Journal Articles


Kwon, T.J., Fu, L., Melles, S.J. (2016)
Computer-aided Civil and Infrastructure Engineering :DOI: 10.1111/mice.12222 

Read Abstract

This study presents an innovative approach to the planning of a critical highway sensor infrastructure – road weather information system (RWIS). The problem is formulated to minimize the spatially averaged kriging variance of hazardous road surface conditions while maximizing the coverage of accident-prone areas. This optimization framework takes explicit account of the value of information from an RWIS network, providing the potential to enhance the overall efficacy of winter maintenance operations and the safety of the travelers. Spatial simulated annealing is used to solve the resulting optimization problem and its performance is demonstrated using a real-world case study from Minnesota, United States. The case study illustrates the distinct features of the proposed model, assesses the effectiveness of the current location setting, and recommends additional stations locations. The findings of our study suggest that the proposed model could become a valuable decision-support tool for planning a new RWIS network and evaluating the performance of alternative RWIS expansion plans.

Melles, S.J., Chu, C., Alofs, K.,M., Jackson, D.A. (2015)
Landscape Ecology 30:919-935

Read Abstract

The Great Lakes currently harbour a number of non-native fishes that are thermally limited to the comparatively warm waters of Lake Erie and Lake Ontario. Climate change could facilitate the inland spread of many non-native species as the Great Lakes and their tributaries warm, putting thousands of inland lakes and streams at risk. We investigated how watershed network configurations, climate change and proposed hydro-power development could influence invasion risk in the Great Lakes Basin. Electric circuit theory was used to model hydrologic accessibility of aquatic ecological networks (i.e., lake, river, and impoundment chains) within tertiary watersheds. Risk of invasion was measured as the product of probability of non-native species spread (hydrologic accessibility) and amount of suitable thermal habitat under an ensemble of air temperature projections. Proposed hydro-power dam sites and their upstream catchments were used to evaluate changes in total risk of invasion given passable, semi-passable, and impassable dams. We show that projected climate change will lead to more coolwater stream and warmwater lake habitat. Overall invasion risk of cool- and warmwater species was highest in southern Ontario and surprisingly in northern watersheds draining into Lake Superior. This risk could be partially mediated by proposed dams if dams reduce connectivity and access to potentially suitable habitat. Our evaluation of mean invasion risk provides a broad-scale comparative tool for management of invasive species control options.

Jones, N.E., Schmidt, B., Melles, S.J. (2014)
Canadian Journal of Fisheries and Aquatic Sciences 71: 1616–1624.

Read Abstract

Extremes of flow and patterns of flow variability limit the distribution and abundance of riverine species via a natural disturbance regime. Using a habitat template approach, we describe the distribution and characteristics of natural flow regimes in Canada based on the severity of flows, flow predictability, and flow variability. Bayesian clustering was used to group 888 gauged watersheds across Canada into 10 classes. Some flow classes were found in all provinces, whereas others showed greater regional grouping related to land physiography (e.g., Canadian Shield and ecozones). Ontario and British Columbia had the greatest diversity of flow classes. Larger river systems tended towards less harsh flow regimes and greater flow regularity than small systems. A stream–lake network pattern, particularly the presence of lakes, decreased the severity of flow. The flow metric flood-free interval was found to be a potentially misleading indicator of reduced disturbance for high-latitude streams in Canada where ice formation and persistence are important stress factors for biota. Most flow stations had an 80% or higher chance of belonging to their primary membership class. Quantifying uncertainty in class assignment can help fellow scientists and resource managers appropriately apply our findings.

Melles, S.J., Jones, N.E., Schmidt, B. (2014)
Environmental Management 53: 549-566.

Read Abstract

Conservation and management of fresh flowing waters involves evaluating and managing effects of cumulative impacts on the aquatic environment from disturbances such as: land use change, point and nonpoint source pollution, the creation of dams and reservoirs, mining, and fishing. To assess effects of these changes on associated biotic communities it is necessary to monitor and report on the status of lotic ecosystems. A variety of stream classification methods are available to assist with these tasks, and such methods attempt to provide a systematic approach to modeling and understanding complex aquatic systems at various spatial and temporal scales. Of the vast number of approaches that exist, it is useful to group them into three main types. The first involves modeling longitudinal species turnover patterns within large drainage basins and relating these patterns to environmental predictors collected at reach and upstream catchment scales; the second uses regionalized hierarchical classification to create multi-scale, spatially homogenous aquatic ecoregions by grouping adjacent catchments together based on environmental similarities; and the third approach groups sites together on the basis of similarities in their environmental conditions both within and between catchments, independent of their geographic location. We review the literature with a focus on more recent classifications to examine the strengths and weaknesses of the different approaches. We identify gaps or problems with the current approaches, and we propose an eight-step heuristic process that may assist with development of more flexible and integrated aquatic classifications based on the current understanding, network thinking, and theoretical underpinnings.

