Prevalence-adjusted optimisation of fuzzy models for species distribution

Most performance criteria which have been applied to train ecological models focus on the accuracy of the model predictions. However, these criteria depend on the prevalence of the training set and often do not take into account ecological issues such as the distinction between omission and commission errors. Moreover, a previous study indicated that model training based on different performance criteria results in different optimised models. Therefore, model developers should train models based on different performance criteria and select the most appropriate model depending on the modelling objective. This paper presents a new approach to train fuzzy models based on an adjustable performance criterion, called the adjusted average deviation (aAD). This criterion was applied to develop a species distribution model for spawning grayling in the Aare River near Thun, Switzerland. To analyse the strengths and weaknesses of this approach, it was compared to model training based on other performance criteria. The results suggest that model training based on accuracy-based performance criteria may produce unrealistic models at extreme prevalences of the training set, whereas the aAD allows for the identification of more accurate and more reliable models. Moreover, the adjustable parameter in this criterion enables modellers to situate the optimised models in the search space and thus provides an indication of the ecological model relevance. Consequently, it may support modellers and river managers in the decision making process by improving model reliability and insight into the modelling process. Due to the universality and the flexibility of the approach, it could be applied to any other ecosystem or species, and may therefore be valuable to ecological modelling and ecosystem management in general.     

Modelling habitat overlap among sympatric mesocarnivores in southern Illinois,USA

Few researchers have developed large-scale habitat models for sympatric carnivore species. We created habitat models for red foxes (Vulpes vulpes), coyotes (Canis latrans) and bobcats (Lynx rufus) in southern Illinois, USA, using the Penrose distance statistic, remotely sensed landscape data, and sighting location data within a GIS. Our objectives were to quantify and spatially model potential habitat differences among species. Habitat variables were quantified for 1-km² buffered areas around mesocarnivore sighting locations. Following variable reduction procedures, five habitat variables (percentage of grassland patches, interspersion–juxtaposition of forest patches, mean fractal dimension of wetland patches and the landscape, and road density) were used for analysis. Only one variable differed (P < 0.05) between red fox and coyote sighting areas (road density) and bobcat and coyote sighting areas (mean fractal dimension of the landscape). However, all five variables differed between red fox and bobcat sighting areas, indicating considerable differences in habitat affiliation between this pair-group. Compared to bobcats, red fox sightings were affiliated with more grassland cover and larger grassland patches, higher road densities, lower interspersion and juxtaposition of forest patches, and lower mean fractal dimension of wetland patches. These differences can be explained by different life history requirements relative to specific cover types. We then used the Penrose distance statistic to create habitat models for red foxes and bobcats, respectively, based on the five-variable dataset. An independent set of sighting locations were used to validate these models; model fit was good with 65% of mesocarnivore locations within the top 50% of Penrose distance values. In general, red foxes were affiliated with mixtures of agricultural and grassland cover, whereas bobcats were associated with a combination of grassland, wetland, and forest cover. The greatest habitat overlap between red foxes and bobcats was found at the interface between forested areas and more open cover types. Our study provides insight into habitat overlap among sympatric mesocarnivores, and the distance-based modelling approach we used has numerous applications for modelling wildlife–habitat relationships over large scales.     

A pattern-oriented modelling approach to simulating populations of grey partridge

A number of wildlife species including the grey partridge (Perdix perdix) have shown dramatic post-war population declines. Multiple drivers have been proposed as reasons for the declines, for example agrochemical use and intensification of agricultural practices, climate, predation, and changes in landscape structure. These drivers may interact in non-linear ways and are inherently spatio-temporal in nature. Therefore models used to investigate mechanisms should be spatio-temporal, of proper scale, and have a high degree of biological realism. Here we describe the development and testing of an agent-based model (ABM) of grey partridge using a well documented pre-decline historical data set in conjunction with a pattern-oriented modelling (POM) approach. Model development was an iterative process of defining performance criteria, testing model behaviour, and reformulating as necessary to emulate system properties whilst ensuring that internal mechanisms were biologically realistic. The model was documented using ODdox, a new protocol for describing large agent-based models. Parameter fitting in the model was achieved to within ±2% accuracy for 15 out of 17 field data patterns used, and within 5% for the remaining two. Tests of interactions between input parameters showed that 62% of parameter pairs tested had significant interactions underlining the complex nature of the model structure. Sensitivity analysis identified chick mortality as being the most sensitive factor, followed by adult losses to hunting and adult overwinter mortality, agreeing in general with previous partridge models. However, the ABM used here could separate individual drivers, providing a better understanding of the underlying mechanisms behind population regulation, and allowing factors to be compared directly. The ABM used is rich in output signals allowing detailed testing and refinement of the model. This approach is particularly suited to systems such as the partridge system where data for comparison to model outputs is readily available. Despite the accurate fit between historical data and model output, making use of the predictive power of the approach the model requires further calibration and testing under modern field conditions.     

