topHRU - threshold optimization for HRUs in SWAT

Michael Strauch, Christoph Schürz, and Robert Schweppe

Contact: michael.strauch@ufz.de

About

topHRU calculates the average Relative Error of Aggregation (aREA) for different levels of input data aggregation within the Soil and Water Assessment Tool (SWAT). Input data aggregation in SWAT is based on HRU thresholds for land use, soil, and slope. topHRU allows identifying pareto-optimal threshold combinations (minimal spatial error for a given number of HRUs) to minimze the trade-off between computation time and model error.

Installing the topHRU package

To install the package please execute following lines in R:

install.packages("devtools")
install.packages(c("abind", "emoa", "dplyr", "ggplot2","plotly"))
devtools::install_github("michstrauch/TopHRU")

Minimum example

The package contains the template dataset hru_data. To see the structure of the required table you can compare with the template dataset:

library(topHRU)
View(hru_demo)

Input data

As INPUT you need the "hrus" table from the "project_name.mdb" from a SWAT project database. The hrus table will become available AFTER the definition of HRUs in an ArcSWAT project WITHOUT applying thresholds (i.e. 0 for land use, soil, and slope). You can export the hrus table either from a .csv file (therefore the hrus table must be extracted from the "project_name.mdb" and saved to "name.csv" prior to loading it in R. Otherwise you can load the table directly from the "project_name.mdb". Therefore the RODBC package must be installed and a 32 bit version of R must be used.

hru_table <- extract_hru("path/to/extracted/hrus.csv") 
#or
hru_table <- extract_hru("path/to/project_name.mdb") 

####Analysis To run an HRU analysis for the template dataset run:

hru_eval <- evaluate_hru(hru_demo)

The resulting list holds the complete results of the HRU analysis,

hru_eval$result_all

and a reduced result table only showing the pareto-optimal threshold combinations,

hru_eval$result_nondominated

####Visualization plot_pareto provides two options for visualization. The default is an interactive visualization of the dominated and nondominated threshold combinations, where the function returns a plotly object. Such plot supports the user in finding an adequate threshold combination for the respective project is available with the following function call:

plot_pareto(hru_analysis = hru_analysis, area_thrs = 0.1, hru_thrs = 2000)

For publication a ggplot object is more useful. Hence setting the parameter interactive = FALSE returns a static plot:

plot_pareto(hru_analysis = hru_analysis, area_thrs = 0.1, hru_thrs = 2000, interactive = FALSE)

The two threshold parameters in the function define the positions of the dashed guide lines that seperate the data points. If these are not needed, no threshold values are provided with the function call.