Manual of Fisheries Survey Methods II January 2000

Manual of Fisheries Survey Methods II: with periodic updates

Chapter 7: Stream Fish Population Estimates by Mark-and-Recapture and Depletion Methods
Roger N. Lockwood and James C. Schneider

TOC

NEXT PAGE PREVIOUS PAGE CITATION

Suggested citation: Lockwood, Roger N. and J. C. Schneider. 2000. Stream fish population estimates by markand-recapture and depletion methods. Chapter 7 in Schneider, James C. (ed.) 2000. Manual of fisheries survey methods II: with periodic updates. Michigan Department of Natural Resources, Fisheries Special Report 25, Ann Arbor.

Chapter 7

TOC NEXT PAGE PREVIOUS PAGE CITATION .

Summing the size group estimates by species to obtain an estimate of the total population within the size range actually sampled. Noting the ratio of marked to unmarked fish by species and size (e. 7. inch group). 7. 3. which is also widely Chapter 7 1 . 6.g.Petersen methods Ricker (1975) discusses the calculation of population estimates in detail.. Allowing at least 1 day for marked fish to recover and become mixed in the population. 7. 5. 2. 5. 4. Mark-and-recapture methods can also be used in deeper streams (as by electrofishing from a boat) if it can be reasonably assumed all targeted fish are vulnerable because either (a) all parts of the stream can be sampled or (b) marked fish are randomly mixed with unmarked fish. Marked fish randomly mix with unmarked fish. 4. There is negligible emigration or immigration during the recapture period. such as a tag or temporary fin clip. Marks are retained during the sampling period and all marks on recaptured fish are recognized. The mark-andrecapture method requires the following conditions: TOC NEXT PAGE PREVIOUS PAGE CITATION 1. 8. Two basic methods are available.1 Mark-and-recapture estimates The mark-and recapture method is generally preferred over the depletion method and has been shown to be unbiased when more than 50% of a population is marked (Jensen 1992). Giving fish identifying marks. Releasing fish in good condition back into the same area. Schneider Estimates of the total number of fish in sections of streams can be made reliably and inexpensively by subsampling a portion of the population. Tabulating data by species and size. The general process for estimating a fish population using the mark-and-recapture method entails: 1. The Chapman variation is very similar to the Bailey modification of the Petersen equation. Collecting a sample of fish of the target species from a discrete section of stream during an initial “marking run”. He recommends a slight variation of the Chapman modification of the Petersen equation because it gives a statistically unbiased estimate for finite populations. Calculating for each combination of species and size group (to compensate for gear selectivity) an estimate of abundance by a Petersen equation. 3. 9. Collecting a random sample of fish during a subsequent “recapture run”.Manual of Fisheries Survey Methods II January 2000 Chapter 7: Stream Fish Population Estimates by Mark-and-Recapture and Depletion Methods Roger N.1. such as we deal with in inland waters. mark-and-recapture and depletion. Lockwood and James C. Marked and unmarked fish have the same mortality rates. Either method is appropriate for shallow streams which can be waded and thoroughly sampled with electrofishing gear. Marked and unmarked fish are equally vulnerable to capture.1 Chapman . 2.

and approximate 95% confidence limits = N ±2(200) = N ±400.Manual of Fisheries Survey Methods II January 2000 used. M = number of fish caught.834 .1).1–If 100 fish were marked and released from the first run. These provide low and high ranges for R which are then substituted in Equation (1) to calculate the lower and upper confidence limits. it is not the best estimator of variance for single estimates (Ricker 1975). Estimation of population N and variance of N. 10 + 1 (100 + 1) 2 (80 + 1)(80 − 10) = 39. marked and released in first sample. Continued on next page. for example when summing estimates and variances for two to more size groups to obtain a total population estimate. Ricker (1975) stated that the probability of a systematic statistical bias in the population estimate can be ignored if recaptures number 3-4 or more. Variance Equation (2) should be used whenever variance estimates are to be combined. Therefore. However. and both produce estimates slightly less than the simple Petersen equation. i. 95% confidence limits = N ± t(Standard error) .e. C = total number of fish caught in second sample (including recaptures). R = number of recaptures in the second sample (fish marked and released in the first sample). (2) Standard error = Variance of N . Chapter 7 2 . N 2 (C − R ) (C + 1)( R + 2) (1) Variance of N = ( M + 1) 2 (C + 1)(C − R ) ( R + 1) ( R + 2) 2 = . t = Student’s t for C-1 degrees of freedom. where. While 95% confidence limits are often used for research. These limits are typically asymmetrical and measure variability more accurately. and the second run contained 80 fish of which 10 were recaptures: N= (100 + 1)(80 + 1) = 744 fish . 68% limits (1 standard error) may be suitable for management purposes. N = population estimate. if necessary. Lower limit ( N L ) = 344 and Upper limit ( N U ) = 1. (10 + 1) 2 (10 + 2) TOC NEXT PAGE PREVIOUS PAGE CITATION Variance of N = Standard error = 39834 = 200 = 68% confidence limit..144. pool data from adjacent size groups to obtain at least 3-4 recaptures per estimate. follow as: N = ( M + 1)( C + 1) R+1 . with the Chapman modification. Example 7. His recommendation is to use either binomial charts or a Poisson distribution (Table 7.

