# Assessing the Effects of Climate Change on Egg Production by Backyard Chickens

Discussion in 'Chicken Behaviors and Egglaying' started by FishChicken, Nov 23, 2014.

1. ### FishChickenOut Of The Brooder

11
1
26
Nov 23, 2014
Introduction

My overall objectives with this observation study were to: (1) quantify the factors that are known to affect the probability of egg laying for my backyard chickens, (2) determine how these factors affect the probability of egg laying, and (3) if temperature is important to egg production, then how might average air temperature increases affect the egg production by my chickens.

Methods
I collected egg counts on my six birds. The birds have always been kept under natural lighting conditions with continual access to food. The birds are allowed to roam daily throughout my 110' x 110' yard. The flock consists of 1 Buff Orpington, 1 Barred Rock, 2 Ameraucana, and 2 Australorps.

In addition to daily egg counts collected at 4pm each day, I also kept track of the birds age (days - the same across all birds), mean air temperatures (C), and the day length (hours). Day length was calculated using an algorithm developed by NOAA to account for the geospatial location of my chicken coop.

I measured the probability of a bird laying an egg each day using a logistic regression, where the daily counts of eggs are assumed independent and binomial distributed. I also use Akaike's Information Criterion (AIC) to determine the best functional form (e.g., quadratic, linear, logarithmic) and the relative importance of each variable to the probability of egg laying (If you want details ask).

The model for the probability of egg laying, p(egg), may be expressed as:

p(egg) = logit^-1[ B0 + B1*TempC + B2*Daylength + B3*Daylength^2 + B4*log(Age) ]

Results

All three of the well-known factors on egg production were determined to be important by AIC model selection. So, I won't go into details here, but if you are interested please do ask. At any rate, day length was the most important variable, followed by bird's age, and then average daily air temperature (Figure 1). The functional form of day length was best expressed as a quadratic, whilst age best followed the natural logarithm, and temperature was best expressed as a straight line. The model explained ~ 50% of the null deviance (variation in the data), and there was no indication of over-dispersion (lack of fit from a binomial distribution). Perhaps most notably, these findings highlight that simply using a light bulb for short periods in the evening hours may be the best way to boost egg production during the winter months. In other words, my observational study here, is matching expectations of what is known about chicken egg production, which is a good thing.

On the logit scale here are the estimated model coefficients...

Coefficients: Estimate SD
B0 - Intercept -12.390 1.161
B1 -TempC
-0.044 0.007
B2 - Daylength
2.623 0.206
B3 - Daylength^2
-0.086 0.008
B4 - log(Age)
-0.987
0.061

These values may be used to estimate the probability of egg production of your own backyard chickens. It would be interesting to see how the model performs on another flock of birds in another location.

Figure 1, Effects plots for my backyard chicken egg-production model.

So, here is the output of the statistical model compared to the observed proportions of eggs that were laid by my chickens. For this comparison, I used only the first year of data to estimate the model parameters to then predict the second year of data; doing so allowed me to assess model performance on an independent data set. Overall, the model that included only 3 factors to predict egg production performed really well when predicting egg counts in the following year (Figure 2).

Figure 2. Illustration of modeled egg production versus the observed egg production by my six chickens.

So, given that the model is temperature-dependent one can use the model in conjunction with increased temperature projections to estimate how average temperature increases might affect egg production. Changing the observed temperatures at my coop by +0, +1, +2, and +3 C provides a set of scenarios whereby average temperatures are increased, but variability in temperature is maintained at current levels. Applying the model to these temperature scenarios allowed me to estimate how temperature increases might affect the production of eggs by my backyard chickens.

I found that each bird might be expected to lay 3 eggs less per year at +1C, 5 eggs less per year at +2C, and 7 eggs less per year at +3C, indicating that any effects of climate change on egg production by backyard chickens are likely to be small (Figure 3).

Figure 3. The projected effects of average increases in air temperature on the egg production of backyard chickens.

Conclusions

Just three factors were needed to capture about 50% of the variation in the egg production of my backyard chickens. In order of importance these factors were day length, the bird's age, and average daily temperature. Although higher temperatures are expected to decrease egg production, the magnitude of this reduction is expected to be small, and can likely be easily offset by simply using a light bulb, or if possible, changing diet, or increasing food intake.

FishChicken

Last edited: Nov 23, 2014
2. ### hennibleOverrun With Chickens

Welcome to BYC
Yes it was interesting, thanks for posting.

1 person likes this.
3. ### centrarchidChicken Obsessed

18,407
3,232
486
Sep 19, 2009
Holts Summit, Missouri
FishChicken,

You need to repeat experiment over multilple years. Additionally it would be better is a lot more hens would be used. Your findings suggest a lot but you do not have enough data to support such with confidence.

1 person likes this.
4. ### sunflourFlock MasterPremium MemberProject Manager

14,802
6,613
506
Jan 10, 2013
Macon,GA

1 person likes this.
5. ### FishChickenOut Of The Brooder

11
1
26
Nov 23, 2014
Centrarchid,

I beg to differ. My study started with the a priori information that day light, temperature, and bird age affect egg production. This has been well established in the agricultural literature for decades, if not centuries - so repeated annual trials to convince me of my study's conclusions are moot. Had my study not been able to detect these effects then I would be concerned. Also, the daily egg counts may be considered independent samples/trials. Under this assumption, I have over 800 repeated samples in relation to the effects I've considered. Thanks for the comment! But I don't really think more samples are needed. Besides this is a common flaw with all observational environmental studies that cannot ever be repeated exactly...

