Modeling vegetation greenness and its climate sensitivity with deep‐learning technology

Climate sensitivity of vegetation has long been explored using statistical or process‐based models. However, great uncertainties still remain due to the methodologies’ deficiency in capturing the complex interactions between climate and vegetation. Here, we developed global gridded climate–vegetation models based on long short‐term memory (LSTM) network, which is a powerful deep‐learning algorithm for long‐time series modeling, to achieve accurate vegetation monitoring and investigate the complex relationship between climate and vegetation. We selected the normalized difference vegetation index (NDVI) that represents vegetation greenness as model outputs. The climate data (monthly temperature and precipitation) were used as inputs. We trained the networks with data from 1982 to 2003, and the data from 2004 to 2015 were used to validate the models. Error analysis and sensitivity analysis were performed to assess the model errors and investigate the sensitivity of global vegetation to climate change. Results show that models based on deep learning are very effective in simulating and predicting the vegetation greenness dynamics. For models training, the root mean square error (RMSE) is <0.01. Model validation also assure the accuracy of our models. Furthermore, sensitivity analysis of models revealed a spatial pattern of global vegetation to climate, which provides us a new way to investigate the climate sensitivity of vegetation. Our study suggests that it is a good way to integrate deep‐learning method to monitor the vegetation change under global change. In the future, we can explore more complex climatic and ecological systems with deep learning and coupling with certain physical process to better understand the nature.     

Implementation of a hierarchical global vegetation classification in ecosystem function models

Abstract. We propose an alternative approach for the currently used biogeographic global vegetation classifications. A hierarchical vegetation classification system is proposed for consistent and routine monitoring of global vegetation. Global vegetation is first defined into six classes based on plant canopy structure and dynamics observable by remote sensing from satellites. Additional biome variability is then represented through a remote sensing derived leaf area index map, and direct climate data sets driving an ecosystem model to compute and map net primary production and evapotranspiration. Simulation results from an ecosystem function model suggest that the six canopy structure-based classes are sufficient to represent global variability in these parameters, provided the spatio-temporal variations in Leaf Area Index and climate are characterized accurately. If a bioclimatically based classification is needed for other purposes, our six class approach can be expanded to a possible 21 classes using archived climatic zones. For example, tropical, subtropical, temperate and boreal labels are defined by absolute minimum temperature. Further separation in each class is possible through changes in water availability defined by precipitation and/or soils. The resulting vegetation classes correspond to many of the existing, conventional global vegetation schemes, yet retain the measure of actual vegetation possible because remote sensing first defines the six biome classes in our classification. Vegetation classifications are no longer an end product but a source of initializing data for global ecosystem function models. Remote sensing with biosphere models directly calculates the ecological functions previously inferred from vegetation classifications, but with higher spatial and temporal accuracy.     

Partially supervised learning using an EM-boosting algorithm

Training data in a supervised learning problem consist of the class label and its potential predictors for a set of observations. Constructing effective classifiers from training data is the goal of supervised learning. In biomedical sciences and other scientific applications, class labels may be subject to errors. We consider a setting where there are two classes but observations with labels corresponding to one of the classes may in fact be mislabeled. The application concerns the use of protein mass-spectrometry data to discriminate between serum samples from cancer and noncancer patients. The patients in the training set are classified on the basis of tissue biopsy. Although biopsy is 100% specific in the sense that a tissue that shows itself to have malignant cells is certainly cancer, it is less than 100% sensitive. Reference gold standards that are subject to this special type of misclassification due to imperfect diagnosis certainty arise in many fields. We consider the development of a supervised learning algorithm under these conditions and refer to it as partially supervised learning. Boosting is a supervised learning algorithm geared toward high-dimensional predictor data, such as those generated in protein mass-spectrometry. We propose a modification of the boosting algorithm for partially supervised learning. The proposal is to view the true class membership of the samples that are labeled with the error-prone class label as missing data, and apply an algorithm related to the EM algorithm for minimization of a loss function. To assess the usefulness of the proposed method, we artificially mislabeled a subset of samples and applied the original and EM-modified boosting (EM-Boost) algorithms for comparison. Notable improvements in misclassification rates are observed with EM-Boost.     