92 ± 6. 18.42 + 1 and NU = We conclude from this example that the population contains about 744 fish.42 and RU = 18. it is expedient to collect all data within a short time period such as one day. inch group 9 has an estimated population of 293 fish with a variance of 12. TOC NEXT PAGE PREVIOUS PAGE CITATION In many studies the investigator may desire to add several population estimates and have ˆ ). 5.100 + 3.e. The appropriate equation for computing a total variance for j inch groups is: ˆ) = Var ( N j Var ( N ) i i =1 (3) Example 7. 7.185.100.92 ± 1.1 are: R + 1. or RL = 5. avoidance of recapture by marked fish) is unknown.427 . and ≥10. and inch group 10+ has an estimated population of 153 with a variance of 3. N L = 421 and N U = 1.2 Depletion estimates The depletion method (also known as the “Zippin” method. This method Chapter 7 3 .1–Continued.392 + 12.501 .960 R + 1. When substituted for R in Equation (1): NL = (100 + 1)(80 + 1) = 421 . If fish are likely to migrate in or out of a study section soon (say in less than 1 week).274.42 .g.274.Manual of Fisheries Survey Methods II January 2000 Example 7.2–Inch group 8 has an estimated population of 357 fish with a variance of 20. and the population being estimated is relatively small (roughly less than 2. the depletion method is superior to the mark-and-recapture method due to a shorter sampling time period. and approximate 95% confidence limits = N i.42 + 1 (100 + 1)(80 + 1) = 1. For example.935 = 36. one might have separate population confidence limits for the total population ( N estimates for trout in inch groups 8. Variance of N ˆ = 36. see Zippin 1958 ) is satisfactory if the stream is very small.935: ˆ = 357 + 293 + 153 = 803 fish . and any systematic error (e.392. which when added give the total number of catchable-size trout. N ˆ = 20. Note that this measurement of error is for random error only. Better 95% confidence limits from the equation given in Table 7. 9.0 = 11. Standard error of N ˆ ± 2(191) = N ˆ ± 382 .. but the statistical error is relatively large and with 95% certainty the true number lies between 421 and 1.000 individuals).427 = 191 ..

Manual of Fisheries Survey Methods II January 2000 requires that an adequate number of fish be removed on each sampling pass so that measurably fewer fish are available for capture and removal on a subsequent pass.. Population estimate N and variance of N are calculated as: p= N= C1 −C2 . For both two-pass and multiple-pass methods. N = population estimate. where. Collection effort and conditions which affect collection efficiency. Vulnerability to capture of fish in a specified sample group must remain constant for each pass (e.1 Two-pass depletion methods TOC NEXT PAGE PREVIOUS PAGE CITATION For two-pass depletion estimates. The following conditions must be met for accurate depletion method estimates: 1. fish are captured and removed during two capture sessions. Record number of fish removed (or marked) by species and size group. 3. Chapter 7 4 .g. Because of differences in gear selectivity. Two types of depletion methods are used.2. 4. as with mark-and-recapture methods. (C1 − C 2 ) C12 C 22 (C1 + C 2 ) . partitioning estimates by species and size groups is recommended. two-pass and multiple-pass. 7. 4. Depletion estimates are made following the general process: 1. size group estimates and their variances are summed. to provide total population estimates. If steps 1 and 2 were completed more than twice. Equations provided here are described in greater detail in Seber and Le Cren (1967). All fish within a specified sample group must be equally vulnerable to capture during a pass. Remove (or mark to simulate removal) fish within a discrete section of stream. (C1 − C 2 )4 (4) (5) Variance of N = (6) Standard error of N = Variance of N . calculate population estimates using multiplepass equations. p = probability of capture. must remain constant. 3. fish do not become more wary of capture). Emigration and immigration by fish during the sampling period must be negligible. If steps 1 and 2 were completed twice. 2. C 2 = number of fish removed in second sample. 2. C1 = number of fish removed in first sample. calculate population estimates using two-pass equations. 5. Repeat steps 1 and 2. C1 C12 . such as water clarity.