Pfeifer, M., Lefebvre, V., Baeten, L., Banks-Leite, C., Betts, M., Brunet, J., Cerezo, A., Cisneros, L., Coomes, D., D’Cruze, N., Duguay, S., Eigenbrod, F., Hadley, A.,  Hanson, T.R., Hawes, J., Heartsill, T., Klingbeil, B., Kolb, A., Kormann, U., Lachat, T., Lantschner, V., Laurance, B., Lens, L., Lugo, A., Marsh, C., Medina-Rangel, G.F., Melles, S., Mezger, D., Owen, C., Phalan, B., Possingham, H., Raheem, D., Ribeiro, D.B., Robinson, D., Robinson, R., Rytwinski, T., Scherber, C., Slade, E., Somarriba, E., Stouffer, P., Struebig, M.J., Tylianakis, J., Tscharntke, T., Urbina Cardona, J.N., Vasconcelos, H., Wells, K., Willig, M., Wood, E., Young, R.P., Bradley, A., Ewers, R. (2014)
Ecology and Evolution manuscript ID: ECE-2013-12-0573

Read Abstract

Habitat fragmentation studies have produced complex results that are challenging to synthesize. Inconsistencies among studies may result from variation in the choice of landscape metrics and response variables, which is often compounded by a lack of key statistical or methodological information. Collating primary datasets on biodiversity responses to fragmentation in a consistent and flexible database permits simple data retrieval for subsequent analyses. We present a relational database that links such field data to taxonomic nomenclature, spatial and temporal plot attributes, and environmental characteristics. Field assessments include measurements of the response(s) (e.g., presence, abundance, ground cover) of one or more species linked to plots in fragments within a partially forested landscape. The database currently holds 9830 unique species recorded in plots of 58 unique landscapes in six of eight realms: mammals 315, birds 1286, herptiles 460, insects 4521, spiders 204, other arthropods 85, gastropods 70, annelids 8, platyhelminthes 4, Onychophora 2, vascular plants 2112, nonvascular plants and lichens 320, and fungi 449. Three landscapes were sampled as long-term time series (>10 years). Seven hundred and eleven species are found in two or more landscapes. Consolidating the substantial amount of primary data available on biodiversity responses to fragmentation in the context of land-use change and natural disturbances is an essential part of understanding the effects of increasing anthropogenic pressures on land. The consistent format of this database facilitates testing of generalizations concerning biologic responses to fragmentation across diverse systems and taxa. It also allows the re-examination of existing datasets with alternative landscape metrics and robust statistical methods, for example, helping to address pseudo-replication problems. The database can thus help researchers in producing broad syntheses of the effects of land use. The database is dynamic and inclusive, and contributions from individual and large-scale data-collection efforts are welcome.

Fortin, M.-J., James, P., MacKenzie, A. Melles, S.J., Rayfield, R. (2012)
Spatial statistics 1: 100-109

Read Abstract

A critical part of ecological studies is to quantify how landscape spatial heterogeneity affects species’ distributions. With advancements in remote sensing technology and GIS, we now live in a data-rich era allowing us to investigate species–environment relationships in heterogeneous landscapes at multiple spatial scales. However, the degree and type of spatial heterogeneity changes depending on the spatial scale at which species–environment relationships are analysed. Here we present the current spatial analytic methods used in ecological studies to quantify ecological spatial heterogeneity. To determine the key spatial scales at which underlying ecological processes act upon species, we recommend use of spectral decomposition techniques such as wavelet analysis or Moran’s eigenvector maps. Following this, a suite of spatial regression methods can be used to quantify the relative influence of environmental factors on species’ distributions. Finally, spatial graph metrics can be employed to quantify the effects of spatial heterogeneity on landscape connectivity across or within species’ ranges and can be used as additional predictors in spatial regression models. We emphasize how spatial statistics, spatial regression, and spatial graph theory can be used to provide insights into how landscape spatial complexity influences species distributions and to better understand species response to global change.

