Our new paper demonstrates how applying novel sampling gear combinations can provide statistical replication that is necessary for fitting a class of hierarchical models of occurrence and abundance that use detection/non-detection data. We apply this method to marine reef fish to describe patterns in occurrence and abundance and account for incomplete and variable detection. The modeling methodology has limited previous use by fisheries researchers because common distributional assumptions are violated by invasive sampling techniques often used for studying fish. However, collecting data with a combination of invasive and non-invasive sampling gears can meet the assumptions of these models, making them available for fisheries research, monitoring and stock assessment. For a copy, please click this link or send me an email (firstname.lastname@example.org).
The problem is that hierarchical models that use detection/non-detection data, such as occupancy models, require that replicate samples vary according to a binomial distribution. This means that each fish is independent and each replicate sample is independent. In fisheries research, this assumption is hard to meet because the methods we use to sample fish are usually invasive. For example, if we collect two consecutive replicate samples of the same population with electrofishing, the vulnerability of the fish to being captured in the second sample will most likely be compromised by being shocked and handled in the first sample! This issue is almost inescapable for fish sampling if fish are to be captured, removed from the water, and handled before release, and thus, the binomial sampling assumption will rarely be met.
One solution to this problem is to employ two sampling methods, simultaneously. The sampling gears we evaluated in our paper were standard baited fish traps (invasive) in combination with camera traps (non-invasive). Camera traps have become increasingly popular as a non-invasive sampling method for terrestrial wildlife and have been used to some degree for fish, but the combination of the trap and the camera can be a powerful fisheries tool because it allows researchers to collect specimens with the trap (for determining size/age compositions, etc., as is often important in fisheries research) and obtain an independent replicate count with the camera. Because the samples are collected simultaneously, any behavioral response by the fish to being trapped is a non-issue for the camera sample, eliminating one major source of noise that can lead to extra-binomial variation in the captures.
The development of useful tools that can reduce bias in monitoring information for fish, such as those presented here, are important advances for fisheries research as fisheries data are often plagued with noise that is difficult to explain, statistically. If you would like to discuss this method or others, please feel free to contact me at email@example.com.