Abstract

I evaluated the utility of landscape variables as predictors of elevations associated with the forest - alpine ecotone in the southeast corner of British Columbia.   Multiple regression showed that average summer temperature, latitude / longitude and snow depth ,all factored by aspect, were related to the elevation of woodland and parkland ecosystems.  While these variables correlated with elevation (p= <0.05), the correlations were stronger on north and southwest aspects.  This may have important implications for interpretation of high elevation ecosytem mapping in BC.





Introduction


The main objective of this study is to conduct a preliminary investigation into the relationship between available landscape-wide environmental variables and elevations associated with the forest-alpine ecotone.  Specifically, elevation points collected at the Engelmann Spruce - Subalpine Fir (ESSF) Woodland  to Parkland transition zone are related to available landscape-wide environmental variables.

 

The study area (formerly know as the Rocky Mtn. Forest district) includes the East Kootenay Mountains and the Rocky Mountain trench area from Golden south to the U.S. border (see Figure 1).  The  ESSF Parkland/Woodland boundary (PWB) begins where continous forest starts to thin out and form open clumps, islands or ribbons of heath and herb meadows. PWB elevations points were identified as areas where trees transition from more than 50% cover to less than 50% cover on large and continuous cool, neutral and warm aspects.

Terms:

  • ESSF woodland is described as open and contiuous, upper-elevation forest.
  • In the ESSF parkland, islands of trees occur together with krummholz, heath, herb meadows, and grasslands.  


See Figure 2 for a map of the high elevation BEC zones in the study area.


BEC Alpine and Subalpine Ecosystems of BC

   Click below for a link to the BEC Alpine project website


Climate and Mountains;

"Climate is the fundamental factor in establishing a natural environment, it sets the stage upon which all physical, chemical, and biological processes operate.  This becomes especially evident at the climatic margins of the earth, i.e., desert and tundra.  Under temperate conditions, the effects of climate are often muted and intermingled so that the relationships between stimuli and reaction are difficult to isolate, but under extreme conditions the relationship becomes more evident.  Extremes constitute the norm in many areas within high mountains; for this reason, a basic knowledge of climatic processes and characteristics is a prerequisite to an understanding of the mountain mileu."


- Andy Bach 
Department of  Geography
 Minnesota State University


Methods

316  PWB elevation points were identified from air photos (or ortho rectified imagery if air photos were not available for certain areas), of which 302 were used in this analysis due to data entry errors.  Elevation points were selected in areas that denoted  large patches of identifiable ESSF woodland to parkland transition zones. See Figure 3 (a thru c) for an example showing PWB points on an ortho photo.

Linear models for these PWB points based on landscape variables (see Table 1), were ranked using Akaike's information criterion (AIC) method (Burnham and Anderson 2003). For model selection, AIC uses the principle of parsimony and enforces the trade-off between underfitting and overfitting (Burnham and Anderson 2003). It is important to recognize that there is substantial uncertainty as to which model is the best for this dataset.  However, a group of a priori candidate models were defined and the model with the lowest AIC value was selected as the best model. AIC value is calculated with the following formula:

AIC = -2(maximized log-likelihood) + 2K

 

where K is the number of parameters and the maximized log-likelihood is the natural logarithm of the likelihood function for a particular model.

The snow depth data was transformed by taking the log of (snowdepth +2) and wind speed was transformed by adding two to all values.  These variables were transformed in order to homogenize the scales of the data and even out the variances.   All analysis was preformed with the program - R 2.3.1 (2006).

Data and Landscape Variable Summary

  • Latitude and Longitude (decimal degrees);
  • Digital Elevation Models (rasterized to 25m cell size) from which slope, aspect, transformed aspect (TA), slope adjusted transformed aspect (TASL), and elevation were calculated;
  • Raster climate data including average temperature (400m cell size), average precipitation (400m cell size), snow depth (950m cell size) and windspeed (4km cell size);
  • Geology Polygons (basic rock type combinations for region); and,
  • Air photos (1:15 000 and 1:20 000) and 2005 orthorecified imagery (0.5m cell size)



Results

Analysis of stepwise forward and backward AIC model selection suggests that stepwise chosen linear model 3 (AIC = 2389) best predicts elevation of the PWB (see Table. 1).

The method of AIC model selection indicates only one a priori candidate model however, because the second best model,  a priori model 3 (AIC = 2402), is  greater than 4 AIC points different from the best AIC model.  For another model to be considered substantially similar the AIC values must be within 2 AIC points (Burnham and Anderson 2003).  Model 7 has a stepwise chosen AIC value of 2406 and an a priori AIC value of 2409.  The model 7 scenario, therefore, performed well but still shows very little similarity to the model 3 scenario.  Adjusted R squared values generally correlate with the trends in AIC values with the best fitting value being 0.3597.  Generally the adjusted R squared values do not suggest that the selected elevation points fit particularily well to the predicted values for each model equation.  Nevertheless, some values are considered significant and in this modeling scenario temperature, snow depth, latitude, longitude and aspect seem to be the most important predictor variables.

Factor analysis of the AIC selected model 3 suggests that north and southwest aspects may show the most consistent elevational trends (see Table 2).


