GENS 2431 - Data Analysis
GENS 2441 - Introduction to GIS
GENS 4951 - Special Topics: Advanced GIS
GENS 3401 - Research Methods in Environmental Science
GENS 3421 - Biogeography
Research
Interests
David is widely interested in conservation biogeography, but specifically how spatial information (e.g., GIS data, remote
sensing imagery) can be used to assist and inform conservation planning. His current research interests include: species distribution modelling, biodiversity monitoring and assessment,
simulation modelling, and GIS. David is particularly interested in questions pertaining to how to use and improve biodiversity monitoring to make statistical inferences and support
decision making, how to best determine the conservation value of different landscape units, and how to select and manage candidate areas as part of a larger reserve network.
Black Duck Joint Venture (BDJV)
2008 Competitive Grant, Black Duck Population Habitat Model for Maritime Canada', Department of
Geography and Environment, Mount Allison University, Sackville, NB, Canada.
Environment Canada, Environmental Damages Fund, 'ABBA: Atlantic Beached Bird Analysis',
Department of Geography and Environment, Mount Allison University, Sackville, NB, Canada.
Held jointly with Bird Studies Canada.
Refereed Publications
Lieske, D.J. and D.J. Bender (2011).
A robust test of spatial predictive models: geographic cross-validation. Journal of Environmental Informatics
17:91-101. [View abstract]
Lieske, D.J. and D.J. Bender (2009).
Accounting for the influence of geographic location and spatial autocorrelation in environmental models:
a comparative analysis using North American songbirds. Journal of Environmental Informatics 13:12-32.
[View abstract]
Lieske, D.J. and D.J. Bender (2009).
Visualizing Species Distributions: the Role of Geostatistics and GIS in Understanding Large-Scale Spatial Variation in Breeding Birds. Chapter 19
In
Representing, Modeling, and Visualizing the Natural Environment, N. Mount, G. Harvey, P. Aplin, and G. Priestnall (eds). CRC Press.
Espie, R.H.M.,
P.C. James, L.W. Oliphant, I.G. Warkentin and D.J. Lieske. (2004). Influence of
nest-site and individual quality on breeding performance in Merlins Falco
columbarius. Ibis 146: 623-631
Lieske, D.J., I.G. Warkentin, L.W. Oliphant, P.C. James, and R.H.M. Espie. (2000). Effects of population density
on survival in Merlins (Falco columbarius). Auk 117: 184-193.
[View abstract][Download paper ]
Lieske, D.J., L.W. Oliphant, P.C. James, I.G. Warkentin, and R.H.M. Espie. (1997). Age of first breeding in
Merlins (Falco columbarius). Auk 114: 288-290. [Download paper ]
Manuscripts
in Review
Lieske, D.J. Seeing the forest and the trees: visualizing geographic and environmental
space using enhanced exploratory data analysis. Submitted to Cartographica.
Lieske, D.J., Mahoney, M., O'Hara, P., Wilhelm, S., Whittam, R., and Pelot, R. Profiling
small-scale oil discharges on Canada's east coast: the impact of Surveillance method and
preliminary spatial trends. Submitted to Marine Pollution Bulletin.
Lieske, D.J., Mahoney, M., O'Hara, P., Wilhelm, S., Whittam, R., and Pelot, R. Profiling
small-scale oil discharges on Canada's east coast: the impact of Surveillance method and
preliminary spatial trends. Submitted to Marine Pollution Bulletin.
Lieske, D.J. and Bender, D. The Role of model complexity, autocorrelation and spatial location in
determining the performance of predictive occurrence models: a test using geographic
cross-validation. Submitted to Journal of Environmental Informatics.
Manuscripts
in Preparation
D.J. Lieske. et al. A model-based atlas of Atlantic Canadian
seabirds-at-sea: mapping distribution for conservation priority-setting at the
regional level.
Theses
Ph.D. Dissertation: Systematic conservation at the regional scale: the role of species distribution models
in priority setting.
[View abstract]
Master's Thesis: Population dynamics of urban Merlins (Falco columbarius).