Melles, S.J., Jones, N., Schmidt, B. (2012)
Freshwater Biology 57: 415-434

Read Abstract

1. We review some of the classic literature on geomorphology and ecology of streams in an effort to examine how theoretical developments in these aquatic sciences have influenced the way fresh flowing waters are classified. Our aim was to provide a historical examination of conceptual developments related to fluvial classification, and to discuss implications for conservation planning and resource management.
2. Periods of conceptual influences can be separated into three overlapping phases each distinguished by theoretical, analytical or technological advances: (i) early Darwinian perspectives; (ii) the quantitative revolution; and, (iii) age of the computer, hierarchy and scale.
3. During the first phase, stream geomorphologists were largely influenced by Darwinian metaphors. The study of stream origin and change through time became more important than the study of stream systems themselves. The idea that streams progress deterministically through successive stages of development seemed to create a veil, most prevalently in North America, that barred analysis of the full scope of variability in these systems for over 50 years.
4. The quantitative revolution brought about many new ideas and developments, including the laws of stream numbers. This period focused on predictive and mechanistic explanations of stream processes, setting the stage for physically based stream classifications that assume that streams can be restored by engineering their physical characteristics.
5. In the most recent ‘age of the computer’, concepts from the fields of geographic information science and landscape ecology have been incorporated into stream ecology and aquatic classification. This has led to investigations in stream and aquatic ecosystems at hierarchical spatial scales and along different dimensions (upstream/downstream, riparian/floodplain, channel/ground water and through time). Yet, in contrast to terrestrial landscapes, flowing waters are not as easily classified into spatially nested hierarchical regions wherein upper levels can be subdivided into smaller and smaller regions at finer spatial scales. Riverscapes are perhaps best described as directionally nested hierarchies: aquatic elements further downstream cannot be rendered equivalently to elements upstream. Moreover, fully integrated aquatic ecosystem classifications that incorporate lake and river networks, wetlands, groundwater reservoirs and upland areas are exceedingly rare.
6. We reason that the way forward for classification of flowing waters is to account for the directionally nested nature of these networks and to encode flexibility into modern digital freshwater inventories and fluvial classification models.

Melles, S.J., Fortin, M.-J., Badzinski, D. and Lindsay, K. (2012)
Avian Conservation and Ecology 7(2): 3

Read Abstract

Theoretical and empirical studies suggest that well-connected networks of forest habitat facilitate animal movement and contribute to species’ persistence and thereby the maintenance of biodiversity. Many structural and functional connectivity metrics have been proposed, e.g., distance to nearest neighboring patch or graph-based measures, but the relative importance of these measures in contrast to nesting habitat at fine spatial scales is not well established. With graph-based measures of connectivity, Euclidean distances between forest patches can be directly related to the preferred gap crossing distances of a bird (functional connectivity). We determined the relative predictive power of nesting habitat, forest cover, and structural or functional connectivity measures in describing the breeding distribution of Hooded Warblers (Setophaga citrina) over two successive breeding seasons in a region highly fragmented by agriculture in southern Ontario. Logistic regression models of nesting occurrence patterns were compared using Akaike’s information criterion and relative effect sizes were compared using odds ratios. Our results provide support for the expectation that nest-site characteristics are indeed related to the breeding distribution of S. citrina. However, models based on nesting habitat alone were 4.7 times less likely than a model including functional connectivity as a predictor for the breeding distribution of S. citrina. Models of nest occurrence in relation to surrounding forest cover had lower model likelihoods than models that included graph-based functional connectivity, but these measures were highly confounded. Graph-based measures of connectivity explained more variation in nest occurrence than structural measures of forest connectivity, in both 2004 and 2005. These results suggest that S. citrina selected nesting areas that were functionally connected at their preferred gap crossing distances, but nesting habitat was a critically important predictor of nest occurrence patterns.