Comparison and ranking of different modelling techniques for prediction of site index in Mediterranean mountain forests

Forestry science has a long tradition of studying the relationship between stand productivity and abiotic and biotic site characteristics, such as climate, topography, soil and vegetation. Many of the early site quality modelling studies related site index to environmental variables using basic statistical methods such as linear regression. Because most ecological variables show a typical non-linear course and a non-constant variance distribution, a large fraction of the variation remained unexplained by these linear models. More recently, the development of more advanced non-parametric and machine learning methods provided opportunities to overcome these limitations. Nevertheless, these methods also have drawbacks. Due to their increasing complexity they are not only more difficult to implement and interpret, but also more vulnerable to overfitting. Especially in a context of regionalisation, this may prove to be problematic. Although many non-parametric and machine learning methods are increasingly used in applications related to forest site quality assessment, their predictive performance has only been assessed for a limited number of methods and ecosystems.In this study, five different modelling techniques are compared and evaluated, i.e. multiple linear regression (MLR), classification and regression trees (CART), boosted regression trees (BRT), generalized additive models (GAM), and artificial neural networks (ANN). Each method is used to model site index of homogeneous stands of three important tree species of the Taurus Mountains (Turkey): Pinus brutia, Pinus nigra and Cedrus libani. Site index is related to soil, vegetation and topographical variables, which are available for 167 sample plots covering all important environmental gradients in the research area. The five techniques are compared in a multi-criteria decision analysis in which different model performance measures, ecological interpretability and user-friendliness are considered as criteria.When combining these criteria, in most cases GAM is found to outperform all other techniques for modelling site index for the three species. BRT is a good alternative in case the ecological interpretability of the technique is of higher importance. When user-friendliness is more important MLR and CART are the preferred alternatives. Despite its good predictive performance, ANN is penalized for its complex, non-transparent models and big training effort.     

Application of a Random Forest algorithm to predict spatial distribution of the potential yield of Ruditapes philippinarum in the Venice lagoon, Italy

We present a modelling framework that combines machine learning techniques and Geographic Information Systems to support the management of an important aquaculture species, Manila clam (Ruditapes philippinarum). We use the Venice lagoon (Italy), the first site in Europe for the production of R. philippinarum, to illustrate the potential of this modelling approach. To investigate the relationship between the yield of R. philippinarum and a set of environmental factors, we used a Random Forest (RF) algorithm. The RF model was tuned with a large data set (n = 1698) and validated by an independent data set (n = 841). Overall, the model provided good predictions of site-specific yields and the analysis of marginal effect of predictors showed substantial agreement among the modelled responses and available ecological knowledge for R. philippinarum. The most influent environmental factors for yield estimation were percentage of sand in the sediment, salinity, and water depth. Our results agree with findings from other North Adriatic lagoons. The application of the fitted RF model to continuous maps of all the environmental variables allowed estimates of the potential yield for the whole basin. Such a spatial representation enabled site-specific estimates of yield in different farming areas within the lagoon. We present a possible management application of our model by estimating the potential yield under the current farming distribution and comparing it to a proposed re-organization of the farming areas. Our analysis suggests a reduction of total yield is likely to result from the proposed re-organization.     

