scholarly journals Natural mortality estimation using tree-based ensemble learning models

2020 ◽  
Vol 77 (4) ◽  
pp. 1414-1426 ◽  
Author(s):  
Chanjuan Liu ◽  
Shijie Zhou ◽  
You-Gan Wang ◽  
Zhihua Hu
Keyword(s):  
Life History ◽  
Ensemble Learning ◽  
Random Forests ◽  
Growth Parameters ◽  
Training Dataset ◽  
Natural Mortality ◽  
Statistical Regression ◽  
Learning Models ◽  
Empirical Methods ◽  
Life History Parameters

Abstract Empirical studies are popular in estimating fish natural mortality rate (M). However, these empirical methods derive M from other life-history parameters and are often perceived as being less reliable than direct methods. To improve the predictive performance and reliability of empirical methods, we develop ensemble learning models, including bagging trees, random forests, and boosting trees, to predict M based on a dataset of 256 records of both Chondrichthyes and Osteichthyes. Three common life-history parameters are used as predictors: the maximum age and two growth parameters (growth coefficient and asymptotic length). In addition, taxonomic variable class is included to distinguish Chondrichthyes and Osteichthyes. Results indicate that tree-based ensemble learning models significantly improve the accuracy of M estimate, compared to the traditional statistical regression models and the basic regression tree model. Among ensemble learning models, boosting trees and random forests perform best on the training dataset, but the former performs a slightly better on the test dataset. We develop four boosting trees models for estimating M based on varying life-history parameters, and an R package is provided for interested readers to estimate M of their new species.

Using r-K Selection Theory to Predict Natural Mortality

1980 ◽  
Vol 37 (12) ◽  
pp. 2266-2271 ◽  
Author(s):  
Donald R. Gunderson
Keyword(s):  
Body Weight ◽  
Life History ◽  
Growth Parameters ◽  
Stepwise Multiple Regression ◽  
Natural Mortality ◽  
Instantaneous Rate ◽  
Life History Parameters ◽  
Selection Theory ◽  
Weight Index ◽  
Body Weight Index

Theory on r-K selection is used as a basis for examining correlations between instantaneous rate of natural mortality (M), gonad-body weight index, age at maturity, longevity, and Bertalanffy growth parameters (k, L∞) for 10 species of marine fish. All correlations were consistent with r-K selection theory. The gonad-body weight index was found to be more highly correlated with M than any of the other life history parameters examined (r2 = 0.62), and stepwise multiple regression showed that additional variables added little to the predictive ability of the model. The gonad-body weight index appears to be quite useful in predicting M, and development of an analogous index on an energetics basis might enhance its utility in this regard.Key words: natural mortality, r-K selection, life history parameters

The Evolution of Life History Parameters in Teleosts

1984 ◽  
Vol 41 (6) ◽  
pp. 989-1000 ◽  
Author(s):  
Derek A. Roff
Keyword(s):  
Growth Rate ◽  
Life History ◽  
Life History Theory ◽  
Larval Survival ◽  
Reasonable Assumption ◽  
Natural Mortality ◽  
Age At First Reproduction ◽  
Life History Parameters ◽  
First Case ◽  
First Reproduction

Empirical studies have shown that in teleosts there is a significant correlation between the life history parameters, age at first reproduction, natural mortality, and growth rate. In this paper 1 hypothesize that these correlations are the result of evolutionary adjustments due to the trade-off between reproduction, growth, and survival. A simple and reasonable assumption is that the costs of reproduction are sufficient to cause the ltmt function to decrease. A simple expression relating the age at first reproduction is derived from this assumption. This formula accounts for a statistically significant portion (60.6%) of the variation in age at first reproduction in 30 stocks of fish. To extend the model to predict the distribution of life history parameters across all teleosts, an explicit cost function is incorporated. The model is analyzed with respect to two fitness measures, the expected lifetime fecundity and malthusian parameter, r. In the first case it is shown that the optimal age at maturity, T, depends only on the natural mortality rate (M) and the growth rate (k). In the second case, T is a function of k and the logarithm of a parameter, In C; the latter is a product of egg and larval survival, maximum body length (Lx), and the proportionality coefficient of the fecundity/length function. Difficulties of measuring egg and larval survival make the testing of the latter case difficult for particular species. However, this method provides a simple formula for the computation of r; this is shown generally to be approximately zero, thereby adding strength to the assumptions of the first analysis. The distribution patterns of T on k and M on k are predicted and compared with the observed pattern. In general, the predictions are validated: however, certain combinations of k and ln C are shown to occur very infrequently. The prediction of such "empty" regions of the parameter space remains a challenge for future development of life history theory.

