Using morphometrics to sex adult and juvenile Soras (Porzana carolina)
Data files
Mar 11, 2024 version files 21.86 KB
-
README.md
-
sora_data.xlsx
Abstract
Determining the sex and age of individuals can be an essential element of conservation management, wildlife monitoring, and demographic analysis. For many members of the family Rallidae, distinguishing between males and females is challenging, even when the bird is in the hand. The Sora (<em>Porzana carolina</em>), a secretive rail that occupies freshwater wetlands throughout the United States and Canada, represents a species that is challenging to sex in the field. Morphometric measurements can help sex birds of an array of species, including rails. However, no comprehensive morphometric model has been fully validated for sexing Soras. We used DNA analysis to confirm the sex of Soras captured in the field and logistic regression models to determine which morphological features were the best predictors of sex. Measurements from 108 Soras (31 hatch year females (HY-F), 29 hatch year males (HY-M), 22 after hatch year females (AHY-F), and 26 after hatch year males (AHY-M)) were used to create our logistic regression model. Color definition and connectivity of the auricular patch to eye or nape was used as an additional characteristic in adult birds. Our top-ranked model was further validated using a sample of 72 individuals exhibiting intermediate traits that would be particularly challenging to distinguish in the field. Our top performing model incorporated culmen length and tarsometatarsus length as the features most predictive of sex and had an overall accuracy of 85%. If higher accuracy is desired, an inconclusive band, which eliminates birds of low model score, i.e. scores indicative of inconclusive sex (below + or - 1.2), can be used. The accuracy of remaining birds (75% of sample) will be increased to 95%. Our model shows that simple measurements of culmen and tarsometatarsus is useful in discriminating the sex of a large percentage of live-caught Soras. This morphometric model will facilitate further demographic studies of this species and may be useful in designing morphometric studies of other species in the family Rallidae.
README: Using morphometrics to sex adult and juvenile Soras (Porzana carolina)
Sora were live caught on the Patuxent River in 2018 and 2020. Measurements of their weight (g), fat score (0-5) , culmen (mm), tarsus (mm), and toe (mm) were taken along with a blood sample. The blood samples were sent to DDC Veterinary (DNA Technology Park, 1 DDC Way, Fairfield, OH 45014) where a DNA blood test based on two conserved chromo-helicase-DNA (CHD) genes located on the avian sex chromosomes was used to determine the sex of each individual.
Measurements from 108 Soras (31 hatch year females (HY-F), 29 hatch year males (HY-M), 22 after hatch year females (AHY-F), and 26 after hatch year males (AHY-M) captured in 2018 were used to create our logistic regression model in Matlab. Our top-ranked model was further validated using a sample of 72 individuals captured in 2020 exhibiting intermediate traits that would be particularly challenging to distinguish in the field. Our top performing model incorporated culmen length and tarsometatarsus length as the features most predictive of sex and had an overall accuracy of 85%. If higher accuracy is desired, an inconclusive band, which eliminates birds of low model score, i.e., scores indicative of inconclusive sex (below + or - 1.2), can be used. The accuracy of remaining birds (75% of sample) will be increased to 95%.
Description of the data and file structure
Our raw data Microsoft Excel file shows the results of the DNA sexing done on live-caught Sora by DDC Veterinary (DNA Technology Park, 1 DDC Way, Fairfield, OH 45014) from 2018 and 2020. For each individual, it shows the year they were caught, the USGS band number, the researcher’s guess at the sex, the actual sex determined by DNA sexing, hatch year status, weight (g), fat score (0-5), and culmen (mm), tarsus (mm), and toe (mm) measurements.
The raw data file sora_data.xlsx has the following columns:
- CaptureYear: capture year for the sora
- Band: band number of the captured sora
- OurGuess: sex identification determined by our banding researchers
- ACTUALSEX: sex determination from DNA testing
- Y: main response variable coded as actualsex male: = Y = 1 and actualsex female: = Y = 0
- Age: age group AHY = after hatchling year and HY = hatchling year
- Age01: numerical coding of age with AHY coded as 0 and HY coded as 1
- Cul: Culmen length in mm
- Tar: Tarsometatarsus length in mm
- Toe: middle toe length in mm
- Wt: mass in grams
- Fat: not used, but on a scale of 0-5 (0 being no visible fat to 5 being large amounts of visible fat)
Data was derived from the following sources:
Data are from DNA sexing done on live-caught Sora by DDC Veterinary (DNA Technology Park, 1 DDC Way, Fairfield, OH 45014) paired with measurement data taken by researchers at Maryland-National Capital Park and Planning Commission’s Patuxent River Park.
Code/Software
Matlab versions 2019b, 2022b, and 2023a were the primary software tools used in the analysis
Scripts assume that all scripts and supporting functions reside in the same folder. It is also assumed that the sora_data.xlsx data file resides in the same folder as the scripts.
Methods
Data was collected at Maryland-National Capital Park and Planning Commission’s Patuxent River Park in Jug Bay wetlands. A morphometric model was created using machine learning algorithms and software including: JMP 14® software (SAS Institute, Inc., Cary, NC), Matlab 2019b (The MathWorks Inc., Natick, MA), and scikit-learn 0.23.2 (open-source python package). The preferred model as selected by Akaike Information Criterion (AIC), Schwartz Information Criterion (SIC), and cross validation.
Usage notes
Microsoft Excel (open-source alternative: Google Sheets), MP 14® software (SAS Institute, Inc., Cary, NC), Matlab 2019b (The MathWorks Inc., Natick, MA), and scikit-learn 0.23.2 (open-source python package).