Melles, S.J., Fortin, M.-J., Lindsay, K., Badzinski, D. (2011)
Global Change Biology 17: 17-31

Read Abstract

Species’ ranges are dynamic, shifting in response to a large number of interrelated ecological and anthropogenic processes. Climate change is thought to be one of the most influential drivers of range shifts, but the effects of other confounded ecological processes are often ignored even though these processes may modify expected range responses to climate change. To determine the relative effects of climate, forest availability, connectivity, and biotic processes such as immigration and establishment, we examine range changes occurring in a species of bird, the Hooded Warbler (Wilsonia citrina). We focus predominantly on the periphery of the species’ northern range in Canada but we also examine data from the entire species’ range. Nesting records in southern Ontario were obtained from two breeding bird Atlases of Ontario separated by a period of 20 years (1981–1985 and 2001–2005), and the rate of range expansion was estimated by comparing the number of occupied areas in each Atlas. Twelve hypotheses of the relationship between the rate of range expansion and factors known to influence range change were examined using model-selection techniques and a mixed modeling approach (zero-inflated Poisson’s regression). Cooler temperatures were positively related to a lack of range expansion indicating that climate constrained the species’ distribution. Establishment probability (based on the number of occupied, neighboring Atlas squares) and immigration from populations to the south (estimated using independent data from the North American Breeding Bird Survey) were also important predictors of range expansion. These biotic process variables can mask the effects of forest availability and connectivity on range expansion. Expansion due to climate change may be slower in fragmented systems, but the rate of expansion will be influenced largely by biotic processes such as proximity to neighboring populations.

Optimizing the spatial pattern of networks for monitoring radioactive releases *Best paper award, Computers and Geosciences (2011)

Melles, S.J., Heuvelink, G.B.M., Twenhöfel, C.J.W. and Stöhlker, U. (2011)
Computers and Geosciences 37: 280-288

Read Abstract

This study presents a method to optimize the sampling design of environmental monitoring networks in a multi-objective setting. We optimize the permanent network of radiation monitoring stations in the Netherlands and parts of Germany as an example. The optimization method proposed combines minimization of prediction error under routine conditions with maximizing calamity detection capability in emergency cases. To calculate calamity detection capability, an atmospheric dispersion model was used to simulate potentially harmful radioactive releases. For each candidate monitoring network, we determined if the releases were detected within one, two and three hours. Four types of accidents were simulated: small and large nuclear power plant accidents, deliberate radioactive releases using explosive devices, and accidents involving the transport of radioactive materials. Spatial simulated annealing (SSA) was used to search for the optimal monitoring design. SSA was implemented by iteratively moving stations around and accepting all designs that improved a weighted sum of average spatial prediction error and calamity detection capability. Designs that worsened the multi-objective criterion were accepted with a certain probability, which decreased to zero as iterations proceeded. Results were promising and the method should prove useful for assessing the efficacy of environmental monitoring networks designed to monitor both routine and emergency conditions in other applications as well.

Baume, O.P., Skoien, J.O., Heuvelink, G.B.M., Pebesma, E.J. Melles, S.J.
International Journal of Applied Earth Observation and Geoinformation 13: 409-419

Read Abstract

Environmental issues such as air, groundwater pollution and climate change are frequently studied at spatial scales that cross boundaries between political and administrative regions. It is common for different administrations to employ different data collection methods. If these differences are not taken into account in spatial interpolation procedures then biases may appear and cause unrealistic results. The resulting maps may show misleading patterns and lead to wrong interpretations. Also, errors will propagate when these maps are used as input to environmental process models. In this paper we present and apply a geostatistical model that generalizes the universal kriging model such that it can handle heterogeneous data sources. The associated best linear unbiased estimation and prediction (BLUE and BLUP) equations are presented and it is shown that these lead to harmonized maps from which estimated biases are removed. The methodology is illustrated with an example of country bias removal in a radioactivity exposure assessment for four European countries. The application also addresses multicollinearity problems in data harmonization, which arise when both artificial bias factors and natural drifts are present and cannot easily be distinguished. Solutions for handling multicollinearity are suggested and directions for further investigations proposed.