Sanitary felling of Norway spruce due to spruce bark beetles in Slovenia: A model and projections for various climate change scenarios

A model is presented to predict sanitary felling of Norway spruce (Picea abies) due to spruce bark beetles (Ips typographus, Pityogenes chalcographus) in Slovenia according to different climate change scenarios. The model incorporates 21 variables that are directly or indirectly related to the dependent variable, and that can be arranged into five groups: climate, forest, landscape, topography, and soil. The soil properties are represented by 8 variables, 4 variables define the topography, 4 describe the climate, 4 define the landscape, and one additional variable provides the quantity of Norway spruce present in the model cell. The model was developed using the M5′ model tree. The basic spatial unit of the model is 1 km², and the time resolution is 1 year. The model evaluation was performed by three different measures: (1) the correlation coefficient (51.9%), (2) the Theil's inequality coefficient (0.49) and (3) the modelling efficiency (0.32). Validation of the model was carried out by 10-fold cross-validation. The model tree consists of 28 linear models, and model was calculated for three different climate change scenarios extending over a period until 2100, in 10-year intervals. The model is valid for the entire area of Slovenia; however, climate change projections were made only for the Maribor region (596 km²). The model assumes that relationships among the incorporated factors will remain unchanged under climate change, and the influence of humans was not taken into account. The structure of the model reveals the great importance of landscape variables, which proved to be positively correlated with the dependent variable. Variables that describe the water regime in the model cell were also highly correlated with the dependent variable, with evapotranspiration and parent material being of particular importance. The results of the model support the hypothesis that bark beetles do greater damage to Norway spruce artificially planted out of its native range in Slovenia, i.e., lowlands and soils rich in N, P, and K. The model calculation for climate change scenarios in the Maribor region shows an increase in sanitary felling of Norway spruce due to spruce bark beetles, for all scenarios. The model provides a path towards better understanding of the complex ecological interactions involved in bark beetle outbreaks. Potential application of the results in forest management and planning is discussed.     

A comparison of four process-based models and a statistical regression model to predict growth of Eucalyptus globulus plantations

In forest management and ecological research, consideration of the impacts and risks of climate change or management optimisation is complex. Computer models have long been applied as tools for these tasks. Process-based forest growth models claim to overcome the limitations of empirical statistical models, but the capacity of different process-based models and modelling approaches have rarely been compared directly. This study evaluates stepwise multiple regression models in comparison to four process-based modelling approaches (3-PG, 3-PG+, CABALA and Forest-DNDC) for greenfield predictions of Eucalyptus globulus plantation growth from 2 to 8 years after planting throughout southern Australia.     

Design of a Domain Specific Language for modelling processes in landscapes

The modelling of processes that occur in landscapes is often confronted to issues related to the representation of space and the difficulty of properly handling time and multiple scales. In order to investigate these issues, a flexible modelling environment is required. We propose to develop such a tool based on a Domain Specific Language (DSL) that capitalises on the service-oriented architecture (SOA) paradigm. The modelling framework around the DSL is composed of a model building environment, a code generator and compiler, and a program execution platform. The DSL introduces five language elements (entity, service, relation, scenario and datafacer) that can be combined to offer a wide range of possibilities for modelling in space and time at different scales. When developing a model, model parts are either built using the DSL or taken from libraries of previously built ones, and adapted to the specific model. The practical usage of the DSL is illustrated first with the Lotka–Volterra model, and then with a landscape modelling experiment on the spread of a mosquito-borne disease in the Sahelian region of West Africa. An interesting characteristic of this approach is the possibility of adding new elements into an existing model, and replacing others with more appropriate ones, thus allowing potentially complex models to be built from simpler parts.     

Temporal limits to simulating the future spread pattern of invasive species: Buddleja davidii in Europe and New Zealand

The spread of invasive species is a major ecological and economic problem. Dynamic spread modelling is a potentially valuable tool to assist regional and central government authorities to monitor and control invasive species. To date a lack of suitable data has meant that most broad scale dispersal models have not been validated with independent datasets, and so their predictive ability and reliability has remained unscrutinised. A dynamic, stochastic dispersal model of the widely invasive plant Buddleja davidii was calibrated on European spread data and then used to project the temporal progression of B. davidii's distribution in New Zealand, starting from several different historical distributions. To assess the model's performance, we constructed an occupancy map based on the average number of simulation realisations that have a population present. The application of Receiver Operating Characteristic (ROC) curves to occupancy maps is introduced, but with specificity substituted by the proportion of available area used in a realisation. A derivative measure, the partial area under these curves when assessed through time (pAUC), is introduced and used to assess overall performance of the spread model. The model was able to attain a high level of model sensitivity, encompassing all of the known locations within the occupancy envelope. However, attempting to simulate the spread of this invasive species beyond a decade had very low model specificity. This is due to several factors, including the exponential process of spread (the further a population spreads the more sites exist from which it can spread stochastically), and the Markovian chain property of the stochastic system whereby differences between realisations compound through time. These features are seen in many reports of spread models, without being explicitly acknowledged. Our measure of pAUC through time allows a model's temporal performance and its specificity to be simultaneously assessed. While the rapid deterioration in model performance limits the utility of this type of modelling for forecasting long-term broad-scale strategic management of biological invasions, it does not necessarily limit its attractiveness for informing smaller scale and shorter term invasion management activities such as surveillance, containment and local eradication.     

