Pond coefficients from other taxa increased mean error by 2.35% against separate fits in 18 prairie duck survey years, 2000–2019
We tested whether a pond coefficient learned from nine duck taxa could predict the tenth taxon as well as its own fitted coefficient. We used ten taxa and 24 fixed prairie regions from the Waterfowl Breeding Population and Habitat Survey. We fitted models to 23 annual transitions before 2000 and tested them in 18 complete survey years in 2000–2019. All models used previous duck shares and fixed spatial preferences. The transferred pond coefficient reduced mean squared Hellinger error by 9.98% against a model without ponds. Its nominal 95% test-year block interval was 6.32% to 13.82%. However, it had 2.35% more error than separate taxon coefficients, with an interval from 1.51% to 3.16%. Mean errors were 0.026544 and 0.025936. Northern shoveler and northern pintail had the largest losses from transfer. All eight planned model refits also had more error with transfer. This is a test of complete transfer of one habitat coefficient. It does not test partial sharing, a taxon with no bird history, future forecasts, or causal habitat effects. The intervals condition on fitted models and source estimates. They omit original survey error. The package has the unchanged source archive, code, provenance, predictions, figures, and repeat instructions.
waterfowl · habitat · species distribution · parameter transfer · prediction · wetlands · reproducibility
References
4- usesCurrent May pond changes reduced spatial prediction error by 12% beyond prior duck shares for ten taxa in 24 prairie strata during 18 test years, 2000–2019 sci:sha256:983106b74e74c7d6d426c6295a2a110b3e58b8602ef62d79ee94c9114252668b
We use its source selection, softmax benchmark, and conditional test-year block procedure. Its baseline and separate habitat models are reproduced. We add transfer of the pond coefficient from the other nine taxa.
- The pond gain depended on northern shoveler in fixed one-year forecasts for ten prairie duck taxa in 2000–2019 sci:sha256:df8cd3e9f89f5e59a149eedc66d6576f8e2f4fc0e0c998d2f2a34f58bc3cd8b3
This related benchmark tests prior ponds in forecasts of total abundance. Our test predicts current spatial shares after current ponds are known.
- doi:10.1111/ele.12757
This work describes joint models for species communities and variation in environmental responses. We test the narrower rule of one common pond coefficient.
- doi:10.1002/wmon.1074
This work uses waterfowl and habitat data for conservation planning in the Canadian prairie region.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- usesPartial sharing of pond coefficients did not lower mean spatial error for ten prairie duck taxa in 2000–2019
We reuse its retained source archive and adapt benchmark code under MIT terms. It left partial sharing untested.
- usesSeparate pond gain and loss slopes gave no clear mean gain in spatial predictions for ten prairie duck taxa in 2000–2019
We reuse its source archive, temporal selection, softmax model family and mean-error benchmark. We add prior pond levels and test separate change slopes.