Patterns from eBird -- Northern Cardinal

The patterns we see from eBird observations are often the result of two factors: detectability and movements. Detectability—the probability of observing a species when it is known to be present—varies as the seasonal behavior of the species varies. Your eBird observations are helping us to amass information on species detectability, which will ultimately help us refine our understanding of the occurrence patterns for all species. Note that most previous research on detectability has been specific to a certain season, so eBird data promise to shed new light on this familiar problem.

The animated occurrence map below is the first of its kind that we have shown on eBird, but we hope to provide more of these in the future. Our team has analyzed eBird observations along with landscape information such as climate, elevation, human population, and habitat. The results of these analyses involve tens of thousands of predictions at various points across the landscape, but are are summarized on an easy-to-understand map surface. The brighter colors indicate a higher likelihood of observing Northern Cardinal, and the duller colors indicate areas where they are less likely to be observed. Each surface shows the estimated probability of an observer recording at least one Northern Cardinal on a typical eBird traveling count (1 hour over 1 km at 7 am) on a given date. These estimates are averaged within a predefined grid square (50x50 km). To show how distributions evolve and change throughout the year we produce a sequence of 100 surface estimates spaced at approximately 3 day intervals.

Figure 1. Animated map of the predicted occurrence of Northern Cardinal in the Lower 48 United States in 2008. Some of the variations reflect minor differences in model performance, but others follow known biological patterns and surely represent changes in detectability.

By studying a species that is resident, we can examine temporal and spatial patterns of detection. This is important because with a better understanding of how birds are detected in the field we can better adjust estimates of actual occurrence for all species. Interestingly, the model results indicate significant variation in Northern Cardinal occurrence throughout the year. Figures 2a–2d show snapshots from different periods that highlight these changes.

Figure 2a. Predicted occurrence of Northern Cardinal in the Lower 48 United States on 5 April 2008.

The predicted occurrence of Northern Cardinal in April is very high and extends to the western and northern limits, surely because April is peak singing time for the species across much of its range. Cardinals are conspicuous since the males perch high and sing often, and the species is widely distributed in discrete territories.

Figure 2b. Predicted occurrence of Northern Cardinal in the Lower 48 United States on 17 June 2008.

During the nesting period cardinals sing less frequently and are more secretive while raising young. They are still present in the same territories, but are simply less conspicuous and harder to detect. Notice the overall darkening of the map compared with April, especially in the Adirondacks, higher Appalachians, and northwestern part of the range where the species' occurrence was lower to begin with.

Figure 2c. Predicted occurrence of Northern Cardinal in the Lower 48 United States on 3 September 2008.

Early September is a time when Northern Cardinals are molting, causing them to become more vulnerable to predators, and most become more secretive. The general appearance of this map is very similar to June, even though young are off the nests now and overall abundance is probably near its peak.

Figure 2d. Predicted occurrence of Northern Cardinal in the Lower 48 United States on 21 January 2008.

The final map (Fig. 2d) shows the model results for January. From December to February, cardinal occurrence does not change much. However, during this time the predictions for the species’ occurrence are interesting to explore. In winter they concentrate around bird feeders and in food-rich thickets, where it is possible to encounter flocks of 10 to 40, sometimes more. These numbers are never matched during other times of year, since cardinals are territorial and do not flock during other seasons. This flocking activity is a sign that cardinals do move around the landscape, and occur in slightly different areas and in different-sized groups at different times of the year, and may even concentrate in certain suburban areas--since these are the places with bird feeders and food-rich ornamental plantings. The predictions seem to recognize the high detectability of cardinal during these times.

Please remember, during our April fund-raising month, that your contributions help not only with running eBird day-to-day and with new initiatives (like our current effort to take eBird worldwide), but also with the research that produces these maps, which we eventually hope to provide for all species. If you believe in eBird and what we do, consider donating directly or supporting us by pledging to our World Series of Birding effort.

NOTE: Real gurus of bird distribution may notice some minor "errors" in the mapped range of cardinal above. Most striking, perhaps is the fact that cardinals are predicted to occur in portions of southeastern California, western Arizona, and southern Nevada where they do not occur. This is partly due to the fact that the similar habitat in these regions make it seem like cardinals "should" occur there; indeed a small (and declining?) population has blinked on and off on the California side of the Colorado River for several decades. More importantly, it also should be stressed that these models' ability to predict species distributions get "smarter" with more eBird data. We have made many appeals for how important data are from underbirded areas, like these southwestern deserts and the Dakotas (cardinals are not as widespread in the Dakotas as they are shown, either). With more data from these areas and more County Birding efforts from these parts of the country, these models will further improve.