New hydroepidemiological models of indicator organisms and zoonotic pathogens in agricultural watersheds

Simple analytical models are derived to assess how a series of cattle animal farms affect the transport and fate of an indicator organism (Escherichia coli) and a zoonotic pathogen (Campylobacter) in a stream. Separate steady-state mass-balance models are developed and solved for the ultimate minimum and maximum concentrations for the two organisms. The E. coli model assumes that the organism is ubiquitous and abundant in the animals’ digestive tracts. In contrast, a simple dose-response model is employed to relate the Campylobacter prevalence to drinking water drawn from the stream. Because faecal indicators are commonly employed to assess the efficacy of best management practice (BMP) interventions, we also employ the models to assess how BMPs impact pathogen levels. The model provides predictions of (a) the relative removal efficacy for Campylobacter and (b) the prevalence of Campylobacter infection among farm animals after implementation of BMPs. Dimensionless numbers and simple graphs are developed to assess how prevalence is influenced by a number of factors including animal density and farm spacing. A significant outcome of this model development is that the numerous dimensional input and parameter variables are reduced to a group of just four dimensionless Campylobacter-related quantities, characterizing: animal density; in-stream attenuation; animal-to-animal transmission; and infection recovery. Calculations reveal that for some constellations of these four quantities there can be a greater-than-expected benefit in that the proportional reduction of stream Campylobacter concentrations post-BMP can substantially exceed the proportional reduction of concentrations of E. coli in that stream. In addition, a criterion for system sterility (i.e., the conditions required for the farm infection rate to decrease with downstream distance) is derived.     

Nutrient criteria for lakes,ponds, and reservoirs: A Bayesian TREED model approach

We develop regional-scale eutrophication models for lakes, ponds, and reservoirs to investigate the link between nutrients and chlorophyll-a. The Bayesian TREED (BTREED) model approach allows association of multiple environmental stressors with biological responses, and quantification of uncertainty sources in the empirical water quality model. Nutrient data for lakes, ponds, and reservoirs across the United States were obtained from the Environmental Protection Agency (EPA) National Nutrient Criteria Database. The nutrient data consist of measurements for both stressor variables (such as total nitrogen and total phosphorus), and response variables (such as chlorophyll-a), used in the BTREED model. Markov chain Monte Carlo (McMC) posterior exploration guides a stochastic search through a rich suite of candidate trees toward models that better fit the data. The Bayes factor provides a goodness of fit criterion for comparison of resultant models. We randomly split the data into training and test sets; the training data were used in model estimation, and the test data were used to evaluate out-of-sample predictive performance of the model. An average relative efficiency of 1.02 between the training and test data for the four highest log-likelihood models suggests good out-of-sample predictive performance. Reduced model uncertainty relative to over-parameterized alternative models makes the BTREED models useful for nutrient criteria development, providing the link between nutrient stressors and meaningful eutrophication response.     

Modelling N mineralization from green manure and farmyard manure from a laboratory incubation study

