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.     

生根粉对梭梭苗木根系生长及成活的影响

在自然条件下,采用随机区组设计研究了生根粉(ABT3号)4个处理(25、50、100、200 mg L-1)对人工种植的梭梭幼苗细根动态、年生长终期根系形态特征以及成活率的影响.结果表明:1)梭梭幼苗一年中细根生长有2次高峰,峰值分别出现在6月和9～10月.经生根粉处理后,梭梭幼苗细根的生长动态与对照基本一致,但在具体月份生根粉明显增加了细根总长度、细根生长速率和细根数量密度.2)ABT3号生根粉可以使梭梭幼苗的存活率达到50%以上,显著高于对照的34.75%,相关分析表明生根粉提高梭梭幼苗成活率是通过增加根系生长来实现的.3)由主成分分析可知,50 mg L-1的生根粉处理作用效果最理想.因此,建议在梭梭种植过程中可使用50 mg L-1的生根粉提高其成活率;在本地,肉苁蓉接种的最佳季节是6月.图2表3参31     

Modeling freshwater fish distributions using multiscale landscape data: A case study of six narrow range endemics

Species distribution models (SDMs) have become integral tools in scientific research and conservation planning. Despite progress in the assessment of various statistical models for use in SDMs, little has been done in way of evaluating appropriate ecological models. In this paper, we evaluate the multiscale filter framework as a suitable theoretical model for predicting freshwater fish distributions in the upper Green River system (Ohio River drainage), USA. The spatial distributions of six fishes with contrasting biogeographies were modeled using boosted regression trees and multiscale landscape data. Species biogeography did not appear to affect predictive performance and all models performed well statistically with receiver operating characteristic area under the curve (AUC) ranging from 0.87 to 0.98. Predictive maps show accurate estimations of ranges for five of six species based on historical collections. The relative influence of each type of environmental feature and spatial scale varied markedly with between species. A hierarchical effect was detected for narrowly distributed species. These species were highly influenced by soil composition at larger spatial scales and land use/land cover (LULC) patterns at more proximal scales. Conversely, LULC pattern was the most influential feature for widely distributed at all spatial scales. Using multiscale data capable of capturing hierarchical landscape influences allowed production of accurate predictive models and provided further insight into factors controlling freshwater fish distributions.     

Multistate models for monitoring individual trees in permanent observation plots

This interdisciplinary research on forest ecosystems begins with some characteristics of ecosystems which are the basis for the derivation of statistical models for the development and vitality of trees. Several ecological problems which could be solved by longitudinal studies are mentioned. Statistical methods for the evaluation of the crowns of spruce trees (Picea abies Karst) in three permanent observation plots in Switzerland are described. In particular, the time-dependent proportional odds model and a transitional model are used. Through application of these multistate models the data give information on the dependence of an ordered categorical response variable on covariates characterizing the ecosystem. The response variable is observed through infrared aerial photographs. This monitoring system gives insight into the dynamic behaviour of the forest ecosystem. The need for more eco-systematically motivated statistical research using longitudinal studies is identified.     

Using information theory as a substitute for stepwise regression in ecology and behavior

In ecological and behavioral research, drawing reliable conclusions from statistical models with multiple predictors is usually difficult if all predictors are simultaneously in the model. The traditional way of handling multiple predictors has been the use of threshold-based removal-introduction algorithms, that is, stepwise regression, which currently receives considerable criticism. A more recent and increasingly propagated modelling method for multiple predictors is the information theoretic (IT) approach that quantifies the relative suitability of multiple, potentially non-nested models based on a balance of model fit and the accuracy of estimates. Here, we examine three shortcomings of stepwise regression, subjective critical values, model uncertainty, and parameter estimation bias, which have been suggested to be avoided by applying information theory. We argue that, in certain circumstances, the IT approach may be sensitive to these issues as well. We point to areas where further testing and development could enhance the performance of IT methods and ultimately lead to robust inferences in behavioral ecology.     

辽西大凌河流域土地利用变化及驱动力分析

从政策、流域综合治理、经济发展和技术进步、农民认知态度等4方面对影响大凌河流域土地利用变化的驱动力进行了分析。同时运用农户问卷调查和驱动力分析结果,选取影响耕地变化的社会经济和人口因子,运用主成分分析和多元迭代回归分析确定影响耕地变化的主要因子,并拟合出耕地变化的最优度模型。研究结果表明:在1987—2002年期间,农田和未利用荒地面积在不断减小,而林地、果园、草地在不断增加,但1995年后变化边际度大大减小;主成分分析表明影响土地利用变化主要影响因子是农业人口(A-POP)、总人口(T-POP)、农村经济收入(TIRE)、农林牧渔收入(IAFAF)和第三产业总产值(GTI);多元迭代回归分析表明耕地面积变化的最优回归模型中主变量是农业人口(A-POP)、总人口(T-POP)、农村经济收入(TIRE),这些变量能够解释95.1%的耕地变化。     