Distance learning techniques for ontology similarity measuring and ontology mapping

Recent years, a large amount of ontology learning algorithms have been applied in different disciplines and engineering. The ontology model is presented as a graph and the key of ontology algorithms is similarity measuring between concepts. In the learning frameworks, the information of each ontology vertex is expressed as a vector, thus the similarity measuring can be determined via the distance of the corresponding vector. In this paper, we study how to get an optimal distance function in the ontology setting. The tricks we presented are divided into two parts: first, the ontology distance learning technology in the setting that the ontology data have no labels; then, the distance learning approaches in the setting that the given ontology data are carrying real numbers as their labels. The result data of the four simulation experiments reveal that our new ontology trick has high efficiency and accuracy in ontology similarity measure and ontology mapping in special engineering applications.     

Estimating plant abundance using inflated beta distributions: Applied learnings from a lichen–caribou ecosystem

Quantifying abundance and distribution of plant species can be difficult because data are often inflated with zero values due to rarity or absence from many ecosystems. Terrestrial fruticose lichens (Cladonia and Cetraria spp.) occupy a narrow ecological niche and have been linked to the diets of declining caribou and reindeer populations (Rangifer tarandus) across their global distribution, and conditions related to their abundance and distribution are not well understood. We attempted to measure effects related to the occupancy and abundance of terrestrial fruticose lichens by sampling and simultaneously modeling two discrete conditions: absence and abundance. We sampled the proportion cover of terrestrial lichens at 438 vegetation plots, including 98 plots having zero lichens. A zero‐inflated beta regression model was employed to simultaneously estimate both the absence and the proportion cover of terrestrial fruticose lichens using fine resolution satellite imagery and light detection and ranging (LiDAR) derived covariates. The probability of lichen absence significantly increased with shallower groundwater, taller vegetation, and increased Sphagnum moss cover. Vegetation productivity, Sphagnum moss cover, and seasonal changes in photosynthetic capacity were negatively related to the abundances of terrestrial lichens. Inflated beta regression reliably estimated the abundance of terrestrial lichens (R² = .74) which was interpolated on a map at fine resolution across a caribou range to support ecological conservation and reclamation. Results demonstrate that sampling for and simultaneously estimating both occupancy and abundance offer a powerful approach to improve statistical estimation and expand ecological inference in an applied setting. Learnings are broadly applicable to studying species that are rare, occupy narrow niches, or where the response variable is a proportion value containing zero or one, which is typical of vegetation cover data.     

Detecting remote evolutionary relationships among proteins by large-scale semantic embedding

Virtually every molecular biologist has searched a protein or DNA sequence database to find sequences that are evolutionarily related to a given query. Pairwise sequence comparison methods--i.e., measures of similarity between query and target sequences--provide the engine for sequence database search and have been the subject of 30 years of computational research. For the difficult problem of detecting remote evolutionary relationships between protein sequences, the most successful pairwise comparison methods involve building local models (e.g., profile hidden Markov models) of protein sequences. However, recent work in massive data domains like web search and natural language processing demonstrate the advantage of exploiting the global structure of the data space. Motivated by this work, we present a large-scale algorithm called ProtEmbed, which learns an embedding of protein sequences into a low-dimensional "semantic space." Evolutionarily related proteins are embedded in close proximity, and additional pieces of evidence, such as 3D structural similarity or class labels, can be incorporated into the learning process. We find that ProtEmbed achieves superior accuracy to widely used pairwise sequence methods like PSI-BLAST and HHSearch for remote homology detection; it also outperforms our previous RankProp algorithm, which incorporates global structure in the form of a protein similarity network. Finally, the ProtEmbed embedding space can be visualized, both at the global level and local to a given query, yielding intuition about the structure of protein sequence space.     

Statistical models of invertebrate distribution on Macquarie Island: a tool to assess climate change and local human impacts

Sub-Antarctic islands are good model systems in which to study the ecological effects of human impacts, particularly global climate change and alien species. Invertebrates form a central component of these ecosystems. We conducted a stratified survey of 69 sites on sub-Antarctic Macquarie Island and used logistic regression models to describe the distribution of 14 abundant invertebrate species. We also developed a statistical model of windspeed based on topography. The distributions of individual species were described by different combinations of aspect, altitude and vegetation type. Ordination of sites based on species composition showed strong effects of altitude and vegetation on invertebrate assemblages. The species distribution models provide a tool for detecting, monitoring and predicting effects of climate change and alien species on biota and ecosystem processes. Accepted: 30 August 1998     

GVC-Net: Global Vascular Context Network for Cerebrovascular Segmentation Using Sparse Labels