Predicting N mineralization from organic manures like farmyard manure (FYM) is more difficult than from fresh organic materials like crop residues, as the manures vary greatly in composition. A laboratory incubation experiment was carried out for 98 days at 30 °C under aerobic conditions to study the effects on N dynamics of Gliricidia (Gliricidia sepium, Jacquin) and FYM application to soil at 5 and 10 g kg⁻¹. Application of Gliricidia induced N mineralization from the start of incubation period, with the amount of N mineralized increasing with rate of application. In contrast, application of FYM resulted in immobilization of mineral N in soil, irrespective of the rate of application. The initial net immobilization from FYM was limited by availability of N in the soil for the higher rate of application.We used the APSIM SoilN module to simulate these contrasting patterns of mineralization of N from Gliricidia and from FYM. The prediction of N mineralized from Gliricidia was better than FYM. The default model parameters specify that the fresh organic matter pools (FPOOL1, FPOOL2 and FPOOL3) have the same C:N ratio and this assumption was ineffective in predicting N mineralized from FYM. The predictive ability of the model improved when this default assumption was modified based on the size of the individual pools (FPOOL1, FPOOL2 and FPOOL3), and the pool's C:N ratios. The modelling efficiency, a measure of goodness of fit between the simulated and observed data, improved markedly for the modified model. The discrepancy between the modelled and observed data was a tendency for the model to underestimate the rate of re-mineralization at the lower rate of application of FYM in the later part of incubation. Unfortunately the appropriate modification to the size and C:N ratios of the FPOOLs could not be determined on the basis of chemical analysis alone. Thus, a true predictive application of the model to a new FYM material is not yet possible.     

Modelling species distributions with penalised logistic regressions: A comparison with maximum entropy models

An important aspect of species distribution modelling is the choice of the modelling method because a suboptimal method may have poor predictive performance. Previous comparisons have found that novel methods, such as Maxent models, outperform well-established modelling methods, such as the standard logistic regression. These comparisons used training samples with small numbers of occurrences per estimated model parameter, and this limited sample size may have caused poorer predictive performance due to overfitting. Our hypothesis is that Maxent models would outperform a standard logistic regression because Maxent models avoid overfitting by using regularisation techniques and a standard logistic regression does not. Regularisation can be applied to logistic regression models using penalised maximum likelihood estimation. This estimation procedure shrinks the regression coefficients towards zero, causing biased predictions if applied to the training sample but improving the accuracy of new predictions. We used Maxent and logistic regression (standard and penalised) to analyse presence/pseudo-absence data for 13 tree species and evaluated the predictive performance (discrimination) using presence-absence data. The penalised logistic regression outperformed standard logistic regression and equalled the performance of Maxent. The penalised logistic regression may be considered one of the best methods to develop species distribution models trained with presence/pseudo-absence data, as it is comparable to Maxent. Our results encourage further use of the penalised logistic regression for species distribution modelling, especially in those cases in which a complex model must be fitted to a sample with a limited size.     

Evaluation of discrete host–parasitoid models for diamondback moth and Diadegma semiclausum field time population density series

The performance of discrete mathematical models to describe the population dynamics of diamondback moth (DBM) (Plutella xylostella L.) and its parasitoid Diadegma semiclausum was investigated. The parameter values for several well-known models (Nicholson–Bailey, Hassell and Varley, Beddington, Free and Lawton, May, Holling type 2, 3 and Getz and Mills functional responses) were estimated. The models were tested on 20 consecutive sets of time series data collected at 14 days interval for pest and parasitoid populations obtained from a highland cabbage growing area in eastern Kenya. Model parameters were estimated from minimized squared difference between the numerical solution of the model equations and the empirical data using Powell's method. Maximum calculated DBM growth rates varied between 0.02 and 0.07. The carrying capacity determined at 16.5 DBM/plant by the Beddington et al. model was within the range of field data. However, all the estimated parameter values relating to the parasitoid, including the instantaneous searching rate (0.07–0.28), per capita searching efficiency (0.20–0.27), search time (5.20–5.33), handling time (0.77–0.90), and parasitism aggregation index (0.33), were well outside the range encountered empirically. All models evaluated for DBM under Durbin–Watson criteria, except the May model, were not autocorrelated with respect to residuals. In contrast, the criteria applied to the parasitoid residuals showed strong autocorrelations. Thus, these models failed to estimate parasitoid dynamics. We conclude that the interactions of the DBM with its parasitoid cannot be explained by any of the models tested. Two factors may be associated with this failure. First, the parasitoid in this integrated biological control system may not be playing a major role in regulating DBM population. Second, and perhaps more likely, poor correlations reflect gross inadequacies in the theoretical assumptions that underlie the existing models.     

Incorporating spatial autocorrelation into cellular automata model: An application to the dynamics of Chinese tamarisk (Tamarix chinensis Lour.)

