Uncertainty in biological monitoring: a framework fordata collection and analysis to account for multiplesources of sampling bias

Authors: V Ruiz-Gutierrez; M B Hooten; Evan HC Grant
Contribution Number: 532

eSummary(1) Biological monitoring programs are increasingly relying upon large volumes of citizenscience data to improve the scope and spatial coverage of information, ch a l l en g i n g thescientific commun i ty to develop design and model - b ase d approaches to improve inference.(2) Recent statistical models in ecology have be en develope d to accommodate false-negativeerrors, although recent work points to false positive errors as equally important sources ofbias. This is of particular concern for the success of any m o n i t or i n g program given rates assmall as 3% could lead to th e overestimation of the occurrence of rare events by as much as50%, and even small false positive r a t es can severely b i as estimates of occurrence dynamics.(3) We present an int eg ra t ed , comput at i o n al l y efficient Bayesian hierar chical model tocorrect for fal se positi ve and negative error s in detection/no n -d et ec ti o n data. Our modelcombines ind ependent, a u x i l i ar y data sources with field observations to i m p r ove t h eestimation of false positive rate s, when a subset of field observatio n s cannot be validated aposteriori or assumed as per fe ct . We evaluated the performance of the model across arange of occurren ce rates, false positive and negative errors, and quantity of auxil i ar y data.(4) Th e mode l perfor m ed well under all simulated scenario s, and we were able t o identifycritical auxiliary data characteristics which resul t ed in improved infer en ce. We applied ourfalse positive m odel to a large-scale, citizen -sci e n ce monitor i n g program for anurans in theNortheastern U.S., using auxiliary data from an experiment d esi g n ed to estimate falsepositive er r o r rates. Not correcting for false positive ra t es resulted in biased estimates ofoccupancy in 4 of the 10 anu r a n species we an a l y zed , leadin g to an overestima t i on of theaverage number of occupi ed survey routes by as much as 70%.Conclusions. The framework we present for da ta collecti o n and analysis is able toefficiently pr ovide reliable inference for occurrence patterns using data from acitizen-science monitorin g program . However, our approach is ap p l i ca b le to data generatedby any type of research and monit or i n g program , independent of skill level or scale, when effort i s placed on obtaining independent info rma t i on on false positive rates

Publication details
Published Date:
Outlet/Publisher: Methods in Ecology and Evolution doi/10.1111/2041-210X.12542
Media Format:

ARMI Organizational Units:
Northeast - Biology
Monitoring and Population Ecology; Quantitative Developments
Notice: PDF documents require Adobe Reader or Google Chrome Browser (recommended) for viewing.