p s = p ). i. when less than 20% of the population is caught per pass). 7.20 (i. Estimation steps are as follows: T= X= k Ci .e.Manual of Fisheries Survey Methods II January 2000 Two-pass depletion estimates are unbiased when p≥0. i =1 k i =1 (7) (8) (k − i )Ci .e. Chapter 7 5 .. i = pass number. 200 200 2 = 381 . Example 7.. k = number of removals (passes). T = total number of fish caught in all passes. X = an intermediate statistic used below. (200 − 95) 4 Variance of N = Standard error of N = 876 = 30 . Estimated population and confidence limits are calculated as: p= N= 200 − 95 = 0. N L = 321 and N U = 441.2 Multiple-pass depletion methods TOC NEXT PAGE PREVIOUS PAGE CITATION This method requires three or more passes on the selected stream section and involves additional calculations to estimate the population.3–On the first pass 200 fish were collected and on the second pass 95 fish were collected. Ci = number of fish caught in ith sample... The multiple pass depletion method relies heavily upon consistent catchability ( p1 = p 2 = p 3 =. and approximate 95% confidence limits = N ± 2(30) = N ± 60 .2. (200 − 95) (200 2 )(95 2 )(200 + 95) = 876 .525 . where. Further description of these equations are found in Carle and Strub (1978).80 and quite unreliable when p≤0.

6>1. Example 7. TOC X = [(3-1) * 300] + [(3-2) * 130] + [(3-3) * 69] = 600 + 130 + 0 = 730. Since estimating N is an iterative process. and estimates of N and variance of N.4–On the first pass 300 fish were removed. For our second try let’s use n = 520: 520 + 1 520 − 499 + 1 (3 * 520 ) − 730 − 499 + 1 + ( 3 − 1) (3 * 520 ) − 730 + 2 + ( 3 − 1) ( 3 * 520 ) − 730 − 499 + 1 + ( 3 − 2 ) ( 3 * 520 ) − 730 + 2 + ( 3 − 2 ) ( 3 * 520 ) − 730 − 499 + 1 + ( 3 − 3) ( 3 * 520 ) − 730 + 2 + ( 3 − 3) = (23. i (9) where n is the smallest integer satisfying Equation (9).6818)(0. These Equations should be setup in a spreadsheet to facilitate selection of n. and 69 on the third: T = 300 + 130 + 69 = 499.Manual of Fisheries Survey Methods II January 2000 The maximum likelihood estimate of N is determined by an iterative process by substituting values for n until: n +1 n − T +1 ∏ i =1 k kn − X − T + 1 + (k − i ) kn − X + 2 + (k − i ) ≤ 1. and variance of N are then estimated by: p= T . Chapter 7 6 . kN − X (10) N (N − T )T Variance of N = T 2 (kp )2 − N (N − T ) (1 − p ) .0 we must select another number for n greater than 499.3998)(0. (11) Standard error of N = Variance of N .4005)(0.0)(0.0.3515)(0. a suggested initial value for n is T. so for our first n let’s try 499 in Equation (9): 499 + 1 499 − 499 + 1 (3 * 499 ) − 730 − 499 + 1 + ( 3 − 1) ( 3 * 499 ) − 730 + 2 + ( 3 − 1) ( 3 * 499 ) − 730 − 499 + 1 + (3 − 2 ) ( 3 * 499 ) − 730 + 2 + (3 − 2 ) ( 3 * 499 ) − 730 − 499 + 1 + ( 3 − 3) ( 3 * 499 ) − 730 + 2 + ( 3 − 3) NEXT PAGE PREVIOUS PAGE CITATION = (500.5540 which is rounded to 21.3990) = 1.3498)= 21. Since 21. p. and subsequent selections for n should progressively increase from T. 130 on the second.3506)(0. Probability of capture. Note that results of Equation (9) are rounded to one decimal place. We know there were at least 499 fish (T) in the population.5.6. Continued on next page.

