Comparison of complex modeling strategies for prediction of a binary outcome based on a few,highly correlated predictors

Motivated by a clinical prediction problem, a simulation study was performed to compare different approaches for building risk prediction models. Robust prediction models for hospital survival in patients with acute heart failure were to be derived from three highly correlated blood parameters measured up to four times, with predictive ability having explicit priority over interpretability. Methods that relied only on the original predictors were compared with methods using an expanded predictor space including transformations and interactions. Predictors were simulated as transformations and combinations of multivariate normal variables which were fitted to the partly skewed and bimodally distributed original data in such a way that the simulated data mimicked the original covariate structure. Different penalized versions of logistic regression as well as random forests and generalized additive models were investigated using classical logistic regression as a benchmark. Their performance was assessed based on measures of predictive accuracy, model discrimination, and model calibration. Three different scenarios using different subsets of the original data with different numbers of observations and events per variable were investigated. In the investigated setting, where a risk prediction model should be based on a small set of highly correlated and interconnected predictors, Elastic Net and also Ridge logistic regression showed good performance compared to their competitors, while other methods did not lead to substantial improvements or even performed worse than standard logistic regression. Our work demonstrates how simulation studies that mimic relevant features of a specific data set can support the choice of a good modeling strategy.     

Liana abundance and diversity increase with rainfall seasonality along a precipitation gradient in Panama

In tropical regions, rainfall gradients often explain the abundance and distribution of plant species. For example, many tree and liana species adapted to seasonal drought are more abundant and diverse in seasonally-dry forests, characterized by long periods of seasonal water deficit. Mean annual precipitation (MAP) is commonly used to explain plant distributions across climate gradients. However, the relationship between MAP and plant distribution is often weak, raising the question of whether other seasonal precipitation patterns better explain plant distributions in seasonally-dry forests. In this study, we examine the relationship between liana abundance and multiple metrics of seasonal and annual rainfall distribution to test the hypothesis that liana density and diversity increase with increasing seasonal drought along a rainfall gradient across the isthmus of Panama. We found that a normalized seasonality index, which combines MAP and the variability of monthly rainfall throughout the year, was a significant predictor of both liana density and species richness, whereas MAP, rainfall seasonality and the mean dry season precipitation (MDP) were far weaker predictors. The strong response of lianas to the normalized seasonality index indicates that, in addition to the total annual amount of rainfall, how rainfall is distributed throughout the year is an important determinant of the hydrological conditions that favor liana proliferation. Our findings imply that changes in annual rainfall and rainfall seasonality will determine the future distribution and abundance of lianas. Models that aim to predict future plant diversity, distribution, and abundance may need to move beyond MAP to a more detailed understanding of rainfall variability at sub-annual timescales.     

无粪便生活污水SMBR快速处理技术

针对传统生活污水水质相对复杂,处理与回用较为困难的问题,提出在收集时排除粪便水,选择厨房、洗漱、洗澡、洗涤的污水,利用浸没式膜生物反应器(SMBR)技术实现无粪便污水的快速处理,再生水回用于冲厕.结果表明:此种污水COD、NH4+-N和TP含量低,具有良好的可生化性,可大大降低处理周期与处理成本.本技术优化的主要工艺条件为:污泥浓度范围7000～9100 mg·L-1,污泥龄(SRT)40～55 d,水力停留时间(HRT)为80min,气水比为12～15,处理效果好,微滤膜对稳定出水水质起到重要作用.在此条件下,COD、NH4+-N和TP去除率分别为85.5%、53.1%和44.9%.出水COD在20～30 mg·L-1,BOD5为1～5 mg·L-1,NH4+-N为2～3.08 mg·L-1,TP为0.59～0.9 mg·L-1,LAS为0.41～0.67 mg·L-1,去除效果较好,再生水水质可达到<城市污水再生利用城市杂用水水质>(GB/T 18920-2002)标准.     

