Tuesday, July 30, 2013

Autumn Leaves are for the Birds: A Preliminary Study of Red Autumn Foliage & An Insectivorous Migratory Bird Species


Coevolution Theory suggests autumn leaf color change in deciduous tree species may be an adaptive characteristic that functions as a signal to potential plant predators and parasites that the tree is well defended. Several additional adaptive plant-predator hypotheses have been proposed, however most do not seem to be well supported due to a number of different reasons.   A potential better explanation is that leaf color change may be a means to attract insectivorous birds.I use a combination between citizen science reports of forest color and bird frequency and digital color analysis of available satellite imagery to examine the relationship between a bird species and autumn leaf color change. I found that image analysis illustrated a positive change in red coloration over the course of the fall, that Ruby-crowned Kinglet frequency positively increased between September and October, and there was a positive correlation between the increase of kinglets and red foliage.


Color change in autumn deciduous foliage was once considered a non-adaptive characteristic due to leaf senescence.  Archetti (2000), however, proposed the Coevolution Theory that suggests autumn leaf color change in deciduous tree species may be an adaptive characteristic that functions as a signal to potential plant predators and parasites that the tree is well defended. Since then several additional adaptive plant-predator hypotheses have been proposed.
 In addition to plant-predator hypotheses plant physiologists have proposed that color change in autumn foliage serves a physiological purpose for the tree. 

•Color as Signal to Insect Leaf Predators/Parasites

Several hypotheses have been proposed which suggest that leaf color may act as a signal to insects (Archetti 2000; Archetti et al. 2009; Hamilton and Brown 2001) or that insects exhibit preference for leaves that have undergone color change (Archetti 2007). However these do not seem to be well supported.

-Hamilton and Brown (2001) found that tree species that express stronger autumn coloration had greater diversity of specialist aphid species.

-Archetti (2008) illustrates that weak trees also produce bright colors, which he suggests means that these weak trees can cheat to take advantage of the color signals.

-No herbivorous insect has yet to be shown to possess photoreceptors for the color red (Chittka and Döring 2007).

-White (2009) suggests that aphids should prefer green and yellow leaves, because they have more nutrients available to the insects than red leaves.

•Color as a Signal to Migratory Insectivorous Birds

Stiles (1984) proposed the idea that plants may use red color cues to signal to birds the availability of fruit and the work of Schmidt and Schaefer (2004) reinforce this idea. They found that Blackcaps showed preference for the red fruits over other colors of fruit, even though they had never experienced any colored fruits before.

A study by Burns and Dalen (2002) suggests that plants with dark berries may rely on the change in fall leaf color, or what Stiles refers to as “foliar fruit flags”, to create a highly visible contrast to attract birds.

Anthocyanin pigments, responsible for red coloration in fall leaves, are actively produced only in autumn, unlike other leaf pigments which become un-masked with the breakdown of chlorophyll (Chittka and Döring 2007; Archetti et al. 2009). The production of this pigment requires energy during a period of time when trees are actively reabsorbing nitrogen and other nutrients from leaves.

Only about 10% of temperate tree species exhibit red coloration in autumn, though this varies regionally, and within tree species there is variation in the production of of different colors in autumn (Archetti et al. 2009).

 I hypothesize that the red coloration in deciduous tree species is not a signal to aphids or other insect predators and parasites, but is rather produced as a response to the presence of these insects and serves as a signal to attract fall migrating insectivorous bird species to function as a control mechanism for insects. 


•Fall Peak Leaf Coloration

I was able to obtain a complete listing of the 2012 peak fall leaf coloration reports through contact with an Ohio Division of Forestry forester (Burdick pers. comm., 2012). Each site was evaluated and scored on a scale of 1 to 5 where 1= “Mostly Green,” 2= “Changing,” 3= “Near Peak,” 4= “Peak,” 5= “Fading.”  The score for each site was assessed once per week between mid-September and the end of October.
I evaluated the number of different sites with each score for each week, and then plotted this to determine when peak autumn foliage occurred in Ohio (Fig. 1).

