Fig 1.
Locations of 50 lakes in western Washington, United States.
The basemap is freely available from the US Census Bureau.
Fig 2.
Map of cumulative VGI user-days at 50 lakes in western Washington for 2015-2019 on eBird (dark blue), iNaturalist (teal), Flickr (yellow), Twitter (orange), and Gaia (red).
Point size corresponds to the number of user-days and point locations are jittered to ease interpretation. The basemap is freely available from the US Census Bureau.
Fig 3.
Scatterplots (lower-left), density plots (diagonal), and Pearson’s correlation coefficients (upper-right) comparing VGI datasets.
Statistical significance is indicated by *(<0.1), **(<0.01), and ***(<0.001).
Fig 4.
Scatterplots of annual cumulative VGI user-days from eBird, iNaturalist, Flickr, Twitter, and Gaia GPS versus annual mean on-site visitation estimates.
Table 1.
The top ten candidate visitation models relating mobile platform use and on-site visitation estimates. K is the number of parameters in the model, Delta AICc is the difference between AICc of the best fitting model and that of the top model, Model Likelihood is the relative likelihood, AICc Weight is the Akaike weight, and the R2 and root mean square error (RMSE) values are from out-of-sample testing. All models with Delta AICc < 5 are listed.
Fig 5.
Observed and predicted in-sample annual mean daily visitation at lakes, predicted as a function of annual VGI user-days from eBird, iNaturalist, Flickr, Twitter, and Gaia.
Predicted values are plotted relative to observed empirical visitation (R2 = 0.89), and the slope line indicates a 1:1 relationship.
Fig 6.
Coefficient estimates from the revealed preference model relating lake visitation to lake attributes.
White circles represent means and bars are 95% confidence intervals of the estimates.