Why Camera Photos Can Mislead a Deer Survey

Counting trail camera images gives you a photograph total. Estimating deer from that total requires assumptions that the memory card cannot confirm. The usual survey depends especially on equal triggering rates per animal for females and males. Equipment introduces another source of variation. Manufacturer websites list trigger delays spanning 0.1 to 1.0 seconds, allowing cameras sharing a location to produce different totals even when deer behavior stays unchanged.

The procedure originated in a paper published in 1997 by Jacobson and his coauthors. We have not read that original study. According to a paper published in 2011, counts of male, female and fawn photographs supply the demographic ratios. Point estimates come from scaling up the tally of males distinguishable as individuals. Antlers and body markings distinguish bucks, and the procedure assigns individual identities only to males. Those recognized bucks therefore set the scale for estimating the remaining deer.

Mississippi State University Extension illustrates the procedure in its P2788 guide. Its example follows a winter survey on 1,000 acres: cameras watched bait for 10 straight days, with a camera assigned to each 100 acres. The tally included 1,195 images of bucks, 2,403 of does and 1,293 of fawns. Antlers and body markings separated the photographed males into 31 distinct bucks. That works out to an average of 38.5 images for each recognized buck. The calculation assigns that same individual photo rate to females and fawns. Our repeat calculation agrees with the publication's arithmetic.

The guide warns that some states prohibit baiting. We explain the procedure here without recommending bait use.

Working through the formula shows that the recognizable male count drops out when calculating the sex ratio. For the doe estimate, multiply 2,403 images of females by 31, then divide by 1,195. Divide that estimate by 31 recognizable bucks and the 31 terms disappear, leaving 2,403 divided by 1,195: 2.01. Thus, the stated buck-to-doe result of 1 to 2 comes straight from photograph totals. It represents the herd's sex ratio only when females and males generate equal photo totals per animal.

Unequal trigger rates bias the female estimate in direct proportion to the difference. More frequent triggering by does pushes the estimate high; less frequent triggering pushes it low. The next table applies calculations to the P2788 example and contains no observations. Assume a doe produces 1.5 times the pictures per animal that a buck produces. Under that assumption, 2,403 female images represent 42 does. Yet the formula returns 62, overshooting by roughly 50%. The actual buck-to-doe relationship would be 1 to 1.3 rather than 1 to 2.0.

The study from 2011 provides a real-world check. Researchers used photographs from New York's Mianus River Gorge Preserve to test an adjusted version of the procedure. Females there appeared in pictures more frequently per animal. Correcting for that difference reduced the ratios of females to males while increasing the ratio of fawns to females. The researchers also found wide uncertainty under every approach. Results from a single preserve cannot establish a rule for your land. They demonstrate that an actual herd can violate the equal-rate assumption.

Our inventory covers 422 trail camera models produced by 16 manufacturers; we present grouped results only. For 98 models, manufacturers publish trigger times on their websites. Those stated delays span 0.1 to 1.0 seconds, and the median is 0.25. The broader pool also includes two manufacturer-site rows with notes referencing independent tests, plus 61 listings from sellers or reviewers. Those listings may copy a manufacturer's specification or give a test result. Together, the pool covers 161 models with delays of 0.05 to 1.5 seconds, a 30-fold difference. We measured none of these times ourselves.

Models also vary in sensing distance and lens coverage. Manufacturer specifications list distances of 70 to 100 feet across 56 models. Of those, 25 list 80 and 26 list 100. Seller and review listings expand the span to 45 to 120 feet across a pool of 106 models. Published view angles appear for just 34 models, spanning 38 to 360 degrees. The set of 98 published trigger times represents 13 manufacturers and 52 product families. Different models within the same product family often list an identical specification, so the sample does not consist of 98 independent camera designs.

A northern Scotland field experiment demonstrates how those differences can affect a count. Fieldwork spanned February to April 2023, with publication in 2025. The comparison used three models, with one unit per model at each of six locations, operating for 50 days. Most observations involved red deer. The camera with the higher technical specifications listed a trigger time of 0.1 second and a daytime detection range of 30 meters, approximately 98 feet. Advertised trigger times for the inexpensive and older units were 0.5 and 0.6 seconds; their ranges were 20 and 18 meters. Compared with the budget and older models, respectively, the camera with the higher technical specifications recorded 262% and 1,223% more distinct deer. The authors describe this gap as larger than any model difference in previously published comparisons. Trigger time, sensing distance and view angle all varied together, preventing the experiment from isolating the responsible specification. The setting also differs from whitetail habitat in timber: these were red deer on open terrain.

For the survey it describes, the extension guide places less weight on trigger speed. Deer at bait offer repeated chances for a picture, so it calls speed an unimportant camera feature for that method. Its directions prescribe a five-minute pause between photographs. Equipment differences therefore matter least when every station in a baited survey uses an identical model. They matter more when comparing cards from different models or interpreting a zero from an unbaited camera. Here, zero means the operating camera took no deer pictures.

Finding no deer pictures on a card offers little support for a conclusion that deer were absent. The final column identifies the information our catalog contains for each source of missed detections.

We collected the camera figures from manufacturer statements and published listings; we did not measure the specifications ourselves. A listing can reproduce a manufacturer's claim or describe an independent test. Our catalog does not distinguish between those possibilities. The 98-model summary of manufacturer pages excludes two manufacturer-page rows whose notes reference an independent test. Those two rows remain part of the combined 161-model results. Six rows combine stated specifications and measured results within a single field; all statistics omit those rows. P2788 supplies a single worked calculation, and the sensitivity sweep applies arithmetic to that example. Because we have not read the paper from 1997, our account of the procedure relies on the extension guide and the paper from 2011. The Scottish experiment concerns red and roe deer on open terrain, and it leaves the model names undisclosed.

We regard a camera tally as a measurement of photographs produced by specific equipment at specific locations. Turning that tally into a deer estimate requires explicit assumptions. For each station, hunters should log the model along with camera-nights, meaning the number of nights that each camera ran. A buck-to-doe calculation from cards should be treated as a ratio of photographs. Our recommendation is to keep the model identical across stations. For anyone conducting the survey, the extension recommends assigning one camera to each 100 acres and operating it for 10 to 14 straight days. When comparing unadjusted totals between models or years, explain what changed. Log a blank card as zero deer photographs during camera operation. Avoid interpreting that zero as deer absence, as explained in our scouting log guide.