Dr Riley Hales, Louis R. Rosas
basininflow is Python package for making netcdf timeseries files of runoff volume accumulation values for each
subbasin in a given weight table. The output is formatted to work with the RAPID software.
The create_inflow_file function takes in six parameters:
1) a list of gridded runoff NetCDF datasets
2) the name of the output subfolder
3) path where the subfolders and outputs are located
4) a weight table
5) a comid_lat_lon_z file
6) is the input forecast data?
The create_inflow_file function is designed to be executed on a linux supercomputer.
The input NetCDF datasets should follow the following conventions. It is expected that the datasets have 3 or 4 dimensions, in the following orders: time, latitude, longitude or time, expver, latitude, longitude (The 4th dimension is often present in the most recent releases of the ECMWF datasets, which has an 'experimental version' dimension. Internally, this dimension is flattened and the data summed. The name of this dimension does not matter). The runoff, longitude, and latitude variables may be any of the following, respectively: 'ro', 'RO', 'runoff', 'RUNOFF', 'lon', 'longitude', 'LONGITUDE', 'LON', 'lat', 'latitude', 'LATITUDE', 'LAT'. Latitude and longitude is expected in even steps, from 90 to -90 degrees and 0 to 360 degrees repectively. Time is expected as 'time'. The difference in time between each dataset should be equal (typically a time step of a day), and should be an interger.
The user may input as many datasets as desired. There is a built in memory check which warns the user if the amount of memory required exceeds 80% of the available RAM and will terminate the process if the amount of memory required exceeds all the available memory.
If the input dataset is a forecast dataset, they typically contain timesteps of 3 hours which then switch to 6 hours after a certain amount of days. In addition, the forecast data is assumed to be cumulative. By setting the 'forecast' variable to True, the forecast data will be switched from cumulative to instaneous, and the time steps will be forced to 3 hours.
Example NetCDFs can be found in /tests/inputs/era5_721x1440_sample_data.
A single csv file containing at least the following 5 columns of data with the following column names:
1) 'area_sqm', the area in square meters of a certain basin that intersects with the gridded runoff data.
2) 'lon', longitude of the centroid of the gridded runoff data cell that intersects with a certain basin
3) 'lat', latitude of the centroid of the gridded runoff data cell that intersects with a certain basin
4) 'lon_index', x index of the gridded runoff data cell that intersects with a certain basin
5) 'lat_index', y index of the gridded runoff data cell that intersects with a certain basin
Weight tables are expected to contain some digits, followed by an 'x' and the some more digits (i.e. weight_123x4567.csv). These digits represent the lat-lon dimensions of the input NetCDFS. There is an internal check which verifies that the weight table and NetCDF dimensions are the same (currently order does not matter). This file can be generated using either these ArcGIS tools or these python scripts.
An example weight table can be found in /tests/inputs.
A csv which has the unique ids (COMIDs, reach ids, LINKNOs, etc.) of the basins/streams modeled in its first column, sorted in the order in which the user wants RAPID to process. Two other columns must be present, labeled 'lat' and 'lon'. These columns represent a point related to each river reach. This file can be generated using either these ArcGIS tools or these python scripts.
An example comid csv can be found in /tests/inputs.
The create_inflow_file outputs a new NetCDF 3 (classic) dataset in a folder called vpu, under then inflows_dir directory. It is desgined to be accepted by rapid and pass all of its internal tests. Example output datasets can be found in /tests/validation.