Use of ecological niche modelling to predict distributions of freshwater fish species in Kansas

Abstract – An essential innovation in aquatic biodiversity research would be a robust approach to accurately predict species’ potential distributions. In this paper, I conduct an analysis to test the efficacy of ecological niche modelling for predicting fish species’ potential distributions using an artificial‐intelligence algorithm, the Genetic Algorithm for Rule‐Set Prediction (GARP). Models of species’ ecological niches are developed using GARP, and projected onto geography to predict species distributions. To test the validity of this approach, I used freshwater fish distribution data for twelve fish species occurring in Kansas. These taxa were chosen to represent phylogenetic, distribution, and habit diversity. I subset these data by omitting half of the counties from model building, and test models using the omitted counties. Models were tested using Receiver Operating Characteristic (ROC) analyses. Of the species tested, all were statistically significant with the models showing excellent predictive ability. Omission errors across taxa ranged from 0 to 17%. This inferential capacity opens doors to many synthetic analyses based on primary point‐occurrence data.     

Modelling of stream fishes in the Great Plains, USA

Abstract –  Predicting species distributions has important implications for the conservation and management of freshwater fishes, particularly in areas such as the Great Plains, USA where human impacts have resulted in extirpations and declines for numerous native species. There are a number of statistical approaches for constructing distributional models; the accuracy of each is likely dependent on the nature of the environmental gradients, species responses to those gradients and the spatial extent of the modelling. Thus, it is important to compare multiple approaches across species and habitats to identify the most effective modelling approach. Using geographical information system (GIS) derived characteristics of stream segments as predictors, we tested the model performance of three methodologies – linear discriminant function analysis, classification trees and artificial neural networks (ANN) – for predicting the occurrence of 38 fish species in a Great Plains river basin. Results showed that all approaches predicted species occurrences with relatively high success. ANN generally were the best models, in that they generated the most significant models (35 of 38 species) and most accurately predicted species presence for the greatest number of species (average correct classification = 81.1%). The importance of GIS variables for predicting stream fish occurrences varied among species and modelling techniques, but were generally strong predictors of species distributions, including the federally endangered Topeka shiner Notropis topeka . In summary, predictive models should be viewed as both competitive and complementary methodologies for establishing quantitative linkages between fish species and their environment. Our study demonstrates the potential utility of such an approach for guiding conservation efforts for stream fishes of the Great Plains, USA.     

基于GAM 的吉尔伯特群岛海域黄鳍金枪鱼栖息地综合指数

为了可持续利用黄鳍金枪鱼(Thunnus albacares)资源,本文利用2009年10月～12月吉尔伯特群岛海域海上实测的34个站点海洋环境垂直剖面数据,黄鳍金枪鱼渔获率数据,应用广义加性模型(generalized additive model,GAM) 进行建模,预测渔获率,并通过wilcoxon检验来判断预测渔获率与名义渔获率是否存在显著相关性。根据预测渔获率估算黄鳍金枪鱼的栖息地综合指数(IHI),通过对各水层IHI均值分析和Pearson相关系数,判断该方法的预测能力。使用2010年11月～2011年1月在吉尔伯特群岛海域实测的16个站点40～80m水层和0～240m水体的环境数据,验证模型。结果表明：(1)拟合的各水层的IHI值分布各不相同,各水层中影响黄鳍金枪鱼分布的因子各不相同,黄鳍金枪鱼主要栖息在40～120m水层;(2)2010年数据验证结果表明,GAM模型的预测能力较好;(3)GAM在筛选影响黄鳍金枪鱼分布的因子时比较有效,能反应黄鳍金枪鱼渔获率与环境因子之间的非线性关系;(4)可通过GAM建立IHI指数模型来分析大洋性鱼类栖息地的空间分布。     