Spatial autocorrelation (SAC) is frequently encountered in most spatial data in ecology. Cellular automata (CA) models have been widely used to simulate complex spatial phenomena. However, little has been done to examine the impact of incorporating SAC into CA models. Using image-derived maps of Chinese tamarisk (Tamarix chinensis Lour.), CA models based on ordinary logistic regression (OLCA model) and autologistic regression (ALCA model) were developed to simulate landscape dynamics of T. chinensis. In this study, significant positive SAC was detected in residuals of ordinary logistic models, whereas non-significant SAC was found in autologistic models. All autologistic models obtained lower Akaike's information criterion corrected for small sample size (AICc) values than the best ordinary logistic models. Although the performance of ALCA models only satisfied the minimum requirement, ALCA models showed considerable improvement upon OLCA models. Our results suggested that the incorporation of the autocovariate term not only accounted for SAC in model residuals but also provided more accurate estimates of regression coefficients. The study also found that the neglect of SAC might affect the statistical inference on underlying mechanisms driving landscape changes and obtain false ecological conclusions and management recommendations. The ALCA model is statistically sound when coping with spatially structured data, and the adoption of the ALCA model in future landscape transition simulations may provide more precise probability maps on landscape transition, better model performance and more reasonable mechanisms that are responsible for landscape changes.     

Multi-metric evaluation of the models WARM,CropSyst, and WOFOST for rice

WARM (Water Accounting Rice Model) simulates paddy rice (Oryza sativa L.), based on temperature-driven development and radiation-driven crop growth. It also simulates: biomass partitioning, floodwater effect on temperature, spikelet sterility, floodwater and chemicals management, and soil hydrology. Biomass estimates from WARM were evaluated and compared with the ones from two generic crop models (CropSyst, WOFOST). The test-area was the Po Valley (Italy). Data collected at six sites from 1989 to 2004 from rice crops grown under flooded and non-limiting conditions were split into a calibration (to estimate some model parameters) and a validation set. For model evaluation, a fuzzy-logic based multiple-metrics indicator (MQI) was used: 0 (best) ≤ MQI ≤ 1 (worst). WARM estimates compared well with the actual data (mean MQI = 0.037 against 0.167 and 0.173 with CropSyst and WOFOST, respectively). On an average, the three models performed similarly for individual validation metrics such as modelling efficiency (EF > 0.90) and correlation coefficient (R > 0.98). WARM performed best in a weighed measure of the Akaike Information Criterion: (worst) 0<wk<1$0<w_k<1$ (best), considering estimation accuracy and number of parameters required to achieve it (mean wk=0.983$w_k=0.983$ against 0.007 and ∼0.000 with CropSyst and WOFOST, respectively). WARM results were sensitive to 30% of the model parameters (ratio being lower with both CropSyst, <10%, and WOFOST, <20%), but appeared the easiest model to use because of the lowest number of crop parameters required (10 against 15 and 34 with CropSyst and WOFOST, respectively). This study provides a concrete example of the possibilities offered using a range of assessment metrics to evaluate model estimates, predictive capabilities, and complexity.     

Evaluating weather effects on interannual variation in net ecosystem productivity of a coastal temperate forest landscape: A model intercomparison

Forest productivity is strongly affected by seasonal weather patterns and by natural or anthropogenic disturbances. However weather effects on forest productivity are not currently represented in inventory-based models such as CBM-CFS3 used in national forest C accounting programs. To evaluate different approaches to modelling these effects, a model intercomparison was conducted among CBM-CFS3 and four process models (ecosys, CN-CLASS, Can-IBIS and 3PG) over a 2500 ha landscape in the Oyster River (OR) area of British Columbia, Canada. The process models used local weather data to simulate net primary productivity (NPP), net ecosystem productivity (NEP) and net biome productivity (NBP) from 1920 to 2005. Other inputs used by the process and inventory models were generated from soil, land cover and disturbance records. During a period of intense disturbance from 1928 to 1943, simulated NBP diverged considerably among the models. This divergence was attributed to differences among models in the sizes of detrital and humus C stocks in different soil layers to which a uniform set of soil C transformation coefficients was applied during disturbances. After the disturbance period, divergence in modelled NBP among models was much smaller, and attributed mainly to differences in simulated NPP caused by different approaches to modelling weather effects on productivity. In spite of these differences, age-detrended variation in annual NPP and NEP of closed canopy forest stands was negatively correlated with mean daily maximum air temperature during July-September (T_amax) in all process models (R² = 0.4-0.6), indicating that these correlations were robust. The negative correlation between T_amax and NEP was attributed to different processes in different models, which were tested by comparing CO₂ fluxes from these models with those measured by eddy covariance (EC) under contrasting air temperatures (T_a). The general agreement in sensitivity of annual NPP to T_amax among the process models led to the development of a generalized algorithm for weather effects on NPP of coastal temperate coniferous forests for use in inventory-based models such as CBM-CFS3: NPP′ = NPP − 57.1 (T_amax − 18.6), where NPP and NPP′ are the current and temperature-adjusted annual NPP estimates from the inventory-based model, 18.6 is the long-term mean daily maximum air temperature during July-September, and T_amax is the mean value for the current year. Our analysis indicated that the sensitivity of NPP to T_amax was nonlinear, so that this algorithm should not be extrapolated beyond the conditions of this study. However the process-based methodology to estimate weather effects on NPP and NEP developed in this study is widely applicable to other forest types and may be adopted for other inventory based forest carbon cycle models.     