 


Discussion / Conclusion

Upper Elevation BEC Zone Mapping Issues in Southeastern BC

Higher elevation BEC subzones are typically mapped in bands interpreted from elevation and aspect rules (Eng and Meidinger 1999). Plant communities, however, are not directly affected by elevation, but rather by variables such as temperature and precipitation that covary with elevation and latitude (Lookingbill and Urban 2005). These variables are difficult to measure at large spatial scales because the forest to alpine ecotone is very heterogeneous due to variation in slope, aspect, topography, geology, solar radiation and past and present glacial events.

 

Establishing broad “elevation/aspect rules” for the PWB boundary in mountainous areas of the former Rocky Mountain District is a subjective task. Identifying where dense patches of trees in the upper ESSF start to thin out to less than 50% cover on air photos is also subjective due to the fuzzy (i.e. vaguely defined) nature of the PWB.  Nevertheless, there are many areas that show reasonably clear transitional boundaries (See Figure 4).
 

General Mapping issues affecting the establishment of the PWB  fall into two categories:

1) Those that make the PWB bands wider or narrower than normal in comparison to the expected elevation band for that climatic area; and, 

2) Local or regional anomalies, where areas with a high cover of trees or parkland meadow do not relate to consistent elevation/aspect rules.

 
Timoney (1995) suggests that edaphic conditions exert a stronger control over tree distribution on nutrient-poor, dry sandy soils in the forest-tundra ecotone than on nutrient-rich, moist soils.  Thus, on unfavorable arctic terrain the forest-tundra ecotone is wider, more patchy and located southward of forest-tundra on favorable terrain (Timoney 1995).  Moist, nutrient-rich terrains, in contrast, may allow a high continous cover of trees to extend northward until absolute climatic thresholds are reached.

 A similar case in alpine environments might occur where higher quality sites support trees up to a species’ climatic limits (mainly controlled by temperature and precipitation) and lower quality sites are controlled more strongly by edaphic processes.  The lower quality sites can exhibit wider transitional band widths and occur at lower elevations than the “normal” macroclimatic thresholds would predict. 

In this project it has been assumed that areas with “normal” conditions will occur within a certain mountainous area (rule polygons) and that they will be prominent enough to establish general rules for macroclimatic thresholds. This may, however, not always be the case due to the influence of prevalent and complex edaphic factors. 

 
Specific Issues:

Difficulties and uninterpretable components in this project include:

  • Areas where bedrock topography (and/or instability) control tree patch size and shape are difficult to assign percent cover to and may not be reflective of “normal” macroclimatic conditions ;
  • Areas with unknown edaphic controls (e.g. pH, nutrient levels) can also confound the establishment of “elevation/aspect rules”;
  • Areas with scattered patches of krummholz (high wind areas?) are not easy to locate on aerial photographs;
  • Areas with frequent or extensive recent fire history are difficult to classify via aerial photographs;
  • Areas in the shadows of mountains are uninterpretable;
  • Areas with blacked-out ortho coverage are difficult to map;
  • Areas where the terrain is extremely steep making the resulting parkland and/or woodland band very narrow, and nearly undistinguishable, challenge the woodland/parkland/alpine concept. The upper boundary of what looks like woodland can reach very high elevations in this case (e.g., >2340m); and,
  • Areas with environmental influences such as xeric moisture regime, coarse texture or high insolation (i.e. rocky south slopes) have low to sparse tree cover and therefore woodland/parkland transitions can be very broad or not readily discernable.


Conclusion

A preliminary investigation into the relationship between available landscape environmental variables and elevations associated with the PWB  reveals that elevation may be most closely related to average summer temperature + latitude + log(May snowdepth +2) + longitude, all factored by general aspect.  It must be noted that landscape variables used in this study are of varying resolutions (25m for aspect to 4km for windspeed) and only provide a cursory look at the myriad environmental variables that may influence the forest to alpine ecotone.  

The findings of this study suggest that north and southwest aspects may show more consistent elevational trends. 

Further investigation in to the relationship between  landscape variables and the PWB could involve plant  species information and more plant relevant landscape variables such as leaf area index, relative solar radiation and soil type, pH, moisture, etc.  As well, this study would have benifitted from a larger sample size and the use of field data collected at a more local scale for accuracy comparisons.


Acknowledgments

Special thanks to the BC Ministry of Forests and Range, the US National Oceanic & Atmospheric Association (NOAA), Wang et al. (2006) and Environment Canada's Wind Energy Atlas program for supplying data for this project.



Literature cited

Burnam, K.P. and D.R. Anderson 2003. Model Selection and Multimodel Inferance: A practical Information-theoretic Approach. Springer; 2nd Edition. 488 pgs.

Eng, M. and D., Miedinger (1999) A Method for Large-scale Biogeoclimatic Mapping in British Columbia (Version 1). Report for Research Branch, BC Ministry of Forests and Range.

Lookingbill, T. and D. Urban (2005) Gradient analysis, the next generation: towards more plant-relevant explanatory variables.  Can. J. For. Res. 35: 1744-1753.

Timoney, K. (1995) Tree and Tundra Cover Anomalies in the Subarctic Forest-Tundra of Northwestern Canada. Arctic. 48(1)13-21.

 
Wang, T., A. Hamann, D.L. Spittlehouse and S.N. Aitken.            2006. Development of scale free climate data                     for western Canada for use in resourse                             managment. Intl. J. Climatology 26: 383-397


Top

Template Modified From Six Shooter Media.       

  Scott T. Black, UBC 2006.                                                 UBC