BVRE-forecast-code

This code reproduces figures from the Beaverdam Reservoir forecasting data assimilation experiments using the FLARE (Forecasting Lake And Reservoir Ecosystems) system in the manuscript by Wander et al. titled "Data assimilation experiments inform monitoring needs for near-term ecological forecasts in a eutrophic reservoir." If you have any questions, contact Heather Wander at [email protected].

Instructions to reproduce manuscript + SI figures:

1. Download or clone github repository to your local computer
2. Run install.R in the workflows/DA_experiments folder to download GLM and FLARE packages and their dependencies
3. Run download_scores.R in the workflows/DA_experiments folder to download driver data and observation files from EDI or GitHub
4. Run BVR_FLARE_ms_figs.R script in the workflows/DA_experiments folder to reproduce manuscript and supplemental figures
5. Run BVR_FLARE_UC_figs.R script in the workflows/DA_experiments folder to reproduce Fig. 9 (proportion of IC uncertainty) and SI figures for forecasts run without initial conditions uncertainty

Instructions to reproduce FLARE forecasts and scores:

1. Run install.R in the workflows/DA_experiments folder to download GLM and FLARE packages and their dependencies

2. Run combined_workflows.R in the workflows/DA_experiments folder to iteratively generate forecasts for every data assimilation frequency and day in the forecast period

   Note Running forecasts for 365 days and all four data assimilation frequencies will take > 10 days.

Instructions for reproducing using Docker

1. Download and install Docker to your computer (https://www.docker.com)

2. At the command line, run docker run --rm -ti -e PASSWORD=yourpassword -p 8787:8787 rqthomas/wander_et_al:latest

3. Open a webbrowser and enter http://localhost:8787. You will see an Rstudio login screen. The user name is rstudio and the password is yourpassword

4. In the Rstudion session: File -> Open project -> select BVRE-forecast-code/BVRE-forecast-code.Rproj

5. Follow the instructions above for reproducing the figures or the forecasts (note: the R packages are already installed in the Docker container so install.R does not need to be run)