputs (24 data points)◦ Confirm resolution for clustering and re-

projection (CRU vs PRISM)◦ Select number of clusters (15-20)◦ Select land cover comparisons and methods◦ Choose future decades to model◦ Confirm emissions scenarios (A1B, A2, B1)◦ Other issues?

Page 4: Climate Change and Biome Shifts  in Alaska and Western Canada

Overview This project is intended to:

◦ a) develop climate and vegetation based biomes for Alaska, the Yukon and the Northwest Territories, and

◦ b) based on climate data, identify areas that are least likely to change and those that are most likely to change over the next 100 years.

This project builds on and makes use of work previously conducted by SNAP, EWHALE, USFWS, TNC, and other partners.

The completed analysis will be used by partners involved in protected areas, land use, and sustainable land use planning.

Page 5: Climate Change and Biome Shifts  in Alaska and Western Canada

Overall objectivesDevelop climate and vegetation based

biomes (based on cluster analysis) for AK, Yukon, NWT, and areas to the south that may represent future climatic conditions for AK,Yukon or NWT.

Model potential climate-induced biome shift.Based on model results, identify areas that

are least or most likely to change over the next 90 years.

Provide maps, data, and a written report summarizing, supporting, and displaying these findings.

Page 6: Climate Change and Biome Shifts  in Alaska and Western Canada

The Scenarios Network for Alaska and Arctic Planning (SNAP)

SNAP is a collaborative network of the University of Alaska, state, federal, and local agencies, NGOs, and industry partners.

Its mission is to provide timely access to scenarios of future conditions in Alaska for more effective planning by decision-makers, communities, and industry.

Page 7: Climate Change and Biome Shifts  in Alaska and Western Canada

SNAP uses data for 5 of 15 models that performed best for Alaska and northern latitudes

PRISM downscaled to 2 km resolution OR CRU downscaled to 10 minutes (18.4 km)

Monthly temp and precip from 1900 to 2100 (historical CRU + projected)

5 models x 3 emission scenarios Available as maps, graphs, charts, raw data On line, downloadable, in Google Earth, or in

printable formats No data yet:

◦ Extreme events◦ Snowpack◦ Coastal/Oceans

SNAP Projections:based on IPCC models

Page 8: Climate Change and Biome Shifts  in Alaska and Western Canada

Phase I: Alaska modelMapped shifts in potential biomes based on current climate envelopes for six Alaskan biomes and six Canadian Ecozones

http://geogratis.cgdi.gc.ca/geogratis/en/collection/detail.do?id=43618

Page 9: Climate Change and Biome Shifts  in Alaska and Western Canada

Phase I Results:Potential Change: Current - 2100(Noting that actual species shifts lag behind climate shifts)

Page 10: Climate Change and Biome Shifts  in Alaska and Western Canada

Improvements over Phase IExtend scope to northwestern CanadaUse all 12 months of data, not just 2Eliminate pre-defined biome/ecozone

categories in favor of model-defined groupings (clusters)◦Eliminates false line at US/Canada border◦Creates groups with greatest degree of intra-

group and inter-group dissimilarity◦Gets around the problem of imperfect mapping

of vegetation and ecosystem types◦Allows for comparison and/or validation against

existing maps of vegetation and ecosystems

Page 11: Climate Change and Biome Shifts  in Alaska and Western Canada

Sampling Extent

Page 12: Climate Change and Biome Shifts  in Alaska and Western Canada

Cluster analysis Cluster analysis is the assignment of a set

of observations into subsets so that observations in the same cluster are similar in some sense.

Clustering is a method of “unsupervised learning” (the model teaches itself)

Clustering is common for statistical data analysis used in many fields

The choice of which clusters to merge or split is determined by a linkage criterion, which is a function of the pairwise distances between observations.

Cutting the tree at a given height will give a clustering at a selected precision.

Page 13: Climate Change and Biome Shifts  in Alaska and Western Canada

Step 1: Create a Dissimilarity Matrix

Distance measure determines how the similarity of two elements is calculated.

Some elements may be close to one another according to one distance and farther away according to another.

In our modeling efforts, all 24 variables are given equal weight, and all distances are calculated in “24-dimensional space” using RandomForest

(similarity matrix proximity matrix, distance matrix)

Taxicab geometry versus Euclidean distance: The red, blue, and yellow lines have the same length in taxicab geometry for the same route. In Euclidean geometry, the green line has length 6×√2 ≈ 8.48, and is the unique shortest path.

