This repository contains the software, processed data, and all behavioral simulations associated with the paper "The long distance flight behavior of Drosophila supports an agent-based model for wind-assisted dispersal in insects." For a pre-print, see: https://www.biorxiv.org/content/10.1101/2020.06.10.145169v1

The original data (~ 0.5 TB) will be made available upon reasonable request upon formal publication.

Processed data is available through this repository, along with instructions for use, under the folder "./processed_field_data"

This readme assumes working knowledge of Ubuntu and python. This code is not actively maintained. It was last updated on 21 April 2021.

Code and data are licensed under a Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) License.

Data in this repository are stored in a variety of data formats, all offering reasonable human-readability, and all readable with open-source tools.

• most processed data and experimental metadata are saved as .json files
• some field data, such as anemometer recordings, are saved as .csv files

The folder "./agent_based_simulations_models_I_through_IV" houses:

• scripts used to simulate the groundspeed-windspeed relationships expected from fixed-azimuth models (I through IV):
  • simulations with free parameters set to values estimated from the literature (as in Fig. 6), in "/simulations"
  • sensitivity analyses to individually optimize each model (as in Supp. Figs 2 and 3) on the basis of their fit to field data, in "/simulations_sensitivity_analysis"
  • finally, comparing these individually-optimized models on the basis of relative fit to field data (as in Supp. Fig 4), in "/bayes_comparing_optimized_simulations"
• scripts used to simulate groundspeed-windspeed relationships from model I, above, with a suite of stochastic variations introduced ("/model_I_with_stochastic_heading_changes")
• an outdated folder ("/models_V_and_VI_NOT_PRESENTED_IN_PNAS_PAPER") containing preliminary attempts at stochastic simulations that we removed from the paper during the PNAS review process (not removed from repository due to low competence with git).

All scripts used to run advection-diffusion simulations were written by Will Dickson, and are packaged in a separate GitHub repository at: https://github.com/willdickson/ad_coyote_lake

The folder "./advection_diffusion_simulations" houses:

• final datasets produced from the above advection-diffusion simulations, in "/advection_diffusion_datasets"
• scripts and intermediate figure sequences used to assess the above datasets with respect to their fit to field data (as in Fig. 7), in "/advection_diffusion_analyses"

If you would like to re-run our analyses starting from our raw camera-trap image stream, please contact the authors (leitchka@gmail.com). Camera data from each experiment comprise 30 to 120 GB of data; we will make available, upon reasonable request, these data from individual experiments or from the entire dataset (~0.5 TB).

Once you have the camera-trap images, you will need to follow the following instructions for making these data accessible to the analysis (below).

• Ubuntu (we used Ubuntu 16)
• Python (2.7)
• OpenCV (3.3.1)

We have provided a directory structure that will facilitate your analysis of any raw camera-trap data you've requested and received. This directory is called "./field_data_and_analysis_scripts." Subfolders with experimental date names (e.g. "2017_10_26") have been prepared for your manual deposition of the raw camera data we provide. This will require two steps:

1. First, copy the data into the folder: "./field_data_and_analysis_scripts/2017_10_26/trapcam_timelapse_TO_BE_MANUALLY_POPULATED"
2. Then, rename the folder into which you've pasted the data, so it simply reads "/trapcam_timelapse"

You'll want the resultant directory structure to look like:

	field_data_and_analysis_scripts		
		2017_10_26
			trapcam_timelapse
				trap_A					<----------------  
					mask.jpg					|
					tl_0007_0001_20161130_125959.jpg		|
					.						|
					.						|
					.       [many more image files]			|
					.					these are the folders/files
					.					provided on request
					tl_0007_3644_20161130_160239.jpg		|
				trap_B							|
					mask.jpg					|
					tl_0009_0001_20161130_103056.jpg		|
					.						|
					.						|
					.       [et cetera]		<----------------				

At this point, using "/field_data_and_analysis_scripts" as your working directory, you'll want to run the script "/run_trapcam_analysis.py," which will prompt your specification of:

• the experiments you'd like to analyze (e.g. 2017_10_26),
• the traps you'd like to analyze (e.g. trap_G), and
• whether you'd like to tinker around with analysis parameters, or if you'd like to use default analysis parameters to perform the final analysis.
  • if you choose the former, the analysis will come up with an "in-trap/on-trap threshold" for each trap analyzed, on the basis of any dips found in a multimodal histogram of detected-contour contrast metrics. This threshold will be saved in the parameter file, (e.g. "/2017_10_26/gaussian_analysis_parameters.json").
  • if you choose the latter, the analysis will use the saved "in-trap/on-trap threshold" to generate time-series estimations of the number of flies in, and on, the trap surface; these results will be saved in e.g. "/2017_10_26/all_traps_final_analysis_output.json"