Skip to content

Latest commit

 

History

History
43 lines (26 loc) · 2.89 KB

README.md

File metadata and controls

43 lines (26 loc) · 2.89 KB
Raw

radargaugemerging

Python package for statistical merging of radar and gauge precipitation measurements.

Package contents

The following scripts/modules have been implemented and tested:

File Description
collect_radar_gauge_pairs.py script for collecting co-located radar-gauge pairs
importers.py reading radar composites
iterate_kalman_mfb.py iterative running of Kalman filter-based mean field bias (MFB) estimation
kalman_mfb.py Kalman filter-based model for mean field bias
radar_archive.py browsing of radar archives
regression.py multivariate polynomial regression

Examples

Run Kalman filter-based MFB

The following example shows how to iteratively apply the mean field bias (MFB) estimator of Chumchean et al. implemented in kalman_mfb.py. Here we assume that the gauge data source is configured to be hourly accumulation and the name of the configuration profile is config. To collect gauge-radar pairs between 2017-08-12 15:00-16:00 UTC, we first run

python collect_gauge_radar_pairs.py 201708121500 201708121600 gaugeradarpairs_201708121600.dat config

This will compute hourly radar rainfall accumulation between the time period and pick the corresponding hourly gauge accumulations ending at 16:00. Using the gauge-radar pair file gaugeradarpairs_201708121600.dat created above, we can initialize the MFB estimator by running

python iterate_kalman_mfb.py 201708121600 gaugeradarpairs_201708121600.dat mfb_state.dat config

This will store the state of the MFB estimator to mfb_state.dat. Then we can collect the gauge-radar pair file for the next hour by running

python collect_gauge_radar_pairs.py 201708121600 201708121700 gaugeradarpairs_201708121700.dat config

Using the previous MFB state, we can then run

python iterate_kalman_mfb.py 201708121700 gaugeradarpairs_201708121700.dat mfb_state.dat config --prevstatefile mfb_state.dat

This will update the MFB state by using the previously predicted MFB and the most recently observed MFB. This can be repeated to iteratively over successive time steps apply the Kalman filter to produce the MFB estimates.

The logarithmic MFB estimate

$$\displaystyle\beta_t=\frac{1}{n}\sum_{i=1}^n\log_{10}\left(\frac{G_{i,t}}{R_{i,t}}\right)$$

is stored in the dictionary contained in the above state file. The dictionary has the key "corr_factor", whose value can be multiplied with radar-measured rain rates/accumulations to obtain the corrected values.