Lieske, D.J. and D.J. Bender. (2006). Visualising
species distributions: the role of geostatistics and GIS in
understanding large-scale spatial variation in breeding birds
[View abstract]
22 October, 2009: Atlantic Society of Fish and Wildlife Biologists, 46th Annual General Meeting, 'Mapping Key Breeding Habitat for the American Black Duck: the Interplay of Wetlands, Landscapes and Scale', Kouchibouguac National Park, New Brunswick.
17 October, 2009: Atlantic Division of the Canadian Association of Geographers (ACAG) 21st Annual Meeting, 'Local-to-landscape: the Influence of Scale on Habitat Selection by the American Black Duck', Dalhousie University, Halifax, Nova Scotia.
23 October, 2008: Mount Allison University Social Sciences Seminar, 'An evaluation of alternative methods for synoptic visualization of species distribution', Sackville, New Brunswick, Canada.
20 October, 2008: Atlantic Division of the Canadian Association of Geographers (ACAG), 20th Annual Meeting, 'Seabirds-at-sea: an evaluation of alternative methods for synoptic visualization of species distribution', Mount Allison University, Sackville, New Brunswick, Canada.
14 March, 2008: Mount Allison University Social Sciences Seminar, 'Visualising species distributions: the role of geostatistics and GIS in understanding large-scale spatial variation in breeding birds', Sackville, New Brunswick, Canada.
16 February, 2008: Spatial Knowledge and Information (SKI) Canada, 1st Meeting, 'Predictive Modeling and GIS for Conservation', Fernie, British Columbia, Canada.
8 February, 2008: BIOL 3811, 'Reserve selection and systematic conservation planning' (guest lecture), Department of Biology, Mount Allison University, Sackville, New Brunswick, Canada.
13 October, 2007: Atlantic Division of the Canadian Association of Geographers (CAG), 'Systematic conservation planning at the regional scale: the role of species distribution models in priority setting', Université de Moncton, Moncton, New Brunswick, Canada.
12 October, 2007: Atlantic Society of Fish and Wildlife Biologists, 44th Annual General Meeting, 'A survey of methods for modelling species occurrence, with an evaluation of their predictive power and generality',
Wolfville, Nova Scotia, Canada.
25 June, 2006: Society for Conservation Biology, 20th Annual Meeting (‘Conservation without Borders’),
‘A comparison of the predictive accuracy of spatially and non-spatially explicit species distribution models’, San Jose, California.
6 April, 2006: GIS Research UK (GISRUK) 2006, 14th Annual Meeting, ‘Visualising species distributions: the role of
geostatistics and GIS in understanding large-scale spatial variation in breeding birds’, University of Nottingham, UK.
18-22 September, 2004: The Wildlife Society, 11th Annual Meeting, ‘A comparison of resampling methods for
evaluation of resource selection functions’ (Poster), Calgary, AB, Canada.
18-20 March, 2004: Western Division Canadian Association of Geographers Annual General Meeting, ‘Biodiversity
Assessment Using Species Richness Hotspots: A Rapid Tool for Conservation Priority Setting?’ (Presentation), Medicine Hat, AB, Canada.
1997: Prairie Universities Biological Seminars (PUBS), ‘Estimation of adult survival in merlins (Falco columbarius)
and the effect of population density’ (Presentation), University of Lethbridge, Lethbridge, AB, Canada.
1996: Biology 458.3 (Ornithology), ‘Bird population dynamics’ (Presentation), University of Saskatchewan,
Saskatoon, SK, Canada.
13-17 August, 1996: Joint Meeting of the American Ornithologist’s Union and Raptor Research Foundation,
‘Density dependence and its effect on population growth and reproduction in Merlins’ (Presentation), Boise State University, Boise
Back to Top
Workshop Presentations
19 Mar., 2008: ‘Workshop on the R Statistical Package’, School for Resource and Environmental Studies (SRES),
Kenneth C. Rowe Management Building, Dalhousie University, Halifax, NS, Canada.
