Scaling-up spatially-explicit ecological models using graphics processors

How the properties of ecosystems relate to spatial scale is a prominent topic in current ecosystem research. Despite this, spatially explicit models typically include only a limited range of spatial scales, mostly because of computing limitations. Here, we describe the use of graphics processors to efficiently solve spatially explicit ecological models at large spatial scale using the CUDA language extension. We explain this technique by implementing three classical models of spatial self-organization in ecology: a spiral-wave forming predator-prey model, a model of pattern formation in arid vegetation, and a model of disturbance in mussel beds on rocky shores. Using these models, we show that the solutions of models on large spatial grids can be obtained on graphics processors with up to two orders of magnitude reduction in simulation time relative to normal pc processors. This allows for efficient simulation of very large spatial grids, which is crucial for, for instance, the study of the effect of spatial heterogeneity on the formation of self-organized spatial patterns, thereby facilitating the comparison between theoretical results and empirical data. Finally, we show that large-scale spatial simulations are preferable over repetitions at smaller spatial scales in identifying the presence of scaling relations in spatially self-organized ecosystems. Hence, the study of scaling laws in ecology may benefit significantly from implementation of ecological models on graphics processors.     

An inverse model for the analysis of the Venice lagoon food web

A steady-state model of the Venice lagoon food web was constructed, based on a comprehensive set of data, which were collected in the years 2001-2005. Energy flows were estimated by means of an inverse methodology of constrained optimization based on the Minimum Norm criterion, i.e. on the minimization of both the sum of squares of the residuals and of the sum of squares of energy flows. The solution was constrained by a set inequalities, which were derived from general eco-physiological knowledge and site specific data on energy flows. The trophic network was represented by thirty-two nodes, including single-species compartments for the species of high economical or ecological relevance. Mass balance equations were weighted, in order to obtain meaningful results in presence of large differences, up to 5 orders of magnitude, among biomasses. A perturbation technique was applied, with the purpose of reducing the risk of finding solutions heavily affected by the set of constraints and of obtaining a more robust representation of the energy flows. The main patterns of energy flow are consistent with those obtained in previous attempts at modelling the Venice lagoon food web. Micro- and macro-phytobenthos account for the largest fraction of the primary production. Energy is then transferred towards higher trophic levels by means of two main pathways: the recycling of dead biomass through the detritus compartment and the direct consumption by grazers. The first pathway is the most important and accounts for approximately two/thirds of the energy transferred to the second trophic level.     

Disentangling the effects of heterogeneity, stochastic dynamics and sampling in a community of aquatic insects

A dynamic and heterogeneous species abundance model generating the lognormal species abundance distribution is fitted to time series of species data from an assemblage of stoneflies and mayflies (Plecoptera and Ephemeroptera) of an aquatic insect community collected over a period of 15 years. In each year except one, we analyze 5 parallel samples taken at the same time of the season giving information about the over-dispersion in the sampling relative to the Poisson distribution. Results are derived from a correlation analysis, where the correlation in the bivariate normal distribution of log abundance is used as measurement of similarity between communities. The analysis enables decomposition of the variance of the lognormal species abundance distribution into three components due to heterogeneity among species, stochastic dynamics driven by environmental noise, and over-dispersion in sampling, accounting for 62.9, 30.6 and 6.5% of the total variance, respectively. Corrected for sampling the heterogeneity and stochastic components accordingly account for 67.3 and 32.7% of the among species variance in log abundance. By using this method, it is possible to disentangle the effect of heterogeneity and stochastic dynamics by quantifying these components and correctly remove sampling effects on the observed species abundance distribution.     

