Ive been consulting with a National Park on how to improve estimates of visitor usage. they have door mounted meters but these are not considered very accurate. Here is a summary of my sample design:
SUMMARY: National Park Administrators are required to file monthly reports of visitor totals at various park facilities such as visitor centers. Although door mounted meters can and are used to estimate these totals, there are issues with the quality of meter counts. Meters are known to both under-count and over-count under some circumstances.
Human observers can provide a more accurate count at much greater cost, albeit also with associated sampling error. However human observers will not count staff and will count chest carried infants, two known examples of how meters can fail to count properly. We propose a double sampling design to analyze the relationship between meter counts and simultaneous human counts. We will estimate the population ratio of total visitors as observed by humans versus the total as recorded by the meters. This ratio estimate allows for calibration of future meter counts. The easy, cheap meter counts will acquire at least some of the accuracy of expensive and time consuming human observer counts. The design calls for 100 hours of meter counts, stratified by visitor center and peak hours. From this primary sample a secondary sample is selected, consisting of 20 hours of ranger observation.