The broken_window2.0: Code for simulations and case study analysis

Résumé fourni par la source

Here we present data and code used to build and verify the performace of the broken_window2.0 algorithm. The broken_window2.0 is an extension of the broken_window (https://doi.org/10.1016/j.ecoinf.2021.101336), and is used to synthesize temporal abundance data, understand how much data is needed for population trends to be reliable, and quantify how likely an observed trend is spurious given the temporal scale (or 'window') of the population data used. Here, we expand the broken_window algorithmic framework to incorporate non-linear models of population dynamics. Specifically, we include sinusoidal dynamics, exponential dynamics, and quadratic dynamics. We test this expanded algorthmic framework using both simulated data and case study data. We simulated data under both low and high noise scenarios and examine true fit recovery across multiple window lengths. Next, we used time-series abundance data for Thymelicus lineola from a ~30 year butterfly monitoring dataset and assess the performance of the three models. To run the broken_window2.0 on your own data, you will need a dataframe with sample years (or whatever time-scale fits your data) in one column, and abundance in the other.T o run the tool yourself, set your working directory to connect to the 'tool folder' then load the 'broken_window2.' function into the R environment by running 'source(broken_window2.R)' - now you can run the broken_window2.0! See the 'CaseStudy_BrokenWindow2_Wrapper' R script for an example. You can also run each population model individually by loading in the individual functions (i.e. 'source(BrokenWindowSinu.R)' to just load in the sinusoidal algorithm). In the 'ToolFolder' zip folder is the individual algorithms, the full 'broken_window2.0' function, and the case study data we used in our analysis. In the 'SimulationsStudy' zip folder are the R scripts we used to individually test the performance of the different population models. Finally, the 'CaseStudyAnalysis' zip folder contains the R scripts we used to assess the performance of individual models within the broken_window2.0 on real-world time-series abundance data.