冷却链大黄鱼货架期预报系统的开发与评估

通过使用Visual Basic编程语言,完成了基于0~10℃腐败希瓦氏菌生长动力学模型建立的冷却链大黄鱼货架期预报系统的设计,验证了货架期预报系统的准确性。采用以自然污染鱼得到的实验数据建立特定腐败菌生长动力学模型的方法,提高了预测的准确性,腐败希瓦氏菌和剩余货架期预测值与实测值的可靠性评估表明,其相关系数R2分别为0.936和0.935,显示开发的冷却链大黄鱼货架期预报系统可以快速可靠地预测冰鲜大黄鱼的鲜度与剩余高品质期和货架期。系统开发过程中采用免安装处理,方便用户使用。     

基于CART算法的长江口鱼种丰富度预测

长江口是西太平洋最大的河口生态系统和典型的生态群落交错区,预测鱼类生物多样性对评价其生态系统有着重要的科学指示意义。结合2012―2013年长江口15个站点的渔业资源和环境调查数据,利用分类与回归树(CART)算法中的回归树算法,构建了长江口鱼种丰富度预测模型。基于1-SE准则,最优决策树的复杂性参数设置为0.067,结果表明,盐度、溶解氧和季节是影响长江口鱼类生物多样性的主要因子。此外,使用2014年的观测数据对回归树模型预测的长江口鱼种丰富度予以验证,均方根误差(RMSE)、平均相对误差(ARE)和平均绝对误差(AAE)值的统计结果显示,回归树模型在春、夏季的预测效能优于秋、冬季,模型总体上呈现出了较好的预测能力,表明利用CART算法对长江口鱼种丰富度进行预测是可行的。     

Fish assemblages in European Western Highlands and Western Plains: a type-specific approach to assess ecological quality of running waters

Abstract  After typological pre-classification of 398 calibration sites, fish-based metric models were used to predict the impact of human activities on river quality in European Western Highlands and Western Plains ecoregions. Calibration sites were grouped into six assemblage types and according to their geomorphology; test sites were assigned to their corresponding assemblage type. Five anthropogenic variables were used to describe the impact level of each site and stepwise discriminant analysis was performed to: (i) avoid redundancy between metrics; (ii) examine how selected metrics discriminated impact classes and (iii) predict ecological status for each site of the given fish type. Globally, this approach predicted the impact class correctly for 64% of sites. The difference between observed and predicted impact was more than one class for only 2.5% of the sites. When validating this approach with an independent data set, differences between observed and predicted impact values never exceeded 2 impact classes, but these differences varied in size among countries.     

Modelling spatial exclusion of a vulnerable native fish by introduced trout in rivers using landscape features: a new tool for conservation management

相似文献   

Oceanographic habitat suitability is positively correlated with the body condition of a coastal‐pelagic fish

Species distribution models are commonly used to determine a species’ probability of occurrence but have not been used to examine the effect of environmental habitat suitability on fish condition, which is considered to be an integrated measure of physiological status. Here, we test for a relationship between oceanographic habitat suitability and the body condition of kingfish (Seriola lalandi) from eastern Australia. We (a) test whether individuals sampled from areas of high‐quality habitat were in better condition than individuals sampled from areas of low‐quality habitat, and (b) assess whether the condition of kingfish responded to oceanographic habitat suitability predicted at varying time‐before‐capture periods. Kingfish habitat was modelled as a function of sea surface temperature, sea‐level anomaly and eddy kinetic energy in a generalized additive modelling framework. Model predictions were made over one‐ to six‐week time‐before‐capture periods and compared to field‐derived kingfish condition data measured using bioelectrical impedance analysis. Oceanographic habitat suitability was significantly correlated with kingfish condition at time‐before‐capture periods ranging from one to four weeks and became increasingly correlated at shorter lead‐times. Our results highlight that (a) fish condition can respond sensitively to environmental variability and this response can be detected using oceanographic habitat suitability models, and (b) climate change may drive extensions in species range limits through spatial shifts in oceanographic habitat quality that allow individuals to persist beyond historical range boundaries without their body condition being compromised.     

