Wednesday, December 15, 2010

Five more dyamic migration maps


We have posted five more dynamic migration maps on the "Occurrence Map" page. The new species are:

Willow Flycatcher
Eastern Bluebird
Blackpoll Warbler
White-throated Sparrow
Western Meadowlark

As before, comments are welcome below. Thanks to all for the comments on the first round of maps.

Marshall Iliff


  1. So far none of the occurrence maps are from waterfowl or shorebirds. I would expect maps for most of these species to be much rougher around the edges. Is that indeed the case? It might be interesting to throw up a waterbird and a shorebird in the next round to illustrate what happens with those types of species.

  2. Marshall - The Western Meadowlark animation seems to show Western Meadowlarks completely withdrawing from Idaho in winter. However, they are regular (if sparse) in winter at some winter locales in Idaho. For example in the Lewis-Clark Valley (north-central) and the Snake R. plain (southwest ). Some of these data should be in eBird. Any ideas as to why they aren't showing up here? thanks!

  3. RAZL DASL: The lack of waterbirds in the occurrence maps is no accident. They ARE rougher around the edges, and we know why. The suite of GIS layers we use for these models includes good (not great) info on terrestrial habitats, but very little on aquatic habitats. We actually have located and imported some great information on aquatic habitats and are looking forward to re-running the models using this info. This is a fairly long process though, since there is a huge amount of data processing to get these new variables into the locality variable dataset, followed by a huge amount of processing to re-run the models. It is a research direction for us and we are excited to tackle it, but it may take a half a year or more. At the same time, we hope to dial in the scale and produce more 3km maps, like the Wood Thrush one up now. Much to do! Thanks for the comments.

  4. Charles: eBird certainly has data on Western Meadowlarks wintering in Idaho. When I look at the map I do see some color persisting all winter long in at least the southern half of Idaho. This seems consistent with what you suggest.

    Regardless, the issue is one of scale, both spatially and in terms of color display. A small 5 km square valley might regularly have wintering meadowlarks, but when averaged with surrounding desert it gets lost with the 25 km square of bad meadowlark habitat. For this reason, prairies show up as hot white since the whole area is good habitat, but areas with isolated agriculture (e.g., Arizona) are just a pale orange wash.

    The other scaling issue is how do we display the very very high occurrences in north Texas in winter (where meadowlarks are abundant ad occur in flocks of hundreds seen from roadsides) to areas like central or northern Idaho where there might be a few known sots for them. The model sometimes does pick these up, since (as you point out) our committed eBirders have reported them using effort-based lists. But compared to occurrence in Texas they might be 30x or 50x or 100x (!) less likely. Is there a good way for us to convey those huge extremes in the scale?

    We think the answer is to give users the control to adjust the scale and look for those signals. On the models we display, not everywhere that looks black is equal to 0.0000 predicted occurrence. Giving users control would allow them to search for the areas where STEM predicts very rare occurrence and where it is actually 0.000.