27-28 Sept., 2006: ‘Workshop on the R Statistical Package’, Geomatics and Landscape Ecology Lab (GLEL),
Carelton University, Ottawa, ON, Canada. http://www.ucalgary.ca/~djlieske/R-Carleton
Lieske, D.J. and D.J. Bender. 2011. A robust rest of spatial predictive models: geographic cross-validation.
Journal of Environmental Informatics, 17: 91-101.
Predictive modeling is an important tool for identifying areas for conservation prioritization. But the reliability of any model depends on how well its
predictions can be generalized beyond the area surveyed. Recent work points to the potential for enhancing predictive power
by incorporating such spatial processes as autocorrelation or the influence of location, so this study addressed two
questions: (1) what affect does model complexity, spatial autocorrelation and spatial location have on model accuracy?
(2) how generalizable are different methods when applied to new geographic test regions? On average, predictive power
declined 22.7% ± 2.7% SE when models were used to predict occurrences in "unsampled" geographic test regions. Overall
variability in performance depended on the method used. AUTO and GAM models tended to be amongst the least variable,
but results depended upon species. Our results suggest that models with complex functional relationships between the
response and predictor variables (such as GAMs fit with up to 5 knots) tended to either improve accuracy, or perform more
consistently across species, but not both at the same time. In general, it is very difficult to accurately extrapolate
model predictions into unsampled geographic areas. However, we found that habitat specialists such as the Sedge Wren were
consistently well predicted, regardless of method, and that autocorrelated regression (using a Gibbs sampler and simulation
of presence/absence) could be more reliably generalized for species showing strong social structure (e.g., patchiness).
GWR was especially sensitive to the plots used to train the model.
Lieske, D.J. and D.J. Bender. 2009. Accounting for the influence of geographic location and spatial autocorrelation
in environmental models: a comparative analysis using North American songbirds. Journal of Environmental Informatics, 13: 12-32.
Environmental models are a critical tool for identifying where organisms occur by estimating the
relationship among species occurrence and important environmental factors. To date, the overwhelming majority of
predictive occurrence models disregard both the impact of spatial autocorrelation (interaction between neighbouring
points) as well as the possibility that model relationships may vary depending on geographic location. To address
this gap, we measured their impact on five bird species observed during seven years of the North American Breeding
Bird Survey. We first built traditional occurrence models (of varying functional complexity) using logistic
regressions and generalized additive models (GAMs). We then compared model accuracy and goodness-of-fit to those
incorporating spatial autocorrelation (ALOG) and spatial dependence (via geographically weighted regression, GWR).
Environmental variables included aspects of land cover, climate, and topography. A residual analysis indicated that
spatial autocorrelation persisted within even the most complex traditional models. In contrast, not only did ALOG
models incorporate this effect (as indicated by a lack of residual autocorrelation), but also offered better predictive
power for some species (+0.118 in the case of the American Crow, relative to the best GAM model). From an
information-theoretic perspective, ALOG models were consistent improvements over traditional models. Adoption of GWR
models also improved predictive accuracy (ranging from +0.078 for the American Crow and +0.008 for the Purple Finch).
However, comparison of their evidence ratios with ALOG models indicated that ALOG models were generally superior.
While we were unable to determine why geographic location influenced species' responses to environmental conditions,
evidence from generalized estimating equations (GEEs) revealed significant within-route correlation (? = 0.54 ± 0.26 SE),
and implicated an observer effect. A combination of broad-scale and fine-scale factors were important for predicting
occurrence, but we demonstrate that the incorporation of spatial factors offers the potential
to measure the spatially explicit outcomes of intra-specific interactions, and regional differences in resource usage.
We recommend that these methods be considered, particularly when evidence points to spatially autocorrelated errors or
when there are a priori reasons to suspect geographic variability in resource selection.
Espie, R.H.M., P.C. James, L.W. Oliphant, I.G. Warkentin, and D.J. Lieske. 2004. Influence of nest-site and individual quality on breeding performance in Merlins Falco
columbarius. Ibis, 146: 623-631.