Functional regression on remote sensing data in oceanography

The aim of this study is to propose the use of a functional data analysis approach as an alternative to the classical statistical methods most commonly used in oceanography and water quality management. In particular we consider the prediction of total suspended solids (TSS) based on remote sensing (RS) data. For this purpose several functional linear regression models and classical non-functional regression models are applied to 10 years of RS data obtained from medium resolution imaging spectrometer sensor to predict the TSS concentration in the coastal zone of the Guadalquivir estuary. The results of functional and classical approaches are compared in terms of their mean square prediction error values and the superiority of the functional models is established. A simulation study has been designed in order to support these findings and to determine the best prediction model for the TSS parameter in more general contexts.     

Statistical treatment for the wet bias in tree-ring chronologies: a case study from the Interior West,USA

Dendroclimatic research has long assumed a linear relationship between tree-ring increment and climate variables. However, ring width frequently underestimates extremely wet years, a phenomenon we refer to as ‘wet bias’. In this paper, we present statistical evidence for wet bias that is obscured by the assumption of linearity. To improve tree-ring-climate modeling, we take into account wet bias by introducing two modified linear regression models: a linear spline regression (LSR) and a likelihood-based wet bias adjusted linear regression (WBALR), in comparison with a quadratic regression (QR) model. Using gridded precipitation data and tree-ring indices of multiple species from various sites in Utah, both LSR and WBALR show a significant improvement over the linear regression model and out-perform QR in terms of in-sample \({R}^{2}\) and out-of-sample MSE. This further shows that the wet bias emerges from nonlinearity of tree-ring chronologies in reconstructing precipitation. The pattern and extent of wet bias varies by species, by site, and by precipitation regime, making it difficult to generalize the mechanisms behind its cause. However, it is likely that dis-coupling between precipitation amounts (e.g., percent received as rain/snow or percent infiltrating the soil) and its availability to trees (e.g., root zone dynamics), is the primary mechanism driving wet bias.     

Prediction of non-methane hydrocarbons in Kuwait using regression and Bayesian kriged Kalman model

This article describes the hierarchical Bayesian approach for predicting average hourly concentrations of ambient non-methane hydrocarbons (NMHC) in Kuwait where records of six monitor stations located in different sites are observed at successive time points. Our objective is to predict the concentration level of NMHC in unmonitored areas. Here an attempt is made for the prediction of unmeasured concentration of NMHC at two additional locations in Kuwait. We will implement a kriged Kalman filter (KKF) hierarchical Bayesian approach assuming a Gaussian random field, a technique that allows the pooling of data from different sites in order to predict the exposure of the NMHC in different regions of Kuwait. In order to increase the accuracy of the KKF we will use other statistical models such as imputation, regression, principal components, and time series analysis in our approach. We considered four different types of imputation techniques to address the missing data. At the primary level, the logarithmic field is modeled as a trend plus Gaussian stochastic residual model. The trend model depends on hourly meteorological predictors which are common to all sites. The residuals are then modeled using KKF, and the prediction equation is derived conditioned on adjoining hours. On this basis we developed a spatial predictive distribution for these residuals at unmonitored sites. By transforming the predicted residuals back to the original data scales, we can impute Kuwait’s hourly non-methane hydrocarbons field.     

Information-theoretic approaches to statistical analysis in behavioural ecology: an introduction

Scientific thinking may require the consideration of multiple hypotheses, which often call for complex statistical models at the level of data analysis. The aim of this introduction is to provide a brief overview on how competing hypotheses are evaluated statistically in behavioural ecological studies and to offer potentially fruitful avenues for future methodological developments. Complex models have traditionally been treated by model selection approaches using threshold-based removal of terms, i.e. stepwise selection. A recently introduced method for model selection applies an information-theoretic (IT) approach, which simultaneously evaluates hypotheses by balancing between model complexity and goodness of fit. The IT method has been increasingly propagated in the field of ecology, while a literature survey shows that its spread in behavioural ecology has been much slower, and model simplification using stepwise selection is still more widespread than IT-based model selection. Why has the use of IT methods in behavioural ecology lagged behind other disciplines? This special issue examines the suitability of the IT method for analysing data with multiple predictors, which researchers encounter in our field. The volume brings together different viewpoints to aid behavioural ecologists in understanding the method, with the hope of enhancing the statistical integration of our discipline.     