Figure 1: Bar chart illustrating the color status of sites over the course of the 2012 autumn season in Ohio.  This figure shows how the number of sites shifted from “Mostly Green” in mid-September to “Peak” in mid-October.  Peak foliage coloration is determined to have occurred around Oct. 17th.

•Evaluating Bird Data

Bird data was obtained using the online citizen science website, eBird (2012).  I conducted a search for bird species reported in Ohio through the “Bar Graphs” feature of the site.  I refined the search to birds reported only in 2012.  This produced a list of 331 species.  I narrowed the list based on the following criteria: primarily arboreal species, migratory in fall, insectivorous, foliage-gleaning foraging behavior.  The remaining list was then searched using the bar graphs of occurrence to find species whose migration coincided with peak leaf color graphs.  The result was one species; the Ruby-crowned Kinglet.  Other species may also meet these criteria, however the Ruby-crowned Kinglet was chosen as a representative species based on perceived best fit to the criteria.

 Once the Ruby-crowned Kinglet was chosen as a representative species a search was conducted to find its frequency within Ohio between September 1 and November 1, 2012 (Fig. 2). Frequency is defined by eBird (2012) as the number of checklists reporting a given species out of the total number of checklists reported.

Figure 2: Bar graph illustrating the frequency of reports of Ruby-crowned Kinglets in Ohio for days in September & October 2012.  A polynomial trend line has been added to illustrate to show peak occurrence of the species during its migration.

•Foliage in Ohio

Color satellite images were obtained from the DigitalGlobe website.  Images were searched using the criteria that they were from Ohio and produced between September and October 2011 & September-October 2012.  DigitalGlobe satellite scans appear as rectangular-like polygon shapes.

 Once the satellite scans were obtained they were then analyzed for color based on a modification of a protocol established by Murakami et al. (2005).  Their protocol proposes the use of Scion Image software, which has since been discontinued and re-configured into a newer software package, ImageJ (Rasband 2012), which was developed by the National Institutes of Health for biological imaging analysis (Ferreira and Rasband 2012).

 Each satellite image was imported into Adobe Photoshop 7.0.   Color saturation was maximized using the image adjustment feature (fig. 3).


                        Figure 3: Example of a satellite image (e.g. Shawnee State Forest in Ohio) maximized for color saturation in Adobe Photoshop 7.0.
Each saturated satellite image was saved in the JPEG format and then imported into ImageJ.  Once in ImageJ the image was manipulated by splitting the color channels into individual 8-bit images representing red, green, and blue.

The blue image was discarded since I was interested primarily in the amount of red and green foliage represented in each image. 
I then worked with the two remaining images individually to analyze the amount of the target colors present in each image.  Using the ImageJ software I created a threshold image for the color.  This resulted in a black & white image where the target areas were white and negative space was black (fig. 4). 
         Figure 4: The single channel image (e.g. red) is converted to a threshold black and white image.  Negative space is colored black. Target areas are white.
In order to analyze the image the threshold image was inverted so that target pixels were colored black and negative space was colored white, and then a particle analysis was conducted on the image.  The particle analysis produced a percent of image area value for the color being analyzed.
This process was repeated for the green channel image, and each satellite image was treated in the same manner.

 The percent area of red and the percent area of green was recorded for each satellite image, and then a ratio of red:green was calculated for each image.  Images with more red foliage would presumably have higher ratios of red:green, and images with more green foliage would have lower ratios of red:green.  The ratios were plotted and then compared to the frequency of Ruby-crowned Kinglets observed in Ohio on the days that the satellite images were made.


A plot of the red:green values, obtained through image analysis, showed a positive increasing trend over the course of the two years. Additionally, a plot of the frequency of occurrence of Ruby-crowned Kinglets in Ohio in the fall for two years also showed a positive increasing trend (fig. 5), suggesting that the species increased over the course of the season due to migration, and this increase coincided with the increase in red foliage in Ohio forests.

A correlation analysis of the two data sets  for both years resulted in r-values of 0.344423 and 0.520657  for 2011 & 2012 respectively suggesting a positive correlation between the increase in red foliage and the occurrence of Ruby-crowned Kinglets in Ohio.