Melles, S.J., Badzinski, D., Csillag, F., Fortin, M.-J., Lindsay, K. (2009)
International Journal of Applied Earth Observation and Landscape Ecology 24:519-531

Read Abstract

Nest locations of breeding birds are often spatially clustered. This tendency to nest together has generally been related to a patchy distribution of nesting habitat in landscape studies, but behavioral studies of species with clustered breeding patterns draw attention to the importance of social and biotic factors. Indeed, it is becoming increasingly apparent that the breeding system of many territorial, migrant birds may be semi-colonial. The reasons for, and extent of, spatial clustering in their breeding systems are not well understood. Our goal was to tease apart the influence of habitat availability and social drivers of clustered breeding in a neotropical migrant species, the hooded warbler (Wilsonia citrina). To test alternative hypotheses related to clustered habitat or conspecific attraction, we combined a habitat classification based on remote sensing with point pattern analysis of nesting sites. Nest locations (n = 150, 1999–2004), collected in a 1213 ha forested area of Southern Ontario (Canada), were analyzed at multiple spatial scales. Ripley’s K and pair-correlation functions g (uni- and bivariate) were used to test whether nests were clustered merely because potential nesting habitat was also clustered, or whether nests were additionally clustered with respect to conspecifics. Nest locations tended to be significantly clustered at intermediate distances (particularly between 240 and 420 m). Nests were randomly distributed within available habitat at larger distance scales, up to 1500 m. A reasonable hypothesis to explain the detected additional clustering, and one that is consistent with the results of several behavioral studies, is that females pack their nests more tightly than the available habitat requires to be situated closer to their neighbors’ mates. Linking spatially explicit, point pattern analysis with strong inference based on Monte Carlo tests may bring us closer to understanding the generality and reasons behind conspecific attraction at different spatial scales.

Melles, S.J., Heuvelink, G.B.M., Twenhöfel, C.J.W., Stöhlker, U. (2008)
Lecture Notes in Computer Science (Eds. O. Gervasi, B. Murgante, A. Laganà, D. Taniar, Y. Mun, M. Gavrilova), vol. 5072, 444-458. 

Read Abstract

This paper applies a recently developed optimization method to examine the design of networks that monitor radiation under routine conditions. Annual gamma dose rates were modelled by combining regression with interpolation of the regression residuals using spatially exhaustive predictors and an anisotropic variogram of the residuals. Locations of monitoring stations were optimized by minimizing the spatially averaged regression kriging standard deviation. Results suggest that the current network design is near optimal in terms of interpolation error in predicted gamma dose rates. When the network was thinned to fewer stations, spatial optimization was more effective at reducing the interpolation error. Given that some EU countries are considering reducing station density in border regions, the analysis reported here may be useful in guiding which stations can be removed.

Melles, S.J. (2005)
Urban Habitats. 3(1): 4-26.

Read Abstract

The unequal distribution of wealth in cities contributes to other forms of spatial, social, and biological inequities in complex, interacting, and self-reinforcing ways. Recent work on urban birds has often focused on community-level correlation studies of short duration in which many points along an urban gradient are surveyed for birds, and the data are related to various ecological variables measured at multiple scales. Spatial variation in urban bird communities may also reflect socioeconomic variables and cultural differences among the human population. The purpose of this paper was to examine whether socioeconomic factors (such as mean family income and ethnic diversity) also relate to the diversity and abundance of birds in Vancouver, British Columbia. I used redundancy analysis to characterize the socioeconomic gradient in a citywide study of the bird community in 44 census-defined neighborhoods. Mean family income, census tract area, and ethnicity were some of the dominant variables that correlated with most of the variation in the bird community. I found no direct relationship between neighborhood age and bird diversity and abundance. Results demonstrate that wealthier neighborhoods have more native species of birds and that these native species increase in abundance as the socioeconomic status of the neighborhood improves. With two-thirds of the world’s population expected to live in cities by 2030, more and more people will grow up surrounded by a depauperate community of birds, and this could adversely affect the way people perceive, appreciate, and understand nature. Ultimately, as city birdlife diminishes and urban dwellers become dissociated from the natural diversity it represents, popular support for preserving and restoring such diversity may wane, allowing ecological conditions to further erode.