Accounting for variation in species detection in fish community monitoring

Long‐term fish assemblage monitoring requires investigators account for within‐year variation in species' detection. An occupancy modelling framework is presented that accounts for variation in species presence and estimates the effort required to minimise within‐year variation. Species detections from snorkelling surveys and an electrofishing survey were used in single‐species occupancy models to determine the importance of site and sampling covariates on species' occupancy (psi) and detection probabilities (p). Community occupancy models estimated species richness in light of undetected species. For most species, models including patch size and reach as psi‐covariates had higher support whereas models including patch size and sampling method as p‐covariates had higher support. The number of sites and repeated surveys required to estimate occupancy accurately varied among fish species. Community models suggested that the observed number of species underestimated actual richness as much as 27% and the addition of an electrofishing survey increased estimated richness by 13%.     

Distribution and richness of diadromous fish assemblages in Western Europe: large-scale explanatory factors

Abstract –  The aim of this study was to analyse the distribution of 14 diadromous fish at the beginning of the 20th century in western Europe. This study was conducted on a set of 41 water basins. Five environmental variables were selected and we used generalised additive models for explaining the presence–absence of species. The richest basins were located in the centre of the study area. Six main assemblage types were identified along a latitudinal gradient; they were constituted of a common species basis but differed by the absence or presence of other species. The 10 single species models produced have moderate to very good discrimination level and they can correctly predict both absence and presence. Temperature is included in all but one model, response curves vary according to the species; surface area is included in six models.     

Using historical fisheries data to predict tuna distribution within the British Indian Ocean Territory Marine Protected Area,and implications for its management

相似文献   

基于卷积神经网络的养殖鱼类品种识别模型

随着机器学习、计算机视觉等技术的发展,卷积神经网络(CNN)越来越多地应用于图像识别领域,但现有的鱼类图像公共数据集资源较匮乏,难以满足深度CNN模型优化及性能提升的需要。实验以大黄鱼、鲤、鲢、秋刀鱼和鳙为对象,采用网络爬虫以及实验室人工拍照采集相结合的方式,构建了供鱼种分类的基础图片数据集,针对网络爬虫手段获取到的鱼类图像存在尺度不一、格式不定等问题,采用图像批处理的方式对所有获取到的图像进行了统一的数据预处理,并通过内容变换以及尺度变换对基础数据集做了数据增强处理,完成了7 993个样本的图像采集与归纳;在权值共享和局部连接的基础上,构建了一个用于鱼类识别的CNN模型,采用ReLU函数作为激活函数,通过dropout和正则化等方法避免过度拟合。结果显示,所构建的CNN鱼种识别模型具有良好的识别精度和泛化能力。随着迭代次数的增加,CNN模型的性能也逐步提高,迭代1 000次达到最佳,模型的准确率为96.56%。该模型采用监督学习的机器学习方式,基于CNN模型,实现了5种常见鱼类的鱼种分类,具有较高的识别精度和良好的稳定性,为养殖鱼类的品种识别提供了一种新的理论计算模型。     

Fish‐Net: Probabilistic models for fishway planning,design and monitoring to support environmentally sustainable hydropower

The construction of fishways for upstream and downstream connectivity is the preferred mitigation measure for hydropower dams and other riverine barriers. Yet empirical evidence for effective design criteria for many species is missing. We therefore assembled a group of international fishway designers and combined their knowledge with available empirical data using a formal expert elicitation protocol and Bayesian networks. The expert elicitation method we use minimizes biases typically associated with such approaches. Demonstrating our application with a case‐study on the temperate Southern Hemisphere, we use the resulting probabilistic models to predict the following, given alternative design parameters: (i) the effectiveness of technical fishways for upstream movement of migratory fish; (ii) habitat quality in nature‐like bypasses for resident fish; and (iii) rates of mortality during downstream passage of all fish through turbines and spillways. The Fish Passage Network (Fish‐Net) predicts that fishways for native species could be near 0% or near 100% efficient depending on their design, suggesting great scope for adequate mitigation. Sensitivity analyses revealed the most important parameters as follows: (i) design of attraction and entrance features of technical fishways for upstream migration; (ii) habitat preferences of resident fish in nature‐like bypasses; and (iii) susceptibility of fish to barotrauma and blade strike during turbine passage. Numerical modelling predicted that mortality rates of small‐bodied fish (50–100 mm TL) due to blade strike may be higher for Kaplan than Francis turbines. Our findings can be used to support environmentally sustainable decisions in the planning, design and monitoring stages of hydropower development.     