ObjectivesCerebrovascular disease is a serious threat to human health. Because of its high mortality and disability rate, early diagnosis and prevention are very important. The performance of existing cerebrovascular segmentation methods based on deep learning depends on the integrity of labels. However, manual labels are usually of low quality and poor connectivity at small blood vessels, which directly affects the cerebrovascular segmentation results.Material and methodIn this paper, we propose a new segmentation network to segment cerebral vessels from MRA images by using sparse labels. The long-distance dependence between vascular structures is captured by the global vascular context module, and the topology is constrained by the hybrid loss function to segment the cerebral vessels with good connectivity.ResultExperiments show that our method performed with a sensitivity, precision, dice similarity coefficient, intersection over union and centerline dice similarity coefficient of 61.24%, 75.58%, 67.66%, 51.13% and 83.79% respectively.ConclusionThe obtained results reveal that the proposed cerebrovascular segmentation network has better segmentation performance for cerebrovascular segmentation under sparse labels, and can suppress the noise of background to a certain extent.     

Analysis of mouse EphA knockins and knockouts suggests that retinal axons programme target cells to form ordered retinotopic maps

I present a novel analysis of abnormal retinocollicular maps in mice in which the distribution of EphA receptors over the retina has been modified by knockin and/or knockout of these receptor types. My analysis shows that in all these cases, whereas the maps themselves are discontinuous, the graded distribution of EphA over the nasotemporal axis of the retina is recreated within the pattern of axonal terminations across rostrocaudal colliculus. This suggests that the guiding principle behind the formation of ordered maps of nerve connections between vertebrate retina and superior colliculus, or optic tectum, is that axons carrying similar amounts of Eph receptor terminate near to one another on the target structure. I show how the previously proposed marker induction model embodies this principle and predicts these results. I then describe a new version of the model in which the properties of the markers, or labels, are based on those of the Eph receptors and their associated ligands, the ephrins. I present new simulation results, showing the development of maps between two-dimensional structures, exploring the role of counter-gradients of labels across the target and confirming that the model reproduces the retinocollicular maps found in EphA knockin/knockout mice. I predict that abnormal distributions of label within the retina lead to abnormal distributions of label over the target, so that in each of the types of knockin/knockout mice analysed, there will be a different distribution of labels over the target structure. This mechanism could be responsible for the flexibility with which neurons reorganise their connections during development and the degree of precision in the final map. Activity-based mechanisms would play a role only at a later stage of development to remove the overlap between individual retinal projection fields, such as in the development of patterns of ocular dominance stripes.     

Deforestation and Reforestation of Latin America and the Caribbean (2001–2010)

Forest cover change directly affects biodiversity, the global carbon budget, and ecosystem function. Within Latin American and the Caribbean region (LAC), many studies have documented extensive deforestation, but there are also many local studies reporting forest recovery. These contrasting dynamics have been largely attributed to demographic and socio‐economic change. For example, local population change due to migration can stimulate forest recovery, while the increasing global demand for food can drive agriculture expansion. However, as no analysis has simultaneously evaluated deforestation and reforestation from the municipal to continental scale, we lack a comprehensive assessment of the spatial distribution of these processes. We overcame this limitation by producing wall‐to‐wall, annual maps of change in woody vegetation and other land‐cover classes between 2001 and 2010 for each of the 16,050 municipalities in LAC, and we used nonparametric Random Forest regression analyses to determine which environmental or population variables best explained the variation in woody vegetation change. Woody vegetation change was dominated by deforestation (?541,835 km²), particularly in the moist forest, dry forest, and savannas/shrublands biomes in South America. Extensive areas also recovered woody vegetation (+362,430 km²), particularly in regions too dry or too steep for modern agriculture. Deforestation in moist forests tended to occur in lowland areas with low population density, but woody cover change was not related to municipality‐scale population change. These results emphasize the importance of quantitating deforestation and reforestation at multiple spatial scales and linking these changes with global drivers such as the global demand for food.     

A modified sign test for comparing paired ROC curves

We develop a permutation test for assessing a difference in the areas under the curve (AUCs) in a paired setting where both modalities are given to each diseased and nondiseased subject. We propose that permutations be made between subjects specifically by shuffling the diseased/nondiseased labels of the subjects within each modality. As these permutations are made within modality, the permutation test is valid even if both modalities are measured on different scales. We show that our permutation test is a sign test for the symmetry of an underlying discrete distribution whose size remains valid under the assumption of equal AUCs. We demonstrate the operating characteristics of our test via simulation and show that our test is equal in power to a permutation test recently proposed by Bandos and others (2005).     