We examined the effects of nest-site quality and bird quality on breeding performance in male and female Merlins Falco columbarius from a long-term study in
Saskatoon, Saskatchewan. In addition, we tested whether nest-site quality was associated with survival, as well as lifetime reproductive success (LRS). For females, nest-site quality had
little influence on any of the measures of breeding performance or survival. Even so, when females switched nest-sites, they tended to move to better ones. Hatch date was repeatable for
the same females occupying different nest-sites but not for the same sites occupied by different females. Among males, birds surviving past each age category tended to occupy nest-sites
of higher quality, and LRS was positively correlated with nest-site quality. The relationship between nest-site quality and LRS was heavily influenced by the poorest nest-sites. When
males switched nest-sites, they too tended to move to ones of higher quality. In addition, chick hatch date was repeatable neither for the same males occupying different sites nor for the
same sites occupied by different males. As with most other raptors, male Merlins provide most of the food for the pair and their young during the breeding season, and differences in
nest-site quality may have affected the effort needed by males to secure food. Female Merlins, however, appear still to have considerable control over the timing of breeding.
Lieske, D.J., I.G. Warkentin, L.W. Oliphant, P.C. James, and R.H.M. Espie. (2000). Effects of population density on
survival in Merlins (Falco columbarius). Auk 117: 184-193.
Accurate estimation of survival probabilities is an
important component of population demographics, and it permits a test of the life-history prediction that densities influence population dynamics via suppression of survival rates. As
part of a long-term study of urban-nesting Merlins (Falco columbarius), we estimated survival rates and tested for the effects of density dependence based on capture histories from
1,354 individuals (43 males and 110 females caught for the first time as adult breeding birds, and 597 males and 604 females caught for the first time as locally produced nestlings).
Overall capture probabilities were 0.55 ± SD of 0.039 per year for adults, 0.10 ± 0.075 per year for juvenile males, and 0.58 ± 0.23
per year for juvenile females. Mean survival
rate of adults was 0.62 ± 0.11 per year and did not differ significantly between males and females. Overall juvenile survival rates were 0.23 ± 0.032 for males and 0.055 ± 0.012 for
females. Band returns suggest that the discrepancy in survival rates between juvenile males and females resulted from higher natal dispersal of females rather than from lower survival.
Survival of adults (but not juveniles) was negatively density dependent, suggesting that density-dependent declines in survival exerted a regulatory effect on population size.
We examined the effect of age on breeding performance in male and
female Merlins (Falco columbarius) from a natural population
using a long-term data set. In the analysis, we examined whether
differences in chick hatch date and brood size associated with
parents of different ages arose due to selection of superior
individuals (differential mortality hypothesis) or to changes within
individuals over time (inadequate experience hypothesis). In addition,
we examined the effect of longevity on production of recruits and
lifetime reproductive success (LRS). In both sexes, breeding performance
improved with age. In females, this was mainly the result of
disproportionate mortality of inferior breeders, with less evidence
to support performance changes within individuals. Among males,
changes in breeding performance with age were largely the result of
improvements within individuals early in their life (between age 1
and 2+). Production of recruits was not dependent on parental age at
the time of breeding for either sex. Recruit production and LRS were
both influenced by longevity, so that longer-lived birds produced
more offspring over their lifetimes and thereby had a greater probability
of producing recruits. The differences between the sexes in terms of
age-dependent breeding performance are likely a consequence of the differing
roles the two parents play in reproduction. Male Merlins provide most of
the food for the pair and their young during the breeding season, and
changes in hunting skill with age may account for individual improvements
in breeding performance.
Lieske, D.J. and D.J. Bender.
Lieske, D.J. and D.J. Bender. Submitted.
Comparative Impact of Spatial Autocorrelation and Location on the Accuracy and Performance of
Species Distribution Models. Ecological Modelling.
Statistical modelling is a critical tool for predicting species occurrence,
but decisions based on such tools are highly sensitive to the accuracy of
the models and the methods used to produce them. To date, the overwhelming
majority of distribution models disregard both the impact of spatial
autocorrelation (proximity to conspecifics) as well as the possibility that
model relationships may exhibit non-stationarity (depend on geographic
location). We measured the impact of autocorrelation and non-stationarity
using five bird species observed during 7 years of the North American
Breeding Bird Survey. We first built non-spatial occurrence models using
logistic regressions and generalized additive models (GAMs), involving land
cover, climatic and topographic variables. We then compared model
accuracy and goodness of fit for models incorporating spatially-lagged
autocorrelation and localized model estimates (via geographically-weighted
regression, GWR).