scholarly journals Preliminary Estimates of Age, Growth and Natural Mortality of Margate, Haemulon album, and Black Margate, Anisotremus surinamensis, from the Southeastern United States

Fishes ◽  
2019 ◽  
Vol 4 (3) ◽  
pp. 44
Author(s):  
Michael L. Burton ◽  
Jennifer C. Potts ◽  
Andrew D. Ostrowski
Keyword(s):  
United States ◽  
Life History ◽  
Atlantic Coast ◽  
Southeastern United States ◽  
Geographic Region ◽  
Fish Growth ◽  
Natural Mortality ◽  
Von Bertalanffy ◽  
Life History Parameters ◽  
Growth Equations

Ages of margate, Haemulon album (n = 415) and black margate, Anisotremus surinamensis (n = 130) were determined using sectioned sagittal otoliths collected from the Southeastern United States Atlantic coast from 1979 to 2017. Opaque zones were annular, forming between January and June for both species, with peaks in occurrence of otoliths with opaque margins in April for margate and March for black margate. The observed ages for margate were 0–22 years, and the largest fish measured 807 mm TL (total length). Black margate ranged in age from 3 to 17 years, and the largest fish was 641 mm TL. Weight–length relationships were: margate, ln(W) = 2.88 ln(TL) − 10.44 (n = 1327, r2 = 0.97, MSE = 0.02), where W is total weight (grams, g); black margate, ln(W) = 3.02 ln(TL) − 11.10 (n = 451, r2 = 0.95, MSE = 0.01). Von Bertalanffy growth equations were Lt = 731 (1 − e−0.23(t+0.38)) for margate, and Lt = 544 (1 − e−0.13(t+2.61)) for black margate. After re-estimating black margate growth using a bias-correction procedure to account for the lack of younger fish, growth was described by the equation Lt = 523 (1 − e−0.18(t+0.0001)). Age-invariant estimates of natural mortality were M = 0.19 y−1 and M = 0.23 y−1 for margate and black margate, respectively, while age-varying estimates of M ranged from 2.93 −0.23 y−1 for fish aged 0–22 for margate and 7.20 − 0.19 y−1 for fish aged 0–18 for black margate. This study presents the first documentation of life-history parameters for margate from the Atlantic waters off the Southeastern United States, and the first published estimate of black margate life history parameters from any geographic region.

scholarly journals Evaluating the predictive performance of empirical estimators of natural mortality rate using information on over 200 fish species

2014 ◽  
Vol 72 (1) ◽  
pp. 82-92 ◽  
Author(s):  
Amy Y Then ◽  
John M Hoenig ◽  
Norman G Hall ◽  
David A Hewitt ◽  
Keyword(s):  
Mortality Rate ◽  
Water Temperature ◽  
Prediction Error ◽  
Cross Validation ◽  
Growth Parameters ◽  
Predictive Performance ◽  
Natural Mortality ◽  
Empirical Methods ◽  
Direct Estimate ◽  
Empirical Estimators

Abstract Many methods have been developed in the last 70 years to predict the natural mortality rate, M, of a stock based on empirical evidence from comparative life history studies. These indirect or empirical methods are used in most stock assessments to (i) obtain estimates of M in the absence of direct information, (ii) check on the reasonableness of a direct estimate of M, (iii) examine the range of plausible M estimates for the stock under consideration, and (iv) define prior distributions for Bayesian analyses. The two most cited empirical methods have appeared in the literature over 2500 times to date. Despite the importance of these methods, there is no consensus in the literature on how well these methods work in terms of prediction error or how their performance may be ranked. We evaluate estimators based on various combinations of maximum age (tmax), growth parameters, and water temperature by seeing how well they reproduce >200 independent, direct estimates of M. We use tenfold cross-validation to estimate the prediction error of the estimators and to rank their performance. With updated and carefully reviewed data, we conclude that a tmax-based estimator performs the best among all estimators evaluated. The tmax-based estimators in turn perform better than the Alverson–Carney method based on tmax and the von Bertalanffy K coefficient, Pauly’s method based on growth parameters and water temperature and methods based just on K. It is possible to combine two independent methods by computing a weighted mean but the improvement over the tmax-based methods is slight. Based on cross-validation prediction error, model residual patterns, model parsimony, and biological considerations, we recommend the use of a tmax-based estimator (M=4.899tmax−0.916, prediction error = 0.32) when possible and a growth-based method (M=4.118K0.73L∞−0.33 , prediction error = 0.6, length in cm) otherwise.

