My first shiny app! Using the results from ensemble models of my scientific initiation to define the priority areas for conservation of Brazil nut
First of all, what made me get started and learn shiny was the facility provided by the app for visualization. This is quite useful when you’ve got an amount of data to visualize and compare. Of course any GIS software is enough to do this job! But it’s not easy to handle and share a set of maps keeping the interactivity from them so that’s why I was searching how to make it properly. And everyone that I know would rather enjoy beautiful interactive graphics or maps to analyse the data. I think this is the best way to deal with the results, it is always good to simplify the visualization before studying it further theoretically.
Well, the maps analysed were results from a prediction made by ensemble models to find out the priority areas for conservation of Bertholletia excelsa using different filters (environmental and geographical) based on 15 variables defined as the best in predicting the suitability of Brazil nut by Daiana tourne et. al 2019 on her doctorate degree using principal component analyses (PCA) combined with expert analyses.
Therefore the main objective is directly correlated with the climatic analyses and soon we will combine with genetic analyses. These results presented here are considering the actual conditions using the variables from different databases as Worldclim, USGS, etc. More details about the research will be posted in a formal article in the future when it gets ready. Is important to say that in this research I have got 3 co-authors from Bioversity International which are: Evert Thomas, Tobias Fremout and Viviana Ceccarelli. Besides them Daiana also is helping in this project as well as my advisor Karina Martins.
The shiny app has a good and organized structure quite easy to understand which helped me a lot to generate the app. However I didn’t use packages as Golem to provide a code cleaner so it was a robust script which I have written without using modules, because although the app has a clear structure and easy to apply modules I wasn’t familiar with it.
The app has 6 tabs with different filters and content as you can see below:
So as I said before the model is an ensemble of different algorithms (MAXENT, RF, GLMNET, etc). Each of these algorithms produces a different presence map. The suitability map shows us the probability (0-1000) to get the presence of the species occurring at the site assuming a specific threshold based on the model filter which could be Environmental or Geographical. The count map indicates how many algorithms predict suitable a point, from 0 algorithms to 8 algorithms which is the maximum number of algorithms which predict suitable the same grid from 10 algorithms used in the ensemble initially. And last but not least the binary map shows us absence and presence (0-1) also considering the same threshold for suitability map.
I will leave here the Shiny app if you would like to take a look further.