Treating these data as depletion data gives C1 = 100.5496 . 100 fish would have been captured on the second pass and Equations (1) and (5) would have given similar results.7) = N ± 23 . 100 fish (C1) were captured and marked.1 and finally 1.4–Continued. When using the two-pass method and fish are removed from the section on the first pass. This verification technique may only be used when 1 or more days elapse between passes. However.1.0 without going over.Manual of Fisheries Survey Methods II January 2000 Example 7.109 (3 * 0.805.1. that is p1 = p 2 =. 120 − 60 Using depletion Equation (5): N= Chapter 7 7 . the maximum likelihood estimate of the population (N) is 546. i. TOC NEXT PAGE PREVIOUS PAGE CITATION As previously mentioned.2. 2 94. Since our goal is to come close to 1. 540.338 = = 136 . approximate 95% confidence limit = N ± 2(11.5–On the first pass 120 fish are marked and released back into the section.. the depletion method assumes that all s capture probabilities for k passes are equivalent. C2 = 70 with N = 333 (using Equation 5). This result is substantially different from the mark-andrecapture estimate of 744 using Equation (1) and is caused by unequal catch probabilities. On the second pass 60 marked fish and 60 unmarked fish are captured: Using mark-and-recapture Equation (1): N= (120 + 1)(120 + 1) = 240 . 60 + 1 120 2 = 240 . N L = 523 and N U = 569. If capture probabilities had been equal. 80 fish were captured of which 70 (C2) were unmarked. 545 and 546 yield estimates for N of 1. On the first pass. Example 7.0. On the second pass.5496) 2 499 − 546 * (546 − 499) * (1 − 0. We can now estimate probability of capture (p) and error statistics: p= Variance of N = 499 = 0..7 . 1. there is no way to verify this assumption. p s = p ..5496) Standard error of N = 136 = 11.e. 1. when fish are marked and released back into the section (to simulate removal) number of fish captured on each pass are expected to be similar. Subsequent trials with n equal to 530. Consider the data presented in Example 7. (3 * 546) − 730 546 * (546 − 499) * 499 12.

Chapter 7 8 . the following values would be plotted: Pass 1 2 3 Sum of previous catches 0 300 430 Catch 300 130 69 400 300 TOC NEXT PAGE PREVIOUS PAGE CITATION Catch 200 100 0 0 100 200 300 400 500 600 700 Sum of previous catches Figure 7.1–Individual catches plotted against the sum of previous catches from example 7.4 where C1 =300. i =0 (12) Example 7.6–Using data from example 7.Manual of Fisheries Survey Methods II January 2000 When more than two passes are made equality of p’s can be verified in numerous ways. Seber and Le Cren (1967) suggest simply plotting each catch against the sum of all previous catches: k −1 Ci . C2 =130 and C3 =69.4.