Relevance of using a vegetation-based method to conserve urban carabid diversity

Urban habitats harbour considerable biological diversity. Ecologists have developed methods to select which habitats should be conserved. The Ecological value, a method based on vegetation, has been created for the urban habitats of Montreal (Quebec, Canada). The main objective of our study was to determine if this method was relevant to assess carabid diversity of Mount Royal Park. This index is calculated using five criteria: uniqueness, representativeness, degree of succession, richness and rarity of the flora, each of which can influence communities of insects. Ground beetles (Coleoptera: Carabidae) were selected because of their success as biological indicators. Despite sampling difficulties, our results demonstrate clearly that the Ecological value method does not represent the high carabid richness of urban open habitats (tall grasses) and their specialised native carabid species. Within forests we found nonetheless that the ecological value index has a significant positive relationship with native carabid abundance. Moreover, maturity and structure of urban forests were positively correlated with carabid abundance and richness. Some urban vegetation characteristics have been shown to influence entomological diversity, but the relevance of using a global floral index to encompass the carabid community seems limited.     

Mapping continuous fields of forest alpha and beta diversity

Question: How to map continuous fields of forest alpha and beta diversity in remote areas, based on easily accessible spatial data. Location: Kyrgyzstan/Central Asia. Methods: The study relied on a combination of predictive mapping and remote sensing. Punctual measurements of alpha diversity were linked to topography and reflectance using regression models. For beta diversity, ordination techniques were employed to extract major vegetation gradients. Scores on the ordination axes were regressed against topography as well as reflectance and subsequently mapped. Beta diversity was mapped as spatial turnover rate along these axes. Results: The diversity maps quantified species counts and turnover in a spatially contiguous manner while taking into account fuzzy transitions. The variance explained by regression models ranged from 51% to 61% in cross‐validation. Many of the observed differences were caused by differences in species shares. The occurrence of walnut, in particular, showed a negative relation to woody species numbers. Conclusion: Mapping biodiversity in remote areas can be based on easily accessible spatial data in combination with a set of calibration field samples. With regard to human influence on walnut dominance, a total removal of human land use would be counterproductive in terms of diversity conservation. The results of this study highlight the need for comprehensive analyses of diversity patterns that include spatially contiguous quantifications of species numbers, shares and turnover rates.     

Species-specific responses to atmospheric carbon dioxide and tropospheric ozone mediate changes in soil carbon

We repeatedly sampled the surface mineral soil (0–20 cm depth) in three northern temperate forest communities over an 11-year experimental fumigation to understand the effects of elevated carbon dioxide (CO₂) and/or elevated phyto-toxic ozone (O₃) on soil carbon (C). After 11 years, there was no significant main effect of CO₂ or O₃ on soil C. However, within the community containing only aspen ( Populus tremuloides Michx.), elevated CO₂ caused a significant decrease in soil C content. Together with the observations of increased litter inputs, this result strongly suggests accelerated decomposition under elevated CO_2. In addition, an initial reduction in the formation of new (fumigation-derived) soil C by O₃ under elevated CO₂ proved to be only a temporary effect, mirroring trends in fine root biomass. Our results contradict predictions of increased soil C under elevated CO₂ and decreased soil C under elevated O₃ and should be considered in models simulating the effects of Earth's altered atmosphere.     

Prospects for tropical forest biodiversity in a human-modified world

The future of tropical forest biodiversity depends more than ever on the effective management of human-modified landscapes, presenting a daunting challenge to conservation practitioners and land use managers. We provide a critical synthesis of the scientific insights that guide our understanding of patterns and processes underpinning forest biodiversity in the human-modified tropics, and present a conceptual framework that integrates a broad range of social and ecological factors that define and contextualize the possible future of tropical forest species. A growing body of research demonstrates that spatial and temporal patterns of biodiversity are the dynamic product of interacting historical and contemporary human and ecological processes. These processes vary radically in their relative importance within and among regions, and have effects that may take years to become fully manifest. Interpreting biodiversity research findings is frequently made difficult by constrained study designs, low congruence in species responses to disturbance, shifting baselines and an over-dependence on comparative inferences from a small number of well studied localities. Spatial and temporal heterogeneity in the potential prospects for biodiversity conservation can be explained by regional differences in biotic vulnerability and anthropogenic legacies, an ever-tighter coupling of human-ecological systems and the influence of global environmental change. These differences provide both challenges and opportunities for biodiversity conservation. Building upon our synthesis we outline a simple adaptive-landscape planning framework that can help guide a new research agenda to enhance biodiversity conservation prospects in the human-modified tropics.     