Comparing species abundance models

Five regression models (Poisson, negative binomial, quasi-Poisson, the hurdle model and the zero-inflated Poisson) were used to assess the relationship between the abundance of a vulnerable plant species, Leionema ralstonii, and the environment. The methods differed in their capacity to deal with common properties of ecological data. They were assessed theoretically, and their predictive performance was evaluated with correlation, calibration and error statistics calculated within a bootstrap evaluation procedure that simulated performance for independent data.     

GIS-based modeling of the geographic distribution of Quercus emoryi Torr. (Fagaceae) in México and identification of significant environmental factors influencing the species’ distribution

The greatest concentration of oak species in the world is believed to be found in Mexico. These species are potentially useful for reforestation because of their capacity to adapt to diverse environments. Knowledge of their geographic distribution and of species–environment relations is essential for decision-making in the management and conservation of natural resources. The objectives of this study were to develop a model of the distribution of Quercus emoryi Torr. in Mexico, using geographic information systems and data layers of climatic and other variables, and to determine the variables that significantly influence the distribution of the species. The study consisted of the following steps: (A) selection of the target species from a botanical scientific collection, (B) characterization of the collecting sites using images with values or categories of the variables, (C) model building with the overlay of images that meet the habitat conditions determined from the characterization of sites, (D) model validation with independent data in order to determine the precision of the model, (E) model calibration through adjustment of the intervals of some variables, and (F) sensitivity analysis using precision and concordance non-parametric statistics applied to pairs of images. Results show that the intervals of the variables that best describe the species’ habitat are the following: altitude from 1650 to 2750 amsl, slope from 0 to 66°; average minimum temperature of January from −12 to −3 °C; mean temperature of June from 11 to 25 °C; mean annual precipitation from 218 to 1225 mm; soil units: lithosol, eutric cambisol, haplic phaeozem, chromic luvisol, rendzina, luvic xerosol, mollic planosol, pellic vertisol, eutric regosol; type of vegetation: oak forest, oak–pine forest, pine forest, pine–oak forest, juniperus forest, low open forest, natural grassland and chaparral. The resulting model of the geographic distribution of Quercus emoryi in Mexico had the following values for non-parametric statistics of precision and agreement: Kappa index of 0.613 and 0.788, overall accuracy of 0.806 and 0.894, sensitivity of 0.650 and 0.825, specificity of 0.963, positive predictive value of 0.945 and 0.957 and negative predictive value of 0.733 and 0.846. Results indicate that the variable average minimum temperature of January, with a maximum value of −3 °C, is an important factor in limiting the species’ distribution.     

Comparison of six correlative models in predictive vegetation mapping on a local scale

A variety of statistical techniques has been used in predictive vegetation modelling (PVM) that attempt to predict occurrence of a given community or species in respect to environmental conditions. We compared the performance of three profile models, BIOCLIM, GARP and MAXENT with three nonparametric models of group discrimination techniques, MARS, NPMR and LRT. The two latter models are relatively new statistical techniques that have just entered the field of PVM. We ran all models on a local scale for a given grassland community (Teucrio-Seslerietum) using the same input data to examine their performance. Model accuracy was evaluated both by Cohen’s kappa statistics (κ) and by area under receiver operating characteristics curve based both on resubstitution of training data and on an independent test data set. MAXENT of profile models and MARS of group discrimination techniques achieved the best prediction.