Evaluation of four models for estimating the population size of largemouth bass in an experimental pond

The utility of four commonly used models for estimating population size in teleosts was tested. Sixty-five individually marked largemouth bass, Micropterus salmoides, were introduced into a concrete pond. Fishing surveys were conducted every 2 days for a period of 19 days. The collected data were then used to estimate the population size under a variety of conditions using the following models: mark/recapture (Petersen method), DeLury (first model), and two models of the software program Capture. Comparison of the actual population size with population estimates obtained using the mark/recapture method showed that the percentage of absolute error was <30% in all cases in which the number of fish caught and marked in the first survey was >30% of the population. Using the DeLury method and Model 1 of Capture, the population estimates were biased toward underestimation, but the error was <30% when the number of fish caught in all surveys was >70%. In contrast, in Model 2 of Capture, the error was relatively small when the percentage of fish caught in all surveys was <70%. These conditions for minimizing errors should be taken into account by fisheries managers when estimating the population size of largemouth bass.     

Prediction of digestible energy content across feed ingredients and fish species by linear regression

In an attempt to develop predictive relationships, apparent digestible energy (ADE) content (n = 361) as biological dependent variable and dietary contents of crude protein (CP), lipid, ash and gross energy (GE) as independent variables, obtained in 40 studies with 65 different feed ingredients and evaluated with 26 fish species, were subjected to linear correlation and multiple linear regression analysis. With dietary CP and GE contents identified as significant predictors, only 58% of the variation in ADE content could be explained. No improvement in accuracy of regression equations was gained by classification of values according to either feed ingredient (animal proteins, plant proteins, cereals) or fish species (water type, water temperature, feed habit). An R ²-value of 0.4570 and mean prediction error (MPE) of 0.2085 between predicted and observed ADE values from eight independent studies (n = 37) illustrated the inability of the derived regression equation to accurately predict ADE contents of feed ingredients. Inclusion of dietary crude fibre and nitrogen-free extract (NFE) contents as independent variables did not improve the accuracy of prediction equations. The inadequacy of the use of linear regression to predict DE content from dietary composition across feed ingredients and fish species with high accuracy is discussed.     

Modelling the oceanic habitats of two pelagic species using recreational fisheries data

Defining the oceanic habitats of migratory marine species is important for both single species and ecosystem‐based fisheries management, particularly when the distribution of these habitats vary temporally. This can be achieved using species distribution models that include physical environmental predictors. In the present study, species distribution models that describe the seasonal habitats of two pelagic fish (dolphinfish, Coryphaena hippurus and yellowtail kingfish, Seriola lalandi), are developed using 19 yr of presence‐only data from a recreational angler‐based catch‐and‐release fishing programme. A Poisson point process model within a generalized additive modelling framework was used to determine the species distributions off the east coast of Australia as a function of several oceanographic covariates. This modelling framework uses presence‐only data to determine the intensity of fish (fish km^?2), rather than a probability of fish presence. Sea surface temperature (SST), sea level anomaly, SST frontal index and eddy kinetic energy were significant environmental predictors for both dolphinfish and kingfish distributions. Models for both species indicate a greater fish intensity off the east Australian coast during summer and autumn in response to the regional oceanography, namely shelf incursions by the East Australian Current. This study provides a framework for using presence‐only recreational fisheries data to create species distribution models that can contribute to the future dynamic spatial management of pelagic fisheries.     

