An evaluation of methods used to estimate carcass composition of common eiders Somateria mollissima

• Main Authors: Jamieson, Sarah E., Gilchrist, H. Grant, Merkel, Flemming R., Falk, Knud, Diamond, Antony W.
• Format: Article in Journal/Newspaper
• Language: English
• Published: 2006
• Subjects: Body condition; Carcass composition; Common eiders; Estimation models; Somateria mollissima; Common Eider
• Online Access: https://pure.au.dk/portal/da/publications/an-evaluation-of-methods-used-to-estimate-carcass-composition-of-common-eiders-somateria-mollissima(52ac4272-afb6-4fb0-8104-ca7adceeb929).html; https://doi.org/10.2981/0909-6396(2006)12[219:AEOMUT]2.0.CO;2; http://www.scopus.com/inward/record.url?scp=33745816667&partnerID=8YFLogxK

To examine how endogenous reserves may influence avian life history, it is often necessary to quantify carcass composition. However, proximate analyses are expensive, time-consuming and difficult to perform under field conditions. Consequently, carcass composition is often estimated from easily measured data. We evaluate methods of estimating carcass composition of the common eider duck Somateria mollissima. We measured, dissected and completed proximate analyses of 92 eiders. Predictive models were derived using multiple regressions of 70 birds, while the remaining 22 were used as an independent test of the models. Each model's accuracy was evaluated by comparing estimates against known values of protein and lipids, using root mean square error (RMSE). Abdominal and leg fat pad mass were highly correlated with total lipid (r = 0.92), and body mass was highly correlated with total protein (r = 0.80). Models that used body mass, fat depots and/or muscle group data were the most accurate (lipids adjusted R 2 = 0.93, RMSE = 14.60; protein adjusted R 2 = 0.74, RMSE = 11.14). By using these equations it is possible, using dissection data, to accurately estimate carcass composition of eiders. If dissection data are not available, one can still estimate carcass composition using equations that require only morphometrics although in our lipid analysis such equations had relatively low accuracy (lipids adjusted R 2 = 0.54, RMSE = 32.74).