from Equation (10): E (C1 ) = Np . Variation in capture probability hinges on numerous factors. + + ⋅⋅⋅ + i E (C1 ) E (C 2 ) E (C k ) (15) The test statistic χ 2 from Equation (15) is compared with χ 02. When variation in capture probability is severe. E (C i ) = N (1 − p ) i −1 p (14) Calculated χ 2 then is: χ2 = [C − E (C k )]2 [C1 − E (C1 )] 2 [C 2 − E (C 2 )]2 . A copy of CAPTURE (White et al. and χ 0 . Appropriate steps should be taken to minimize immigration and emigration of fish.185 + 1.7–Using the data from example 7. with k-2 degrees of freedom (df).5496) = 300 . and estimates of N and variance of N using Equations (7-11) are invalid. more computationally intense methods such as those given by Schnute (1983) or White et al (1982) are necessary. Heggberget and Hesthagen (1979) used the 2-pass depletion method to estimate Atlantic salmon Salmo salar and brown trout Salmo trutta in two small Norway streams and suggested that populations were underestimated by as much as 50% due to electrical current avoidance on the second pass. 300 135 61 NEXT PAGE PREVIOUS PAGE CITATION χ2 = 2 where df = 3-2=1. E (C 3 ) = 546(1 − 0. if χ 2 ≥ χ 02. (13) and for i>1.4: C1 =300. effect of fish movement on population estimates can be reduced by sampling longer sections.95 .841 . 1982). When sampling streams where blocking nets are not practical.5496) 2 (0. The first step is to calculate the expected number of fish collected on each pass using the estimated population N. as by using blocking nets on small streams to greatly reduce fish movement. TOC E (C 2 ) = 546(1 − 0. C3 =69. Of particular concern is increased wariness when fish are exposed to electrofishing.5496)1 (0.5496) = 61 . 1982) is Chapter 7 9 .95 . from Equation (9). Table 2. This follows the χ 2 test form (Observed .5496: E (C1 ) = 546(0. Note that two degrees of freedom are lost because N is estimated (Snedecor and Cochran 1991:77).5496) = 135 .95 .95 = 3. N=546 and p=0. Peterson and Cederholm (1984) found that probability of capture for shocked juvenile coho salmon Oncorhynchus kisutch was similar to previous capture rates after a minimum of 1 hour recovery time.049 = 1. and estimated probability of capture p. (300 − 300) 2 (135 − 130) 2 (61 − 69) 2 + + = 0. C 2 =130. On the other hand. then probability of capture was significantly different with 95% certainty. probability of capture did not differ significantly (at the 95% level of certainty) between passes.95 we show no significant difference (95% certainty) between capture probabilities and Equations (7-11) are appropriate.000 + 0. Since χ 2 < χ 02. the amount of effort used on each pass should be as constant as possible and estimates should be stratified by species and size group to avoid gear selectivity. If χ 2 < χ 02.234 .Expected) 2 / Expected . Example 7.Manual of Fisheries Survey Methods II January 2000 A goodness of fit test may also be used to evaluate equality of capture probability (White et al. To minimize error.

the marked fish may not have recovered and become mixed. and ignoring marked fish for depletion method analysis or counting them as “recaptures” for mark-and-recapture method analysis. >30%) of the population be captured in each sample. Note however. If marked fish are more difficult to capture than unmarked fish on the second pass. the population will be underestimated. The option is left open by marking and releasing fish after the first pass. it is wise to design sampling procedures to retain the option of computing mark-and-recapture estimates while conducting depletion sampling. if more than one pass is made per day. noting their recapture in subsequent passes. Smallmouth bass Micropterous dolomieu in particular are difficult to recapture and mark-and-recapture methods are not recommended (Lyons and Kanehl 1993). The assumptions of the depletion method are rigorous regarding constant fish catchability for each sample and that more than 20% (better. thereby violating the basic assumption of the mark-and-recapture method. If marked fish are more easily captured than unmarked fish on the second pass. Mark-and-recapture methods also have essential constraints. the population will be overestimated. Therefore. TOC NEXT PAGE PREVIOUS PAGE CITATION Chapter 7 10 . This interactive software calculates an appropriate estimate of N and variance of N when capture probabilities are different.Manual of Fisheries Survey Methods II January 2000 available at the Institute for Fisheries Research. The assumptions of the depletion method are most suspect in large streams (more likely to have refuges) and for large fish (more likely to be agile or wary).