Common challenges for ecological modelling: Synthesis of facilitated discussions held at the symposia organized for the 2009 conference of the International Society for Ecological Modelling in Quebec City, Canada, (October 6-9, 2009)

The eleven symposia organized for the 2009 conference of the International Society for Ecological Modelling (ISEM 2009) held in Quebec City, Canada, October 6-9, 2009, included facilitated discussion sessions following formal presentations. Each symposium focused on a specific subject, and all the subjects could be classified into three broad categories: theoretical development, population dynamics and ecosystem processes. Following discussions with the symposia organizers, which indicated that they all shared similar issues and concerns, the facilitated discussions were task-oriented around four basic questions: (1) key challenges in the research area, (2) generating and sharing new ideas, (3) improving collaboration and networking, and (4) increasing visibility to decision-makers, partners and clients. Common challenges that emerged from the symposia included the need for improved communication and collaboration among different academic disciplines, further progress in both theoretical and practical modelling approaches, and accentuation of technology transfer. Regarding the generation and sharing of new ideas, the main issue that emerged was the type of positive interactions that should be encouraged among potential collaborators. The usefulness of the Internet, particularly for the sharing of open-source software and conducting discussion forums, was highlighted for improving collaboration and networking. Several communication tools are available today, and it is important for modellers to use them more intensively. Visibility can be increased by publishing professional newsletters, maintaining informal contacts with the public, organizing educational sessions in primary and secondary schools, and developing simplified analytical frameworks and pilot studies. Specific issues raised in each symposium are also discussed.     

Null models reveal preferential sampling, spatial autocorrelation and overfitting in habitat suitability modelling

Nowadays, species are driven to extinction at a high rate. To reduce this rate it is important to delineate suitable habitats for these species in such a way that these areas can be suggested as conservation areas. The use of habitat suitability models (HSMs) can be of great importance for the delineation of such areas. In this study MaxEnt, a presence-only modelling technique, is used to develop HSMs for 223 nematode species of the Southern Bight of the North Sea. However, it is essential that these models are beyond discussion and they should be checked for potential errors. In this study we focused on two categories (1) errors which can be attributed to the database such as preferential sampling and spatial autocorrelation and (2) errors induced by the modelling technique such as overfitting, In order to quantify these adverse effects thousands of nulls models were created. The effect of preferential sampling (i.e. some areas where visited more frequenty than others) was investigated by comparing model outcomes based from null models sampling the actual sampling stations and null models sampling the entire mapping area (Raes and ter Steege, 2007). Overfitting is exposed by a fivefold cross-validation and the influence of spatial autocorrelation is assessed by separating test and training sets in space. Our results clearly show that all these effects are present: preferential sampling has a strong effect on the selection of non-random species models. Crossvalidation seems to have less influence on the model selection and spatial autocorrelation is also strongly present. It is clear from this study that predefined thresholds are not readily applicable to all datasets and additional tests are needed in model selection.     

Occupational Health Care in Small and Medium-Sized Enterprises—How Many Doctors Do We Need and How Do We Ensure Good Care?

The translation of the framework directive 89/391/EEC (Council Directive 89/391/EEC) into national law aims at supplying occupational protection and health care to all employees of large, small, and medium-sized enterprises (SMEs) likewise, depending on assessed exposure. Prior incomplete protection of the German workforce with bias against the SMEs requires an assessment of quantitative and qualitative adequacy of present occupational health care practice and also of future needs.

Therefore, the Federal Institute for Occupational Safety and Health of Germany initiated a study to evaluate the present state of occupational health care in 4 regions with different geographic and economic structure. Based on these data the future demand for adequately trained occupational physicians will be estimated by employing a statistical method that allows for including a large number of modifying variables (economic development, demographic change, etc.).

Expected result of the applied technique are the estimated minimal and maximal number of occupational physicians that have to be trained to meet future demand. In the same study models of best practice will be identified and evaluated for general application. The project started in October 1997 and will be finished by January 2000.     