The roles of climate, phylogenetic relatedness, introduction effort, and reproductive traits in the establishment of non-native reptiles and amphibians

We developed a method to predict the potential of non-native reptiles and amphibians (herpetofauna) to establish populations. This method may inform efforts to prevent the introduction of invasive non-native species. We used boosted regression trees to determine whether nine variables influence establishment success of introduced herpetofauna in California and Florida. We used an independent data set to assess model performance. Propagule pressure was the variable most strongly associated with establishment success. Species with short juvenile periods and species with phylogenetically more distant relatives in regional biotas were more likely to establish than species that start breeding later and those that have close relatives. Average climate match (the similarity of climate between native and non-native range) and life form were also important. Frogs and lizards were the taxonomic groups most likely to establish, whereas a much lower proportion of snakes and turtles established. We used results from our best model to compile a spreadsheet-based model for easy use and interpretation. Probability scores obtained from the spreadsheet model were strongly correlated with establishment success as were probabilities predicted for independent data by the boosted regression tree model. However, the error rate for predictions made with independent data was much higher than with cross validation using training data. This difference in predictive power does not preclude use of the model to assess the probability of establishment of herpetofauna because (1) the independent data had no information for two variables (meaning the full predictive capacity of the model could not be realized) and (2) the model structure is consistent with the recent literature on the primary determinants of establishment success for herpetofauna. It may still be difficult to predict the establishment probability of poorly studied taxa, but it is clear that non-native species (especially lizards and frogs) that mature early and come from environments similar to that of the introduction region have the highest probability of establishment.     

Analysis and prediction of $$\mathrm{PM}_{10}$$ concentration levels in Tunisia using statistical learning approaches

Over the past years, the health impact of airborne particulate matter \(\mathrm{PM}_{10}\) has become a very topical subject. Thereby, a lot of research effort in the environmental sciences goes towards the modeling and the prediction of ambient \(\mathrm{PM}_{10}\) concentrations. In this paper, we are interested in the statistical classification of the daily mean \(\mathrm{PM}_{10}\) concentration in Tunisia according to the authority regulation. We consider two monitoring stations: a big industrial station and a traffic station. The main goal of this work is to determine the pertinent predictors of \(\mathrm{PM}_{10}\) concentration within a nonlinear multiclass framework. To do this, we used two popular statistical learning methods; the support vector machines (SVM) and the random forests (RF). The statistical results obtained on the real datasets, show that RF outperform SVM for the purpose of variable selection even with a reduced number of observations compared to the number of explicative variables. It was also demonstrated that the \(\mathrm{PM}_{10}\) concentration measured yesterday is the most relevant predictor of its present-day value. Moreover, we found that the more delayed values of \(\mathrm{PM}_{10}\) concentration may be crucial to get an accurate prediction.     

Building statistical models to analyze species distributions.

Models of the geographic distributions of species have wide application in ecology. But the nonspatial, single-level, regression models that ecologists have often employed do not deal with problems of irregular sampling intensity or spatial dependence, and do not adequately quantify uncertainty. We show here how to build statistical models that can handle these features of spatial prediction and provide richer, more powerful inference about species niche relations, distributions, and the effects of human disturbance. We begin with a familiar generalized linear model and build in additional features, including spatial random effects and hierarchical levels. Since these models are fully specified statistical models, we show that it is possible to add complexity without sacrificing interpretability. This step-by-step approach, together with attached code that implements a simple, spatially explicit, regression model, is structured to facilitate self-teaching. All models are developed in a Bayesian framework. We assess the performance of the models by using them to predict the distributions of two plant species (Proteaceae) from South Africa's Cape Floristic Region. We demonstrate that making distribution models spatially explicit can be essential for accurately characterizing the environmental response of species, predicting their probability of occurrence, and assessing uncertainty in the model results. Adding hierarchical levels to the models has further advantages in allowing human transformation of the landscape to be taken into account, as well as additional features of the sampling process.     

Utility of Dynamic-Landscape Metapopulation Models for Sustainable Forest Management

Abstract:  We evaluated the utility of combining metapopulation models with landscape-level forest-dynamics models to assess the sustainability of forest management practices. We used the Brown Creeper ( Certhia americana ) in the boreal forests of northern Ontario as a case study. We selected the Brown Creeper as a potential indicator of sustainability because it is relatively common in the region but is dependent on snags and old trees for nesting and foraging; hence, it may be sensitive to timber harvesting. For the modeling we used RAMAS Landscape, a software package that integrates RAMAS GIS, population-modeling software, and LANDIS, forest-dynamics modeling software. Predictions about the future floristic composition and structure of the landscape under a variety of management and natural disturbance scenarios were derived using LANDIS. We modeled eight alternative forest management scenarios, ranging in intensity from no timber harvesting and a natural fire regime to intensive timber harvesting with salvage logging after fire. We predicted the response of the Brown Creeper metapopulation over a 160-year period and used future population size and expected minimum population size to compare the sustainability of the various management scenarios. The modeling methods were easy to apply and model predictions were sensitive to the differences among management scenarios, indicating that these methods may be useful for assessing and ranking the sustainability of forest management options. Primary concerns about the method are the practical difficulties associated with incorporating fire stochasticity in prediction uncertainty and the number of model assumptions that must be made and tested with sensitivity analysis. We wrote new software to help quantify the contribution of landscape stochasticity to model prediction uncertainty.     