Resilience of tropical rain forests: tree community reassembly in secondary forests

Understanding the recovery dynamics of ecosystems presents a major challenge in the human-impacted tropics. We tested whether secondary forests follow equilibrium or non-equilibrium dynamics by evaluating community reassembly over time, across different successional stages, and among multiple life stages. Based on long-term and static data from six 1-ha plots in NE Costa Rica, we show that secondary forests are undergoing reassembly of canopy tree and palm species composition through the successful recruitment of seedlings, saplings, and young trees of mature forest species. Such patterns were observed over time within sites and across successional stages. Floristic reassembly in secondary forests showed a clear convergence with mature forest community composition, supporting an equilibrium model. This resilience stems from three key factors co-occurring locally: high abundance of generalist species in the regional flora, high levels of seed dispersal, and local presence of old-growth forest remnants.     

Mechanical Restoration of California Mixed‐Conifer Forests: Does it Matter Which Trees Are Cut?

The montane ecosystems of northern California have been subjected to repeated manipulation and active fire suppression for over a century, resulting in changes in community structure that contribute to increased wildfire hazard. Ecosystem restoration via reduction of stand density for wildfire hazard mitigation has received substantial attention in recent years; however, many ecological questions remain unanswered. This study compares belowground effects of two alternative forest thinning treatments designed to restore the large, old tree component of late‐seral structure, one of which focuses on restoring Pinus ponderosa dominance (Pine‐preference) and the other of which promotes development of large trees regardless of species (Size‐preference). We evaluated forest floor and soil chemical and microbial parameters in six experimental thinning treatment units of 40 ha each in the Klamath National Forest of northern California 5–6 years after thinning. Inorganic N availability, soil organic C content, phenol oxidase activity, and forest floor C:N ratio were greater in the Size‐preference treatment, whereas forest floor N and soil pH were greater in the Pine‐preference treatment. Our results indicate that these two thinning strategies produce differences in the soil environment that has the potential to affect growth rates of trees that remain, as well as the growth and survivorship of newly established seedlings. Thus, which species/individuals are removed during structural restoration of these mixed‐conifer forests matters both to the belowground components of the ecosystem today and the vegetation and productivity of the ecosystem in future decades.     

Soil organic matter dynamics during 80 years of reforestation of tropical pastures

Our research takes advantage of a historical trend in natural reforestation of abandoned tropical pastures to examine changes in soil carbon (C) during 80 years of secondary forest regrowth. We combined a chronosequence approach with differences in the natural abundance of ¹³C between C3 (forest) and C4 (pasture) plants to estimate turnover times of C in the bulk soil and in density fractions. Overall, gains in secondary forest C were compensated for by the loss of residual pasture-derived soil C, resulting in no net change in bulk soil C stocks down to 1 m depth over the chronosequence. The free light fraction (LF), representing physically unprotected particulate organic matter, was most sensitive to land-use change. Reforestation replenished C in the free LF that had been depleted during conversion to pastures. Turnover times varied with model choice, but in general, soil C cycling rates were rapid for the 0–10 cm depth, with even the heavy fraction (HF) containing C cycling in decadal time scales. Turnover times of C in the free LF from the 0–10 cm depth were shorter than for the occluded and HFs, highlighting the importance of physical location in the soil matrix for residence time in the soil. The majority of the soil C pool (82±21%) was recovered in the mineral-associated density fraction. Carbon-to-nitrogen ratios and differences in natural abundance ¹⁵N of soil organic matter (SOM) showed an increasing degree of decomposition across density fractions with increasing mineral association. Our data show that the physical distribution of C in the soil has a large impact on soil C turnover and the ability of soils to maintain SOM stocks during land-use and land-cover change.