scholarly journals What Drives State-Sponsored Violence?: Evidence from Extreme Bounds Analysis and Ensemble Learning Models

2018 ◽  
Author(s):  
Danilo Freire ◽  
Gary Uzonyi
Keyword(s):  
Ensemble Learning ◽  
Random Forests ◽  
Large Scale ◽  
Mass Killing ◽  
Learning Models ◽  
Extreme Bounds Analysis ◽  
Mass Killings

The literature on state-sponsored violence has grown significantly over the last decades. Although scholars have suggested a number of potential correlates of mass killings, it remains unclear whether the estimates are robust to different model specifications, or which variables accurately predict the onset of large-scale violence. We employ extreme bounds analysis and distributed random forests to test the sensitivity of 40 variables on a sample of 177 countries from 1945 to 2013. The results help clear the brush around mass killings, as few variables in this literature are robust determinants of atrocity. However, support for an opportunity logic persists as greater constraints on a government limit its ability to employ barbarous tactics. It appears that the Conflict Trap applies to government atrocity. Atrocity breeds atrocity, while wealthy stable democracies tend to avoid episodes of mass killing.

Biological reference points for fish stocks in a multispecies context

2001 ◽  
Vol 58 (11) ◽  
pp. 2167-2176 ◽  
Author(s):  
Jeremy S Collie ◽  
Henrik Gislason
Keyword(s):  
Life History ◽  
Growth Rates ◽  
Prey Species ◽  
Prey Availability ◽  
Mortality Rates ◽  
Reference Points ◽  
Natural Mortality ◽  
Fishing Mortality ◽  
Predator Species ◽  
Life History Parameters

Biological reference points (BRPs) are widely used to define safe levels of harvesting for marine fish populations. Most BRPs are either minimum acceptable biomass levels or maximum fishing mortality rates. The values of BRPs are determined from historical abundance data and the life-history parameters of the fish species. However, when the life-history parameters change over time, the BRPs become moving targets. In particular, the natural mortality rate of prey species depends on predator levels; conversely, predator growth rates depend on prey availability. We tested a suite of BRPs for their robustness to observed changes in natural mortality and growth rates. We used the relatively simple Baltic Sea fish community for this sensitivity test, with cod as predator and sprat and herring as prey. In general, the BRPs were much more sensitive to the changes in natural mortality rates than to growth variation. For a prey species like sprat, fishing mortality reference levels should be conditioned on the level of predation mortality. For a predator species, a conservative level of fishing mortality can be identified that will prevent growth overfishing and ensure stock replacement. These first-order multispecies interactions should be considered when defining BRPs for medium-term (5–10 year) management decisions.

scholarly journals Some explorations of the life history ratios to describe length composition, spawning-per-recruit, and the spawning potential ratio

2014 ◽  
Vol 72 (1) ◽  
pp. 204-216 ◽  
Author(s):  
Adrian Hordyk ◽  
Kotaro Ono ◽  
Keith Sainsbury ◽  
Neil Loneragan ◽  
Jeremy Prince
Keyword(s):  
Life History ◽  
Growth Parameters ◽  
Estimation Method ◽  
Cost Effective ◽  
Analytical Models ◽  
Small Scale ◽  
Natural Mortality ◽  
Von Bertalanffy ◽  
Cost Effective Method ◽  
Spawning Potential Ratio