Page 14: Climate Change and Biome Shifts  in Alaska and Western Canada

Methods: Partitioning Around Medoids (PAM)

The dissimilarity matrix describes pairwise distinction between objects.

The algorithm PAM computes representative objects, called medoids whose average dissimilarity to all the objects in the cluster is minimal

Each object of the data set is assigned to the nearest medoid.

PAM is more robust than the well-known kmeans algorithm, because it minimizes a sum of dissimilarities instead of a sum of squared Euclidean distances, thereby reducing the influence of outliers.

Page 15: Climate Change and Biome Shifts  in Alaska and Western Canada

Clustering limitationsPAM must compare every data point to

every other data point in order to create a dissimilarity matrix and create medoids

Adding additional data points affects processing requirements exponentially

Thus, in creating clusters, we were limited to approximately 20,000 data points

Total area is approximately 19 million square kilometers

This meant selecting one data point for every 20 km by 20 km

Page 16: Climate Change and Biome Shifts  in Alaska and Western Canada

Resolution limitationsData are not available at the same resolution

for the entire area◦ for Alaska, Yukon, and BC, SNAP uses 1961-1990

climatologies from PRISM, at 2 km, ◦ for all other regions of Canada SNAP uses

climatologies for the same time period from CRU, at 10 minutes lat/long (~18.4 km)

◦ In clustering these data, both the difference in scale and the difference in gridding algorithms led to artificial incongruities across boundaries.

◦ One solution to both resolution and clustering limitaitons is to cluster across the whole region using CRU data, which is available for the entire area.

Page 17: Climate Change and Biome Shifts  in Alaska and Western Canada

PRISM data Unlike other statistical methods in use today, PRISM was

written by a meteorologist specifically to address climate Moving-window regression of climate vs. elevation for

each grid cell Uses nearby station observations Spatial climate knowledge base weights stations in the

regression function by their physiographic similarity to the target grid cell

PRISM is well-suited to mountainous regions, because the effects of terrain on climate play a central role in the model's conceptual framework

The primary effect of orography on a given mountain slope is to cause precipitation to vary strongly with elevation.

The topographic facet is an important climatic unit and elevation is a primary driver of climate patterns

Page 18: Climate Change and Biome Shifts  in Alaska and Western Canada

CRU data The station climate statistics were interpolated using thin-

plate smoothing splines (ANUSPLIN) Trivariate thin-plate spline surfaces were fitted as functions

of latitude, longitude and elevation to the station data The inclusion of elevation as a co-predictor adds

considerable skill to the interpolation, enabling topographic controls on climate

Local topographic effects such as rain shadows cannot be resolved unless: (1) a predictor that is a proxy for this influence is incorporated in the interpolation, and/or (2) there are sufficient stations to capture this local dependency as a function of latitude, longitude and elevation.

In regions with sparse data, the station networks used to create these data sets are clearly unable to capture this sort of detail

Page 19: Climate Change and Biome Shifts  in Alaska and Western Canada

Re-projecting CRU clusters to PRISM

CRU is available for entire study area, and offers a good fit at a broader scale

PRISM offers a better fit at fine scales, with better accuracy re altitude

Best of both:◦Cluster results from CRU data were used to

train an RF classification model. ◦RF then classified the full PRISM datasets

(where available)according to these clusters

Page 20: Climate Change and Biome Shifts  in Alaska and Western Canada

Comparison of results using various methodsResults were derived from the

following train/down-model groups:◦Trained to 15km sample of 2km data and

down-modeled to the 2km resolution over AK, YT, BC.

◦Trained to 20km sample of 10min data down-modeled to the 2km resolution over AK, YT, BC.

◦Trained to 15km sample of 2km data and down-modeled to the 10min resolution over full study region.

Page 21: Climate Change and Biome Shifts  in Alaska and Western Canada

Comparison of results: 5 clusters

Trained to PRISM data

Trained to CRU, re-projected to PRISM data

Page 22: Climate Change and Biome Shifts  in Alaska and Western Canada

Comparison of results: 10 clusters

Trained to PRISM data

Trained to CRU, re-projected to PRISM data

Page 23: Climate Change and Biome Shifts  in Alaska and Western Canada

Comparison of results: 15 clusters

Trained to PRISM data

Trained to CRU, re-projected to PRISM data

Page 24: Climate Change and Biome Shifts  in Alaska and Western Canada

Comparison of results: 20 clusters

Trained to PRISM data

Trained to CRU, re-projected to PRISM data

Page 25: Climate Change and Biome Shifts  in Alaska and Western Canada