Machine learning highlights the importance of primary and secondary production in determining habitat for marine fish and macroinvertebrates

1. Species distribution models for marine organisms are increasingly used for a range of applications, including spatial planning, conservation, and fisheries management. These models have been constructed using a variety of mathematical forms and drawing on both physical and biological independent variables; however, what might be called first-generation models have mainly followed the form of linear models, or smoothing splines, informed by data collected in the context of fish surveys.
2. The performance of different classes of variables were tested in a series of species occurrence models built with machine learning methods, specifically evaluating the potential contribution of lower trophic level data. Random forest models were fitted based on the classification of the absence/presence for fish and macroinvertebrates surveyed on the US Northeast Continental Shelf.
3. The potential variables included physical, primary production, secondary production, and terrain variables. For accepted model fits, six variable importance measures were computed, which collectively showed that physical and secondary production variables make the greatest contribution across all models. In contrast, terrain variables made the least contribution to these models.
4. Multivariable analyses that account for all performance measures reinforce the role of water depth and temperature in defining species presence and absence; however, chlorophyll concentration and some specific zooplankton taxa, such as Metridia lucens and Paracalanus parvus, also make important contributions with strong seasonal variations.
5. Our results suggest that lower trophic level variables, if available, are valuable in the creation of species distribution models for marine organisms.

     

A simple model for predicting ecological impacts of introduced aquatic organisms: a case study of common carp, Cyprinus carpio L., in the Sepik-Ramu River Basin, Papua New Guinea

Abstract Practical models for predicting the impacts of introduced biota are urgently required to assess the benefits and risks of introductions. The simple method described predicts the ecological consequences of an introduction through potential competition between species based on elementary niche classifications. The model's predictions are tested against field data for common carp, Cyprinus carpio L., which escaped from fish farms during a study of the feasibility of fish stock enhancement in a large river basin. Recorded effects of carp are based on pre- and post-carp gillnet catches, observations of local villagers, socio-economic censuses and other sources of data. The predicted ecological interactions of common carp are in broad agreement with those recorded. The method can assist pre-introduction assessments and is particularly useful where limited data exist. In this particular study, common carp have contributed significantly to an improved capture fishery by fortuitous accident; production from aquaculture was negligible. High niche overlap between carp and certain resident fish species is predicted and this may be of more consequence in highland than lowland regions. Organisms introduced/transferred for aquaculture should be assessed within a much wider forum. The intended impacts of an introduction/ transfer, which in the case study were entirely erroneous, is an area in need of considerable attention.     

Evaluation of the robustness of predictive yield models based on catchment characteristics using GIS for reservoir fisheries in Sri Lanka

Land-use patterns in the catchment areas of Sri Lankan reservoirs, which were quantified using Geographical Information Systems (GIS), were used to develop quantitative models for yield prediction. The validity of these models was evaluated through the application to five reservoirs that were not used in the development of the models, and by comparing with the actual fish yield data of these reservoirs collected by an independent body. The robustness of the predictive models developed was tested by principal component analysis (PCA) on limnological characteristics, land-use patterns of the catchments and fish yields. The predicted fish yields in five Sri Lankan reservoirs, using the empirical models based on the ratios of forest cover and/or shrub cover to reservoir capacity or reservoir area were in close agreement with the observed fish yields. The scores of PCA ordination of productivity-related limnological parameters and those of land-use patterns were linearly related to fish yields. The relationship between the PCA scores of limnological characteristics and land-use types had the appropriate algebraic form, which substantiates the influence of the limnological factors and land-use types on reservoir fish yields. It is suggested that the relatively high predictive power of the models developed on the basis of GIS methodologies can be used for more accurate assessment of reservoir fisheries. The study supports the importance and the need for an integrated management strategy for the whole watershed to enhance fish yields.