Skip to Main Content
banner NOAA logo            NOAA's National Marine Fisheries Service
            NOAA Fisheries Toolbox

row of waves


 Model List

CSA Model

Line Art Drawing of a FishCollie-Sissenwine Analysis


|  About  |  Downloads  |  References  |  Version History  |



Version 4.3 (January 2014)

Core Application


The Collie-Sissenwine Analysis (CSA) model (sometimes called catch-survey analysis or the DeLury model) is a relatively simple two-stage stock assessment model that estimates abundance, fishing mortality and recruitment using total catch numbers and survey data (Collie and Sissenwine 1983; Conser 1995).  The “recruit” stage group consists of animals that recruit at, just before, or during the current time step.  The rest of the population comprises the “post-recruit” stage group.  The two stages may correspond to age groups, length groups or any other natural division (e.g. genders in hermaphroditic species).  Typically, both groups are assumed fully available to the fishery but this assumption can be relaxed in practice if fishing mortality rates are viewed as rates for fully recruited animals.


CSA (Version 4) was updated during 2013 and previous versions are no longer supported.  The update uses maximum likelihood rather than weighted sums of squares to estimate parameters.  The user must supply survey and year specific CVs that measure the precision of survey observations in likelihood calculations.  The updated version uses Baranov’s catch equation exclusively to simulate the population.  Pope’s approximation is no longer available because accuracy of the approximation degrades at high mortality rates.  As in previous versions, natural morality in each year is specified by the user and not estimable in the model.  The updated model does not allow for estimation of process errors.  Process errors tend to be difficult to estimate because they are typically aliased (cannot be distinguished from) time varying survey variances, natural mortality, recruitment and selectivity parameters already in the CSA model. Multiple surveys and survey types are now allowed, with time of the year that each survey occurred required.

The calculation engine is written in AD Model Builder. The program is based on previous implementations DELPOP8 by Dr. Jeremy Collie at the University of Rhode Island, an APL version by Dr. Ramon Conser retired from SWFSC, and Fortran versions developed by the Population Dynamics Branch at NEFSC. 


Back to Top 




Collie, J. S. and M. P. Sissenwine, 1983. Estimating population size from relative abundance data measured with error. Can. J. Fish. Aquat. Sci., 40: 1871-1879.


Back to Top 


Version History


CSA Version 4.2.2 (September 2013)

Completely new version of CSA written in ADMB. New features include ability to use multiple surveys of different types, the timing within the year of each survey is required, uncertainty associated with each survey observation is required, and removal of process error estimation.


CSA Version 3.1.1 (October 2008)

Program fix for calculations when the number of completed iterations does not equal the number of requested iterations.  Prior version 3.1.0


Back to Top 


Older Version Changes 


 CSA Version 3.03 (May 2005)

  • CSA has been updated to remove the time of the survey from the process equation when an exact solution of the catch equation is selected for either the observed error only or process & observed error cases.  After review, it was concluded that the time of survey can not be adequately handled in the process equations as the model is presently designed.  Users should adjust catch data to be on a consistent time interval with the survey indices entered.  Users should enter zero for both time of catch and time of survey.  If users are using the Pope's approximation, then this change is not affected.

  • A new feature has been added for bootstrapping which allows users to drop from the bootstrap realizations cases where the fishing mortality is infeasible.  

  • Additionally, there were changes to the output report to echo input values not previously reported.


CSA Version 3.01 (March 2005)

  • Added a new enhancement that incorporates an exact solution of the catch equation into both the observed and process & observed error models. The Pope approximation is available for comparison with earlier models.

  • A bug was fixed that now allows users to constrain upper & lower limits on catchability.


Back to Top 


NOAA Fisheries Toolbox Version 3.1

website last modified May 16, 2016


Printer friendly view

Link disclaimer | Email webmaster | Privacy policy |     File Modified Mar 28, 2008