Melles, S.J., Glenn, S., Martin, K. (2003)
Conservation Ecology now Ecology and Society 7(1): 5

Read Abstract

For birds in urban environments, the configuration of local habitat within the landscape may be as critical as the composition of the local habitat itself. We examined the relative importance of environmental attributes (e.g., tree cover, composition, and number of tree species) measured at different spatial scales in relation to urban bird species richness and abundance. We expected that some bird species and nesting guilds would have a closer association with landscape-level features (within 1000 m), such as proximity to large forested areas, than with local-scale habitat measures (within 50 m). To investigate this, avian community data were collected at 285 point-count stations in 1997 and 1998 along four roadside transects located in Vancouver and Burnaby, British Columbia, Canada. Transects (5–25 km in length) bisected three large parks (>324 ha) and proceeded along residential streets in urban and suburban areas. In total, 48 bird species were observed, including 25 common species. Species richness declined in relation to a gradient of increasing urbanization, as measured by local- and landscape-level habitat features. We further examined the significance and importance of local- vs. landscape-level habitat attributes using logistic regression and found that both scales explained the presence/absence distributions of residential birds. Local-scale habitat features such as large coniferous trees, berry-producing shrubs, and freshwater streams were of particular importance in estimating the likelihood of finding bird species. Landscape measures, particularly forest cover (within 500 m) and park area (measured at different scales as a function of distance from point-count stations) significantly improved likelihood estimations based solely on local-scale habitat features. Our results suggest that both local- and landscape-scale resources were important in determining the distribution of birds in urban areas. Parks, reserves, and the surrounding residential areas should be integrated into urban planning and development designs to maintain resident avifauna and overall species diversity in urban environments.

Book Chapters

Heuvelink G.B.M., Griffith D., Hengl T., Melles S.J. (2012)
John Wiley & Sons

Read Abstract

Sampling design optimization in space has been well covered in the geostatistical literature. This chapter discusses a criterion that simultaneously minimizes the variance of the estimation error of a linear trend and that of the interpolation error of the kriging residual. Minimization of the space-time average universal kriging variance is achieved for monthly temperatures of the Upper Austria region. The chapter considers three optimization scenarios: optimization of a static design, in which 10 locations are selected from 35 current stations such that increases in the average space-time universal kriging variance caused by the thinning are as small as possible; optimization of a static design in which the 10 stations can be located anywhere within the Upper Austria region; and, optimization of a dynamic design, in which the 10 station locations may be changed at the start of each year. The chapter reviews the space-time universal kriging and spatial simulated annealing.

Fortin, M.-J., Melles, S.J. (2009)
Springer, New York. Chapter 6,  Pages: 137-160

Refereed Conference Proceedings

Melles, S.J., Jones, N.E., Schmidt, B., Rayfield, B. (2011)
Procedia Environmental Sciences, 1st conference on Spatial Statistics, Enschede, the Netherlands. 6 pp.

Read Abstract

We present two approaches to predict movement of surface waters through heterogeneous terrain: least-cost graphs and circuits. These methods were compared with the commonly applied ‘deterministic eight’ (D8) approach by examining distances from extracted drainage networks to known-flow boundaries. Overlap statistics and classification accuracy estimates indicated that least-cost graphs and the circuit-based approaches hold some promise and avoid many of the pitfalls associated with removing depressions from a DEM. The majority of stream segments were correctly identified by all methods, but errors of omission and commission were fairly high.

Melles, S.J., Benoy, G., Booty, B., Leon, L., Vanrobaeys, J., Wong, I. (2010)
Proceedings of International Environmental Modelling and Software Society, 2010 International Congress on Environmental Modelling and Software Modelling for Environment’s Sake, Fifth Biennial Meeting, Ottawa, Canada. 6 pp. 

Read Abstract

Lake Winnipeg is Canada’s 6th largest freshwater lake and is subject to an increasing rate of eutrophication as a result of nonpoint source pollution from numerous sources including farms and municipal wastewaters. We examine scenarios designed to compare the relative effects of wetland restoration and position on modelled nutrient loadings to Lake Winnipeg from a pilot watershed in the Lake’s Basin, the La Salle River watershed. Scenarios were examined using the Soil and Water Assessment Tool (SWAT). SWAT is a well-known watershed scale hydrologic model designed to assess non-point source pollution loadings to contributing streams across a wide range of scales. Modelled results suggested that increasing wetland cover to historic levels decreased yearly nutrient loadings by 9-21% for both TN and TP. When 25% of historic wetland areas were restored at subwatershed outlets, equivalent or better nutrient reductions were attained. But placing all wetland area at the watershed outlet did not result in as substantial nutrient reductions. There was a larger range of uncertainty when wetlands were modelled across all subwatersheds than when the entire wetland area was modelled at the outlet. These results may indicate that wetland position is as important as wetland amount in terms of nutrient reductions.