Heat Gain From Thermal Radiation Through Protective Clothing With Different Insulation,Reflectivity and Vapour Permeability

The heat transferred through protective clothing under long wave radiation compared to a reference condition without radiant stress was determined in thermal manikin experiments. The influence of clothing insulation and reflectivity, and the interaction with wind and wet underclothing were considered. Garments with different outer materials and colours and additionally an aluminised reflective suit were combined with different number and types of dry and pre-wetted underwear layers. Under radiant stress, whole body heat loss decreased, i.e., heat gain occurred compared to the reference. This heat gain increased with radiation intensity, and decreased with air velocity and clothing insulation. Except for the reflective outer layer that showed only minimal heat gain over the whole range of radiation intensities, the influence of the outer garments’ material and colour was small with dry clothing. Wetting the underclothing for simulating sweat accumulation, however, caused differing effects with higher heat gain in less permeable garments.     

Deforestation in the drylands of Africa: Quantitative modelling approach

This paper has developed a two-sector model of forestry that gives some insights into the interactions of poverty, land tenure insecurity, economic growth, trade and deforestation in the context of the drylands of Africa. Externally, the model results show that the world demand for non-timber forestry products can lead to a win-win situation whereby both trade and environmental sustainability can be promoted. Monopolistic measures to control the supply of these products, for short-term price gains, are harmful economically and environmentally. By contrast, productivity-raising measures have a favourable impact on both the economy and the environment. Internally, the model results show that the growth of the rest of the economy has driven the economic growth of the forestry sector, but at the same time, its environmental degradation. Poverty and land tenure insecurity also contribute to this degradation. Fortunately, policies promoting basic needs satisfaction and property rights security are shown to be effective in mitigating environmental degradation in the forestry sector. The paper results contribute towards a more sustainable exploitation of forestry.Readers should send their comments on this paper to: BhaskarNath@aol.com within 3 months of publication of this issue.     

USE OF TOPOLOGICAL INFORMATION IN HYDROGRAPH ESTIMATION1

ABSTRACT: Discharge hydrographs computed from the theory of linear flow through topologically random channel networks are compared to actual discharge hydrographs for basins in semiarid regions of central Wyoming. The basins drained by the channel networks range in size from 0.69 to 10.8 square miles. Topological information consisting of stream-network magnitude and link -length distribution parameters are used in calibrating celerity values that ensure that the peak discharge and excess rainfall volume of the resulting computed hydrographs match the peak discharge and excess-rainfall volume of the actual hydrographs. Results indicate that, for a given peak discharge and excess-rainfall volume in a basin, the sensitivity of the calibrated celerity values to excess-rainfall duration is small if the ratio of excess-rainfall volume to peak discharge is at least 1.75 times the excess-rainfall duration.     

Predicting Future Threats to the Native Vegetation of Robinson Crusoe Island, Juan Fernandez Archipelago, Chile

Abstract:   The Juan Fernandez Archipelago, a Chilean national park and biosphere reserve, is 700 km west of continental Chile. Invasive plant species pose major threats to the native, highly endemic vegetation, especially on Robinson Crusoe Island (Mas a Tierra), where there is a permanent settlement. We used historical and recent vegetation map data as well as recent vegetational point data to reconstruct vegetational changes on the island since the early twentieth century. In addition, we used logistic regression models with environmental variables or their surrogates (elevation, solar radiation, topographic similarity index, slope position) to estimate the potential distributions of the worst invaders. Native vegetation has been affected most severely by Acaena argentea , Aristotelia chilensis , Rubus ulmifolius , and Ugni molinae , leading to a significant decrease of endemic plants in the montane forests and native shrublands. The native forest has decreased by approximately one-third. The area affected by Aristotelia chilensis increased from 6.5% to 14% of the total island area. Ugni molinae , once rare, is now abundant (4.6%), as is Acaena (11.9%). Rubus ulmifolius , not present 80 years ago, now covers about 7% of the island's surface. The distributions of all studied species except Rubus ulmifolius are significantly ( p < 0.05) controlled by the environmental factors we examined. Inferring from potential distributions of Aristotelia chilensis and Ugni molinae , 50% of the native montane forest could be invaded or replaced (with a probability of >60%) by these plants. Based on the invasion speed of the past 80 years, this would take another 80 years if conservation measures do not succeed.