An update to the bamr R package (built by Mark Hagemann while at UMass) that uses more geomorphically-informed prior knowledge for discharge inversion.
From bamr: "The bamr package facilitates Bayesian AMHG + Manning discharge estimation using stream slope, width, and partial cross-section area. It includes functions to preprocess and visualize data, perform Bayesian inference using Hamiltonian Monte Carlo (via models pre-written in the Stan language), and analyze the results."
geoBAMr expands upon this project by defining prior river knowledge using river classification frameworks. Internal to geoBAMr, geomorphic river types are assigned to rivers using stream widths, which in turn determine which priors are fed into the BAM algorithm. geoBAMr uses the identical Bayesian model as used in bamr.
First, you need to have installed rstan from source on your local machine. To do that, follow the directions at this link verbatim. Otherwise, an error will be thrown during package installation. This only needs to be done the first time you wish to install geoBAMr.
Following that, you can install geoBAMr:
#First get devtools package
if (!require("devtools")) {
install.packages("devtools")
library("devtools")
}
#Then install from github
devtools:: install_github("craigbrinkerhoff/geoBAMr", force=TRUE)
Note that if you need to reinstall this package, uninstall the current version first. Also, check if R has created the folder '~R/win-library/3.6/00LOCK-geoBAMr'. If so, delete it and reinstall.
The best way to get started is to follow the examples in the included vignettes at the bamr website.
To read about the river classification frameworks available and how to implement them, go to the geoBAMr website and click on 'Getting Started'.
-
The Sacramento test case in the bamr package is not included with
geoBAMr. -
geoBAMris far more memory intensive thanbamr, and as such the precompiled stan model will crash if you are using medium-to-large matrices of river widths/slopes. This is likely the issue if the following error is thrown:
Error in unserialize(socklist[[n]]) : error reading from connection
Error in serialize(data, node$con, xdr = FALSE) :
error writing to connection
The workaround is to compile the stan model yourself! Workflow is identical except you 1) add a line of code and 2) amend the bam_estimate() function call. See below (after downloading the stan model from this repo 'master.stan'):
geoBAM_model <- stan_model("~//master.stan", model_name = 'geoBAM_model')
run_bam <- bam_estimate(bamdata=bamdata, bampriors = priors, stanmodel=geoBAM_model)
If you are dead set on running geoBAMr as is (generally for simplicity's sake), one solution is to find more memory! i.e. run geoBAMr on a computing cluster.
- If both
bamrandgeoBAMrare installed, make sure to explictly call functions by package as they have the same names. Otherwise, chaotic confusion will ensue!
For example:
geoBAMr:: bam_estimate()
bamr:: bam_estimate()
For any questions regarding this package, I am reachable at cbrinkerhoff@umass.edu