Median regression models for longitudinal data with dropouts

Summary .  Recently, median regression models have received increasing attention. When continuous responses follow a distribution that is quite different from a normal distribution, usual mean regression models may fail to produce efficient estimators whereas median regression models may perform satisfactorily. In this article, we discuss using median regression models to deal with longitudinal data with dropouts. Weighted estimating equations are proposed to estimate the median regression parameters for incomplete longitudinal data, where the weights are determined by modeling the dropout process. Consistency and the asymptotic distribution of the resultant estimators are established. The proposed method is used to analyze a longitudinal data set arising from a controlled trial of HIV disease ( Volberding et al., 1990 , The New England Journal of Medicine 322, 941–949). Simulation studies are conducted to assess the performance of the proposed method under various situations. An extension to estimation of the association parameters is outlined.     

Analysis of vegetation index changes and driving forces in inland arid areas based on random forest model: a case study of the middle part of northern slope of the north Tianshan Mountains

全球变化背景下的干旱区植被变化受气候变化和人类活动双重影响。定量评价植被变化特征及其驱动机制, 对监测干旱区区域生态环境变化, 促进区域可持续发展有重要意义。由于复杂多样的人类活动难以量化, 有关这方面的研究多局限于植被对气候变化的响应, 而对人类活动影响考虑不足, 导致关于这方面的认识存在较大的偏差和不确定性。该文首先提出与土地利用相关的人类活动量化表征方法; 然后运用多元线性回归模型和随机森林模型中的较优模型, 分析气候变化和具体的人类活动对北天山北坡中段归一化植被指数(NDVI)的影响。主要结果: (1) 2000-2015年期间北天山北坡中段年NDVI总体呈增加趋势; 基于随机森林构建的NDVI与气候因子和人类活动的模型拟合精度明显优于多元线性回归模型, 其决定系数(R²)至少提高了24%; (2)研究期内与耕地有关的人类活动对北天山北坡中段NDVI分布及时空变化的影响呈增加的特征, 在2000-2015年期间人类活动对NDVI变化的贡献率为0.59, 超过了气候因子。该项研究为气候变化和人类活动对植被的影响研究提供了新思路, 也为干旱区生态环境保护和恢复提供了科学依据。     

Smoothed quantile regression analysis of competing risks

Censored quantile regression models, which offer great flexibility in assessing covariate effects on event times, have attracted considerable research interest. In this study, we consider flexible estimation and inference procedures for competing risks quantile regression, which not only provides meaningful interpretations by using cumulative incidence quantiles but also extends the conventional accelerated failure time model by relaxing some of the stringent model assumptions, such as global linearity and unconditional independence. Current method for censored quantile regressions often involves the minimization of the L₁‐type convex function or solving the nonsmoothed estimating equations. This approach could lead to multiple roots in practical settings, particularly with multiple covariates. Moreover, variance estimation involves an unknown error distribution and most methods rely on computationally intensive resampling techniques such as bootstrapping. We consider the induced smoothing procedure for censored quantile regressions to the competing risks setting. The proposed procedure permits the fast and accurate computation of quantile regression parameter estimates and standard variances by using conventional numerical methods such as the Newton–Raphson algorithm. Numerical studies show that the proposed estimators perform well and the resulting inference is reliable in practical settings. The method is finally applied to data from a soft tissue sarcoma study.     

A pseudo-bayesian shrinkage approach to regression with missing covariates

Summary We consider the linear regression of outcome Y on regressors W and Z with some values of W missing, when our main interest is the effect of Z on Y, controlling for W. Three common approaches to regression with missing covariates are (i) complete‐case analysis (CC), which discards the incomplete cases, and (ii) ignorable likelihood methods, which base inference on the likelihood based on the observed data, assuming the missing data are missing at random ( Rubin, 1976b ), and (iii) nonignorable modeling, which posits a joint distribution of the variables and missing data indicators. Another simple practical approach that has not received much theoretical attention is to drop the regressor variables containing missing values from the regression modeling (DV, for drop variables). DV does not lead to bias when either (i) the regression coefficient of W is zero or (ii) W and Z are uncorrelated. We propose a pseudo‐Bayesian approach for regression with missing covariates that compromises between the CC and DV estimates, exploiting information in the incomplete cases when the data support DV assumptions. We illustrate favorable properties of the method by simulation, and apply the proposed method to a liver cancer study. Extension of the method to more than one missing covariate is also discussed.     