Environmental variables were intrinsically autocorrelated, with Moran's I
values in excess of 0.80 at the smallest spatial neighbourhoods. More
elaborate models, based on polynomial or GAM methods, reduced the amount
of autocorrelation in residual errors but were unable to eliminate it.
Consistent with the high levels of residual autocorrelation observed for
the American Crow, the autologistic (ALOG) model not only eliminated
autocorrelation, but substantially improved predictive accuracy (+0.118,
relative to the best GAM model). The remaining species showed a tendency
for only minor changes in predictive accuracy, although from an
information-theoretic perspective, ALOG models were universal improvements
over non-spatial models. Adoption of non-stationary GWR models also
improved predictive accuracy, ranging from +0.078 for the American Crow
and +0.008 for the Purple Finch. The GWR approach resulted in the highest
predictive accuracy for all species except the American Crow, but
a comparison of the evidence ratios of GWR and ALOG models indicated that
ALOG models were consistently favoured over GWR models. While all
GWR models exhibited significant levels of non-stationarity, the
mechanisms contributing to this could only be partially assessed. Based
on a generalized estimating equation (GEE) model, significant
within-route correlation occurred for the American Crow (ρ = 0.54 ± 0.26 SE),
implicating a broad-scale observer effect on the probability of observing
and recording the presence of this species.
A combination of broad-scale and fine-scale factors were important
for predicting occurrence, but we demonstrate that the incorporation
of spatial factors offers the potential to measure the spatially-explicit
outcomes of intra-specific interactions, and regional differences in
resource usage. We recommend that these methods be considered,
particularly when evidence points to spatially autocorrelated errors, or
there are a priori reasons to suspect geographic variability in resource
selection.
Abstracts for Theses
Lieske, D.J. 2007. Systematic Conservation at the Regional Scale: the
Role of Species Distribution Models in Priority Setting. Ph.D. Dissertation, University of Calgary, Calgary, Alberta, 207p.
Present day rates of species loss are of world-wide concern, with species distribution
modelling (SDM) being an important means to identify areas of high
conservation value and direct conservation efforts more efficiently. The
key purpose of this thesis was to implement a full cycle of species
distribution modelling to evaluate the advantages and limitations of
different modelling approaches for identifying candidate
areas for conservation. Key objectives included: (1) assessment of
broad- scale spatial pattern and prevalence of spatial autocorrelation
for a set of breeding birds, observed during the North American Breeding
Bird Survey; (2) examination of the potential for improving predictive
accuracy through the incorporation of autocorrelation and non-stationarity;
(3) evaluation of model predictive accuracy, bias, and generalisability;
(4) analysis of the sensitivity of automated reserve selection to modelling
method, reserve selection algorithm, and information quality.
Assessment of broad-scale pattern indicated that patchy abundance
distributions were common, and nearly universally autocorrelated
(24 of 27 species, or 89%). Modelling results demonstrated that
predictive accuracy was generally related to model complexity.
However, the response in goodness-of-fit was more complicated
and depended upon the species in question. The impact of spatial
autocorrelation depended on the species, with the American Crow
benefiting the most from the application of a spatial autologistic
approach. Accuracy assessment, based on random test points,
confirmed that autologistic models provided substantially higher
predictive power, as did the non-stationary models produced by
geographically weighted regression (GWR). From the perspective of
generalisability, the simplest models were the least vulnerable to
predictive bias and were also the most consistently accurate
across species. The GWR method was the most sensitive to the
geographic locations used to train the model. The reserve
selection analysis identified that the the combination
of a greedy selection algorithm with high quality information was
effective for maximising the habitat value of the reserve network.
The value of SDMs for conservation planning will be maximised if they are as
biologically plausible as possible. Autologistic regression and GWR, on a case-by-case
basis, can lead to more accurate selection of high value habitat, thereby improving
our ability to support long-term species persistence.
.jpg)