More hens would be good, but then with more hens you get 'large-flock effects', and so my inferences to backyard chickens would be weaker. Perhaps what would be best is data on multiple flocks of backyard chickens in other locations (which I mention), we could then add flock as a random effect in the model to make inference to a greater population of backyard flocks :^).

Feel free to share your egg count data and I will add to my data set....

Thanks Again!

Last edited: Nov 23, 2014
6. ### centrarchidChicken Obsessed

18,407
3,232
486
Sep 19, 2009
Holts Summit, Missouri
More small flocks would be how you increase sample size. Your sample size is as is has only one replicate. Conditions of the flocks maintenance would have to be standardaized otherwise you number of flocks will need to be larger still. Also control for age of birds or treat them in the analysis as a function of age.

Literature I am familiar with dealing with environmental and agricultural issues also always involve multiple locations over multiple years.

Sadly, I do not keep data of the quality you need as flock too big and husbandry conditions are very different. Parties like me will also be a pain because of the importance natural forages has with respect to nutrition, The forage quality even on my site varies greatly as a function of season and year as well as timing of mowing events.

7. ### FishChickenOut Of The Brooder

11
1
26
Nov 23, 2014
True, inferences from my study may only apply to my flock, and since repeated samples on my flock are not possible, this is not a concern. However, given that my findings from my flock match expectations throughout the ages, my findings may also apply more broadly to other flocks of chickens. Increasing the number of flocks in the study would improve the estimates, especially the expected error. Including different diets in the estimate would be implicit with including different flocks as a random effect in the model, either way not sure this is necessary, or would drastically change conclusions. It would be nice if more people collected and shared egg count data, we might find something new, or at a minimum, really nail-down the estimate of the effects on egg production for back yard producers.

If only we could include flock number, flock size, location, temperature, egg count, date, elapsed time, and bird age in the data set.

In the end, the amount of caution you should employ depends on what you want to do with my model results, and where you want to draw your inferences. I started with a priori expectations, confirmed these expectations, and then projected these expectations given a set of plausible assumptions. In the end, I let my data speak, and that is what they had to say...

Cheers,
FishChicken

Last edited: Nov 23, 2014
8. ### centrarchidChicken Obsessed

18,407
3,232
486
Sep 19, 2009
Holts Summit, Missouri
You are pushing me now. You are tossing statistical terms about like a statistician that needs a translator to explain concepts to the reader.

The purpose behind putting such results out there is to help others understand what might go on with their flocks. If data holds only for your flock then effort put into study may not be of value to others.

Your results seem too perfect. Based on what I know you would expect egg production to drop at both high and low temperature extremes. This means you would have higher probability of egg production at some intermediate range of temperatures. You data presented in the upper left graph of Figure - Effects on Chicken Egg Production indicates a simple linear relationship where egg number is negatively correlated with temperature. My birds, which are genetically and age-wise quite variable, show a major drop in egg production at the low end of the spectrum you include in your analysis. Mine effectively stop when at even lower temperatures..

9. ### FishChickenOut Of The Brooder

11
1
26
Nov 23, 2014
Sorry to seem pushy Centrarchid! It certainly wasn't my intention. What I was trying to say is that you're basically going to need a whole lot of evidence or a very clever experiment to show that day length, temperature, and a bird's age are NOT important to chicken egg production. So taking more samples to show that they are important to egg production is pointless. However, to better improve the estimate of the expected error in egg production across many chicken flocks - you hit right on the head - need more data on other flocks...

Also, I agree one might expect temperature to be parabolic - or quadratic whereby egg production falls above and below an optimum temperature. But the AIC results were equivocal between the model with temperature expressed as a straight line or as a quadratic. Since the quadratic requires an extra term/coefficient I choose the parsimonious model where temperature was more simply expressed as a straight line. The effect of temperature is marginal, this is evidenced by the wide confidence intervals about the expected trend (Figure 1 - upper left panel)

The model is actually far from perfect. It explains just 50% of the variation (which isn't bad for using just 3 independent factors), but still we're only capturing half the variation. Also, there is systematic lack of model fit in winter time (see Figure 2). Much of this lack of model fit to the data is presumably due to a threshold-like effect of daylight on egg production, which is not fully captured by my statistical model (Figure 2). Namely, below about 9 hours of daylight I actually observed 0 eggs whereby p(eggs)=0, but the model estimates some non-zero probability. Its where the model doesn't fit the data that we learn the most...

Sorry, If I used overly statistical terms, I was under the impression that you might know what I meant. Im trying to keep it simple...

Thanks again,
FishChicken

Last edited: Nov 23, 2014
10. ### centrarchidChicken Obsessed

18,407
3,232
486
Sep 19, 2009
Holts Summit, Missouri
I understand the statistics, but suggest you use a language that is suitable for a broader audience. You are also putting a lot of weight into the statistical significance even though design does not appear appropriate for teasing out what is biologically significant. Too many things are varying together,, especially for the sample size and the possibly variables you could control for. Photoperiod, temperature and cohort age all are varying in a manner making simply date and cohort age being just as valid as predictors of egg production. Model would certainly be simpler