Figure 5: A plot of red:green ratio values from color analysis of satellite images (green line) illustrating a positive increasing trends from September through October for 2011 & 2012.  The plot of Ruby-crowned Kinglet frequencies shows an increasing trend from September through October for both years as well.


This study establishes a possible correlation between the presence of red foliage in autumn and a migratory insectivorous bird species, the Ruby-crowned Kinglet. This preliminarily may support the idea that trees develop red foliage as a means to attract birds, specifically Ruby-crowned Kinglets, during fall migration. 

The development of red foliage may be in response to insect parasites/predators, like aphids, though further research is needed to elucidate this influence on leaf color development. If this idea is supported, then it expands Stiles foliar fruit flag hypothesis (1984) beyond the idea that plants attract birds through the use of red foliage to help disperse seeds. Trees may also attract birds to help control insect loads during a time of year when they are most stressed.

While it is possible that the presence and increase of migratory Ruby-crowned Kinglets in Ohio is coincidental as leaves reach peak coloration in the autumn, the positive r values suggests that there is at least some degree of correlation between the two.

The use of satellite imagery to study fall leaf coloration change is a useful tool.   However it is a rather crude tool because I was not able to differentiate between forest types or tree species.  Additionally, because I was reliant on available satellite scans temporal and spatial coverage was rather limited.

Citizen science bird data has been utilized extensively for many different kinds of studies, however it is not without its limitations.  While eBird allows data to be searched at different temporal and spatial scales researchers are still limited based on where birders have chosen to make & report observations .  This means that some areas are not covered and reported, or have limited coverage, if they are not well birded.



Archetti, M.  2000.  The origin of autumn colors by coevolution.  Journal of Theoretical Biology. 205:625-630.

Archetti, M.  2007.  Autumn colours and the nutrient retranslocation hypothesis: A theoretical assessment. Journal of Theoretical Biology 244:714-721.

 Archetti, M. 2008. Decoupling vigour and quality in the autumn colours game: Weak individuals can signal, cheating can pay.  Journal of Theoretical Biology. 256:479-484.

 Archetti, M., T.F. Döring, S.B. Hagen, N.M. Hughes, S.R. Leather, D.W. Lee, S. Lev-Yadun, Y. Manetas, H.J. Ougham, P.G. Schaberg and H. Thomas.  2009.  Unravelling the evolution of autumn colours: an interdisciplinary approach.  Trends in Ecology and Evolution. 24:166-173.

Burns, K. C. and J. I. Dalen. 2002. Foliage color contrasts and adaptive fruit color variation in a bird-dispersed plant community. Oikos. 96:463-469.

 Chittka, L. and T.F. Döring. 2007.  Are autumn foliage colors red signals to aphids?  PLoS Biology. 5:1640-1644.

 eBird. 2012. eBird: An online database of bird distribution and abundance [web application]. eBird, Ithaca, New York. Available: http://www.ebird.org. (Accessed: Nov. 2012).

Ferreira, T. and W.S. Rasband. “ImageJ User Guide — IJ 1.46”, imagej.nih.gov/ij/docs/guide/, 2010–2012

Hamilton, W.D. and S.P. Brown. 2001.  Autumn tree colours as a handicap signal.  Proceedings of the Royal Society of London B. 268:1489-1493.

Murakami, P.F., M.R. Turner, A.K. van den Berg, P.G. Schaberg. 2005. An instructional guide for leaf color analysis using digital imaging software. Gen. Tech. Rep. NE-327. Newtown Square, PA: U.S. Department of Agriculture, Forest Service, Northeastern Research Station.

Rasband WS. ImageJ, U.S. National Institutes of Health, Bethesda, Maryland, USA, imagej.nih.gov/ij/,1997–2012.

Stiles, E.W. 1984. Fruit for all seasons. Natural History. 8:43-54

Schmidt, V. and H. M. Schaefer. 2004. Unlearned preference for red may facilitate

recognition of palatable food in young omnivorous birds. Evolutionary Ecology Research. 6:919-925.

White, T.C.R.  2009. Catching a red herring: autumn colors and aphids. Oikos. 118:1610-1612.