0 .182 2.0 32.5 63.447 2.4 16.2 28.0 13.92 ± 1.2 6.9 10.6 2.1–Poisson distribution of lower and upper 95% confidence limit coefficientsa for number of recaptures (R).131 2.1 32.086 2.4 41.3 54.074 2.9 59.6 42.5 35.8 22.0 1. use the following equation (Ricker 1975) for 95% limit coefficients: R + 1.Manual of Fisheries Survey Methods II January 2000 Table 7.6 30.4 15.8 65.365 2.6 34.6 29.7 49.2 53.2 2.069 2.1 26.4 9.4 35.0 0.4 30.4 6.0 52.8 14.6 15.5 55.1 37.7 11.4 29.110 2.064 2.6 1.201 2. and Student’s 95% confidence t values for number of degrees of freedom (df).4 20.0 45.228 2.7 28.3 31.9 TOC NEXT PAGE PREVIOUS PAGE CITATION Substitute the coefficients for R in Formula (1).2 9.0 27.3 62.8 18.7 13.6 7.8 3.0 21.5 48.2 40.5 24.960 R 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 a Lower 0.080 2.7 8.1 18.2 21.0 22.2 11.093 2.7 23.2 34.6 56.1 46.2 13.7 5.179 2.3 36.8 51.8 33.776 2.120 2.145 2.571 2.306 2. Chapter 7 11 .0 33.6 36.2 61.4 4.706 4.8 32.0 26.8 17.1 0.101 2.96 R + 1.4 19.7 5.8 57.0 Upper 38.2 8. For larger values of R.5 24.3 47.8 10.7 21.0 4.8 27.9 7.303 3.6 19. Poisson distribution Upper R 3.0 39.0 17.8 44.2 0.160 2.8 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 Student’s t value df t95 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 60 ∞ 12.3 25.8 26.6 64.5 12.262 2.000 1.1 14.3 23.2 Lower 17.0 60.

878 6.95 3.408 3.507 16.074 2.231 8.524 10.383 9.064 7.665 4.991 7. see χ 2 tables in texts such as Snedecor and Cochran (1991).488 11. TOC NEXT PAGE PREVIOUS PAGE CITATION Chapter 7 12 .592 14.781 χ 02.070 12.815 9.70 1.Manual of Fisheries Survey Methods II January 2000 Table 7. Degrees of freedom 1 2 3 4 5 6 7 8 9 10 a χ 02.307 For additional degrees of freedom or alternate levels of certainty.2–Percentiles of the χ 2 distribution for 70% and 95% certaintya.067 15.919 18.841 5.656 11.

Cooper and J.. Integrated area sampling and mark-recapture experiments for sampling fish populations. L. G. Estimating population parameters from catches large relative to the population..Manual of Fisheries Survey Methods II January 2000 7. 1979. T. Los Alamos: Los Alamos National Laboratory. Salmo salar L. 1983. TOC NEXT PAGE PREVIOUS PAGE CITATION Snedecor. Journal of Animal Ecology 36: 631-643. Capture-recapture and removal methods for sampling closed populations. and D.3 References Carle. J. Anderson. G. D. Population estimates of young Atlantic Salmon.. Journal of Wildlife Management 22: 82-90. 1984. Le Cren. Forest Service. F. Minnesota. Ames. 1978. 1967. A comparison of the removal and mark-recapture methods of population estimation for juvenile coho salmon in small stream. Peterson. T. A. 1993. Salmo trutta L. C. and brown trout. Heggberget. Biometrics 34: 621-630... E. Lyons. J. Statistical methods. P. 1958. N. Fisheries Research Board of Canada. R. Ryckman Rewritten 8/1998 by R. eighth edition. N. and M. Drottningholm. Computation and interpretation of biological statistics of fish populations... Kanehl. Biometrics 48: 1201-1205. 58: 27-33. Zippin. Rep. A comparison of four electroshocking procedures for assessing the abundance of smallmouth bass in Wisconsin streams... C. General Technical Report NC-159. Jensen. W. 1992. W. North American Journal of Fisheries Management 4:99-102. 1982. St. Bulletin 191. The removal method of population estimation. Strub. Schnute. A. A new approach to estimating populations by the removal method. C. C. Lockwood and J. R. Freshwater Res. 1991. and Cederholm. Seber. G. J. White. and Hesthagen. Inst. L. Iowa State University Press. S. G. Cochran. Schneider Chapter 7 13 . and W. Written 3/1976 by G. L. U. Iowa. F. Ricker. K. A new method for estimating population size from removal data. R. Department of Agriculture. P. Canadian Journal of Fisheries Aquatic Sciences 40: 2153-2169. and E. D. G.. Otis. Paul. and P. 1975. P. Burnham. by electrofishing in two small streams in north Norway.

Manual of Fisheries Survey Methods II January 2000 TOC NEXT PAGE PREVIOUS PAGE CITATION Chapter 7 14 .