Free drawing distributions with shinysense

2019 Jun 07 @travishinkelman.com

One of the areas of expertise at Cramer Fish Sciences is watershed and habitat restoration. In the context of that work, we are often faced with estimating how much rearing habitat is needed to support a specified number of juvenile salmonids (or how many juvenile salmonids are supported by a specified amount of rearing habitat). Our typical approach for generating these estimates is to use the territory size-fork length relationship from Grant and Kramer (1990). To simplify that calculation, we often assume that all the fish are the same size.

However, I recently learned about an R package, shinysense, that allows the user to draw a line on a plot with their mouse in a Shiny app. Using this approach, I built a little shiny app (embedded below but better when viewed in separate browser window) [1] that allows the user to draw an empirical fork length distribution [2] and then calculates habitat need or habitat capacity based on the drawn fork length distribution.

It is remarkably straightforward to use shinysense to free draw a distribution. The basic idea is that a set of x-axis values and the limits of the y-axis are specified. When a line is drawn, the y-axis values are mapped to the specified x-axis values and available on the server side of the Shiny app. In the Juvenile Salmonid Habitat Calculator, the y-axis values are arbitrary and re-scaled to sum to one. A histogram and table of the distribution give the user feedback on their drawing.

The big picture idea for the Juvenile Salmonid Habitat Calculator has been bouncing around in the back of my mind for several months, but I wasn't sure how to implement the fork length distribution part of the app (based on my existing Shiny knowledge). When I found shinysense, I wasn't searching for it, but I was thrilled to realize that it would allow me to easily implement my idea for the Juvenile Salmonid Habitat Calculator.


[1] Shiny app is hosted on shinyapps.io and code is available through GitHub.

[2] Here is an example of fitting a theoretical distribution to a drawing.