The importance of spatial autocorrelation,extent and resolution in predicting forest bird occurrence

Concerns about declines in forest biodiversity underscore the need for accurate estimates of the distribution and abundance of organisms at large scales and at resolutions that are fine enough to be appropriate for management. This paper addresses three major objectives: (i) to determine whether the resolution of typical air photo-derived forest inventory is sufficient for the accurate prediction of site occupancy by forest birds. We compared prediction success of habitat models using air photo variables to models with variables derived from finer resolution, ground-sampled vegetation plots. (ii) To test whether incorporating spatial autocorrelation into habitat models via autologistic regression increases prediction success. (iii) To determine whether landscape structure is an important factor in predicting bird distribution in forest-dominated landscapes. Models were tested locally (Greater Fundy Ecosystem [GFE]) using cross-validation, and regionally using an independent data set from an area located ca. 250 km to the northwest (Riley Brook [RB]). We found significant positive spatial autocorrelation in the residuals of at least one habitat model for 76% (16/21) of species examined. In these cases, the logistic regression assumption of spatially independent errors was violated. Logistic models that ignored spatial autocorrelation tended to overestimate habitat effects. Though overall prediction success was higher for autologistic models than logistic models in the GFE, the difference was only significantly improved for one species. Further, the inclusion of spatial covariates did little to improve model performance in the geographically discrete study area. For 62% (13/21) of species examined, landscape variables were significant predictors of forest bird occurrence even after statistically controlling for stand-level variability. However, broad spatial extents explained less variation than local factors. In the GFE, 76% (16/21) of air photo and 81% (17/21) of ground plot models were accurate enough to be of practical utility (AUC > 0.7). When applied to RB, both model types performed effectively for 55% (11/20) of the species examined. We did not detect an overall difference in prediction success between air photo and ground plot models in either study area. We conclude that air photo data are as effective as fine resolution vegetation data for predicting site occupancy for the majority of species in this study. These models will be of use to forest managers who are interested in mapping species distributions under various timber harvest scenarios, and to protected areas planners attempting to optimize reserve function.     

Two sides of the same coin? Rare and pest plants native to the United States and Canada

Plant biodiversity is at risk, with as many as 10% of native species in the United States being threatened with extinction. Habitat loss has led a growing number of plant species to become rare or threatened, while the introduction or expansion of pest species has led some habitats to be dominated by relatively few, mostly nonindigenous, species. As humans continue to alter many landscapes and vegetation types, understanding how biological traits determine the location of species along a spectrum from vulnerability to pest status is critical to designing risk assessment protocols, setting conservation priorities, and developing monitoring programs. We used boosted regression trees to predict rarity (based on The Nature Conservancy global rankings) and pest status (defined as legal pest status) from data on traits for the native vascular flora of the United States and Canada including Hawaii, Puerto Rico, and the Virgin Islands (n approximately = 15,000). Categories were moderately to highly predictable (AUCpest = 0.87 on 25% holdout test set, AUCrarity = 0.80 on 25% holdout test set). Key predictors were chromosome number, ploidy, seed mass, and a suite of traits suggestive of specialist vs. generalist adaptations (e.g., facultative wetland habitat association and phenotypic variability in growth form and life history). Specifically, pests were associated with high chromosome numbers, polyploidy, and seed masses ranging from 0.1 to 100 mg, whereas rare species were associated with low chromosome numbers, low ploidy, and large (>1000 mg) seed masses. In addition, pest species were disproportionately likely to be facultatively associated with wetlands, and variable in growth form and life history, whereas rare species exhibited an opposite pattern. These results suggest that rare and pest species contrast along trait axes related to dispersal and performance in disturbed or novel habitats.     

Sparse regression for analyzing the development of foliar nutrient concentrations in coniferous trees

Analyzing and predicting the development of foliar nutrient concentrations are important and challenging tasks in environmental monitoring. This article presents how linear sparse regression models can be used to represent the relations between different foliar nutrient concentration measurements of coniferous trees in consecutive years. In the experiments the models proved to be capable of providing relatively good and reliable predictions of the development of foliage with a considerably small number of regressors. Two methods for estimating sparse models were compared to more conventional linear regression models. Differences in the prediction accuracies between the sparse and full models were minor, but the sparse models were found to highlight important dependencies between the nutrient measurements better than the other regression models. The use of sparse models is, therefore, advantageous in the analysis and interpretation of the development of foliar nutrient concentrations.