Abstract Evaluating the status of data-poor fish stocks is often limited by incomplete knowledge of the basic life history parameters: the natural mortality rate (M), the von Bertalanffy growth parameters (L∞ and k), and the length at maturity (Lm). A common approach to estimate these individual parameters has been to use the Beverton–Holt life history invariants, the ratios M/k and Lm/L∞, especially for estimating M. In this study, we assumed no knowledge of the individual parameters, and explored how the information on life history strategy contained in these ratios can be applied to assessing data-poor stocks. We developed analytical models to develop a relationship between M/k and the von Bertalanffy growth curve, and demonstrate the link between the life history ratios and yield- and spawning-per-recruit. We further developed the previously recognized relationship between M/k and yield- and spawning-per-recruit by using information on Lm/L∞, knife-edge selectivity (Lc/L∞), and the ratio of fishing to natural mortality (F/M), to demonstrate the link between an exploited stock's expected length composition, and its spawning potential ratio (SPR), an internationally recognized measurement of stock status. Variation in length-at-age and logistic selectivity patterns were incorporated in the model to demonstrate how SPR can be calculated from the observed size composition of the catch; an advance which has potential as a cost-effective method for assessing data-poor stocks. A companion paper investigates the effects of deviations in the main assumptions of the model on the application of the analytical models developed in this study as a cost-effective method for stock assessment [Hordyk, A. R., Ono, K., Valencia, S., Loneragan, N. R., and Prince, J. D. 2015. A novel length based empirical estimation method of spawning potential ratio (SPR), and tests of its performance, for small-scale, data-poor fisheries. ICES Journal of Marine Science, 72: 217–231].

Comparative rates of growth of the Port Jackson shark throughout its southern Australian range

2012 ◽  
Vol 63 (8) ◽  
pp. 687 ◽  
Author(s):  
Christopher Izzo ◽  
Kate R. Rodda
Keyword(s):  
Life History ◽  
Growth Parameters ◽  
South Australia ◽  
Growth Curves ◽  
Von Bertalanffy ◽  
Life History Parameters ◽  
Growth Functions ◽  
Port Jackson ◽  
Rates Of Growth ◽  
Size At Birth

Port Jackson sharks are distributed throughout southern Australia, with evidence suggesting that potential subpopulations exist. If subpopulations are evident, then phenotypic variation among groups should result in differences in life-history parameters. The present study tested for patterns of spatial variability of life-history parameters among regional Port Jackson shark populations. Rates of growth from Port Jackson sharks caught in the gulf waters of South Australia were calculated on the basis of counts of vertebral increments. Growth parameters were obtained by fitting the length-at-age data to von Bertalanffy and Gompertz growth functions. While the derived growth curves fit the length-at-age data well (r2 ranged from 0.87 to 0.91), parameters showed considerable differences between the two functions, with the von Bertalanffy function providing the more realistic estimates of growth (combined sexes: k = 0.081 year–1, L∞ = 1232 mm total length and t0 = –1.937 years). Life-history parameters for South Australian Port Jackson sharks were collated with the available data for the species, facilitating comparisons among regional populations. Growth curves among populations varied significantly; however, considerable overlap in the length ranges of size at birth and sizes at maturity among populations were evident. Overall, the data presented here do not provide definitive support for the presence of subpopulations across the distribution of the Port Jackson shark, suggesting that molecular analysis maybe required to directly test for structuring.

A Bayesian approach for overcoming inconsistencies in mortality estimates using, as an example, data for Acanthopagrus latus

2004 ◽  
Vol 61 (7) ◽  
pp. 1202-1211 ◽  
Author(s):  
Norman G Hall ◽  
S Alex Hesp ◽  
Ian C Potter
Keyword(s):  
Life History ◽  
Bayesian Approach ◽  
Growth Parameters ◽  
Probability Distributions ◽  
Total Mortality ◽  
Growth Parameter ◽  
Life History Parameters ◽  
Acanthopagrus Latus ◽  
Marine Embayment ◽  
Mortality Estimates

Reliable estimates of natural (M) and total mortality (Z) are essential for effective fisheries management. However, estimates of M, which are frequently determined from life history parameters, are imprecise and often inconsistent with the values of Z derived from life history parameters and other analyses. This is exemplified by the mortality estimates derived for Acanthopagrus latus in a large marine embayment. Thus, such estimates, calculated for M for this population from a growth parameter and from growth parameters and water temperature, were both 0.70·year–1, whereas those for Z, calculated from maximum recorded age, relative abundance analysis, and a simulation based on maximum age and sample size, ranged from 0.18 to 0.30·year–1. These results are clearly inconsistent. A Bayesian approach was therefore developed that combines the posterior probability distributions of the various mortality estimates and thereby produces integrated and consistent estimates for M and Z. The application of our Bayesian approach to the data for A. latus yielded lower values for M than for Z. Our approach is equally applicable to other fish species.

Sign in / Sign up

Export Citation Format

Share Document