Leon, L..F., Booty, W., Wong, I., McCrimmon, C. Melles, S., Benoy, G., Vanrobaeys, J. (2010)
Proceedings of International Environmental Modelling and Software Society, 2010 International Congress on Environmental Modelling and Software Modelling for Environment’s Sake, Fifth Biennial Meeting, Ottawa, Canada. 6 pp. 

Read Abstract

Estimating non-point source pollution from watersheds and the effects of mitigation measures (e.g. beneficial management practices or BMPs) is an important step in managing and protecting water quality, not only at the basin level where it originates, but also at the receiving waters such as reservoirs, lakes or oceans. Lake Winnipeg is a prime example of such land-lake interactions, where eutrophication and increased algal blooms in the lake are fueled, as evidence suggests, from agricultural sources of nutrients in the region. Over the years, simulation models at the watershed level have been applied to aid in the understanding and management of surface runoff, nutrients and sediment transport processes. Similarly, models with different degrees of complexity are used to simulate the aquatic ecology and water quality in lakes. The Soil and Water Assessment Tool (SWAT) is a widely known watershed model, which provides estimations of runoff, sediment yield, and nutrient loads at a sub-basin level. Here we examine the application of SWAT to one of three pilot watersheds on the Lake Winnipeg basin in order to investigate the impacts and uncertainties of different BMPs on nutrient loading in the targeted catchment areas. We also explore avenues for scaling and propagating such loads and uncertainties into the receiving lake models.

Melles, S.J., Heuvelink, G.B.M., Twenhöfel, C.J.W. van Dijk, A., Hiemstra, P., Baume, O., Stöhlker, U. (2009)
Proceedings of StatGIS conference, Milos, Greece. 6 pp. 

Read Abstract

The design of radiation monitoring networks were optimized by combining a geostatistical assessment of routine prediction error with simulation modelling to assess network signalling function in emergency settings. A physical atmospheric dispersion model was used to simulate radioactive releases throughout the study area under different accident scenarios and varying weather conditions (e.g. small nuclear power plant accidents and mock human-caused radioactive emissions).  Network signalling function was defined as the ability to detect radioactivity above a critical threshold within 3 hours of a nuclear release.  Spatial simulated annealing was used to obtain optimal monitoring designs by moving stations around and accepting those designs that reduced a weighted sum of two criteria (prediction error of mean annual background radiation and network signalling function). Results were promising and the method should prove useful for assessing the efficacy of hazard monitoring networks designed to detect the unlikely event of a nuclear emergency.

Melles, S.J., Beekhuizen, J. de Bruin, S. Heuvelink, G.B.M. van Dijk, A. Twenhöfel, C.J.W.  (2008)
In: IfGI prints. Proceedings of the 6th Geographic Information Days (Eds. E. Pebesma, M. Bishr, T. Bartoschek), 32, 189-198. 

Read Abstract

To prepare for the unlikely event of a nuclear emergency, atmospheric dispersion models are used to forecast release concentrations and radionuclide deposition rates.  By  combining  model  predictions  with  spatial analyses  of  real-time  gamma-dose  rate  measurements at  permanent  ( n = 153)  and  mobile  device  stations  ( n  =  8),  more  informed  intervention  decisions  can  be  considered.  The  main  objective  of  this  research  was  to  deter-mine where to optimally locate mobile measuring devices in order to minimize  costs  associated  with  faulty  decisions.  Incorrect  decisions  may  be  the result  of  uncertainty  in  predicting  the  spread  of  a  potential  radioactive plume.  Spatial  simulated  annealing  was  used  to  optimally  position  mobile devices  by minimizing a weighted sum of expected false positive and false negative  areas  (i.e.,  false classification into safe and unsafe zones). Results indicated that the optimal placement of mobile devices tended to be in areas at, or just inside, the edge of an advancing radioactive plume.

Papers in revision

Optimizing Road Weather Information System (RWIS) – a novel approach

Fu, L., Kwon, T.J., Melles, S.J., Perchanok, M.S.
(Submitted full paper upon acceptance of abstract, PIARC World Road Association Congress to be held in Seoul, Korea, Nov. 2-6, 2015). 16 pp.