Vegetation pattern of mountains in West Greenland – a baseline for long-term surveillance of global warming impacts

Background: Steep environmental gradients, coupled with predicted high temperature rises in the Arctic make arctic mountain vegetation highly suitable for surveillance of changes related to global warming. However, guidelines and baselines for such a purpose are widely lacking since arctic mountain vegetation has been little explored.

Aims: We explore options for long-term surveillance on the basis of a detailed analysis of extant plant community patterns and their underlying environmental conditions in the mountainous inland of West Greenland.

Methods: Distribution, abundance and site conditions of vegetation types were analysed, using 664 vegetation samples and detailed vegetation maps in four altitudinal belts.

Results: Most plant communities had a restricted elevation distribution and were confined to special habitats predominantly defined by mesotopography and soil moisture.

Conclusions: Based on the strong linkage to habitat conditions, horizontal and vertical changes of species distribution and vegetation pattern are excellent indicators for inferring underlying environmental changes on three different scales. The recommendations given concerning climate sensitive species and plant communities, ecotones for setting up observation sites as well as stratification of analysis by habitats can be the basis for establishing long-term surveillance programmes on arctic mountain vegetation.     

Time‐lag effects of global vegetation responses to climate change

Climate conditions significantly affect vegetation growth in terrestrial ecosystems. Due to the spatial heterogeneity of ecosystems, the vegetation responses to climate vary considerably with the diverse spatial patterns and the time‐lag effects, which are the most important mechanism of climate–vegetation interactive effects. Extensive studies focused on large‐scale vegetation–climate interactions use the simultaneous meteorological and vegetation indicators to develop models; however, the time‐lag effects are less considered, which tends to increase uncertainty. In this study, we aim to quantitatively determine the time‐lag effects of global vegetation responses to different climatic factors using the GIMMS3g NDVI time series and the CRU temperature, precipitation, and solar radiation datasets. First, this study analyzed the time‐lag effects of global vegetation responses to different climatic factors. Then, a multiple linear regression model and partial correlation model were established to statistically analyze the roles of different climatic factors on vegetation responses, from which the primary climate‐driving factors for different vegetation types were determined. The results showed that (i) both the time‐lag effects of the vegetation responses and the major climate‐driving factors that significantly affect vegetation growth varied significantly at the global scale, which was related to the diverse vegetation and climate characteristics; (ii) regarding the time‐lag effects, the climatic factors explained 64% variation of the global vegetation growth, which was 11% relatively higher than the model ignoring the time‐lag effects; (iii) for the area with a significant change trend (for the period 1982–2008) in the global GIMMS3g NDVI (P < 0.05), the primary driving factor was temperature; and (iv) at the regional scale, the variation in vegetation growth was also related to human activities and natural disturbances. Considering the time‐lag effects is quite important for better predicting and evaluating the vegetation dynamics under the background of global climate change.     

A global goodness-of-fit statistic for Cox regression models

In this paper, a global goodness-of-fit test statistic for a Cox regression model, which has an approximate chi-squared distribution when the model has been correctly specified, is proposed. Our goodness-of-fit statistic is global and has power to detect if interactions or higher order powers of covariates in the model are needed. The proposed statistic is similar to the Hosmer and Lemeshow (1980, Communications in Statistics A10, 1043-1069) goodness-of-fit statistic for binary data as well as Schoenfeld's (1980, Biometrika 67, 145-153) statistic for the Cox model. The methods are illustrated using data from a Mayo Clinic trial in primary billiary cirrhosis of the liver (Fleming and Harrington, 1991, Counting Processes and Survival Analysis), in which the outcome is the time until liver transplantation or death. The are 17 possible covariates. Two Cox proportional hazards models are fit to the data, and the proposed goodness-of-fit statistic is applied to the fitted models.     

Statistical models for monitoring and predicting effects of climate change and invasion on the free-living insects and a spider from sub-Antarctic Heard Island

Terrestrial ecosystems of sub-Antarctic islands are particularly sensitive to global and local human impacts, including climate change and species invasion. Invertebrates form a central component of these ecosystems. We conducted a stratified survey of 60 sites on sub-Antarctic Heard Island and used Poisson regression models to describe the spatial distribution and abundance of five of the ten free-living species captured. Acari and Collembola were not considered. Five species were not caught in traps in sufficient numbers to model. The distributions of species were described by altitude, vegetation type and aspect. The resulting distribution models can be used to both monitor and predict the effects of climate change and species invasion on this unique and valuable ecosystem.