A Gaussian mixture method for specific differential phase retrieval at X-band frequency

Geyi, Wen; Fox, Neil I.; Market, Patrick S.

doi:10.5194/amt-12-5613-2019

Cited by 6 publications

(10 citation statements)

References 64 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Most MeteoSwiss operational products are estimated over a Cartesian grid of 1 km 2 (in the Swiss LV03 coordinate system) at a temporal resolution of 2 to 5 min. For numerical prediction, MeteoSwiss runs the COSMO model which is a mesoscale-limited area model that is operated and developed by several weather services in Europe (e.g., Switzerland, Italy, Germany, Poland, Romania and Russia) (Seifert et al, 2011;Doms et al, 2011;Baldauf et al, 2011;Wolfensberger and Berne, 2018). COSMO analysis runs are available over Switzerland every hour 1 on a 3D irregular grid.…”

Section: Collocated Gauge-radar Databasementioning

confidence: 99%

“…To create a database of 4 years of radar data, more than 10 million radar PPI scans have to be processed (five radars, 20 elevations every 5 min). Because of this, it is computationally impossible to use an advanced K dp estimation method, such as a Kalman filter method (Schneebeli et al, 2014) or a Gaussian-mixture regression (Wen et al, 2019). Hence a simple method is used in this study which is also used in Wolfensberger et al (2018) and is similar to the one in Timothy et al (1999).…”

Section: Appendix A: Computation Of K Dp and A Hmentioning

confidence: 99%

“…Because of this, it is computationally impossible to use an advanced K dp estimation method, such as a Kalman filter method (Schneebeli et al, 2014) or a Gaussian-mixture regression (Wen et al, 2019). Hence a simple method is used in this study which is also used in Wolfensberger et al (2018) and is similar to the one in Timothy et al (1999). The raw total differential phase shift dp is first corrected for the system offset and then filtered with a moving median filter to give an estimate of the total differential phase shift on propagation dp .…”

Section: Appendix A: Computation Of K Dp and A Hmentioning

confidence: 99%

See 2 more Smart Citations

RainForest: a random forest algorithm for quantitative precipitation estimation over Switzerland

et al. 2021

View full text Add to dashboard Cite

Abstract. Quantitative precipitation estimation (QPE) is a difficult task, particularly in complex topography, and requires the adjustment of empirical relations between radar observables and precipitation quantities, as well as methods to transform observations aloft to estimations at the ground level. In this work, we tackle this classical problem with a new twist, by training a random forest (RF) regression to learn a QPE model directly from a large database comprising 4 years of combined gauge and polarimetric radar observations. This algorithm is carefully fine-tuned by optimizing its hyperparameters and then compared with MeteoSwiss' current operational non-polarimetric QPE method. The evaluation shows that the RF algorithm is able to significantly reduce the error and the bias of the predicted precipitation intensities, especially for large and solid or mixed precipitation. In weak precipitation, however, and despite a posteriori bias correction, the RF method has a tendency to overestimate. The trained RF is then adapted to run in a quasi-operational setup providing 5 min QPE estimates on a Cartesian grid, using a simple temporal disaggregation scheme. A series of six case studies reveal that the RF method creates realistic precipitation fields, with no visible radar artifacts, that appear less smooth than the original non-polarimetric QPE and offers an improved performance for five out of six events.

show abstract

Section: Collocated Gauge-radar Databasementioning

confidence: 99%

Section: Appendix A: Computation Of K Dp and A Hmentioning

confidence: 99%

Section: Appendix A: Computation Of K Dp and A Hmentioning

confidence: 99%

See 1 more Smart Citation

RainForest: a random forest algorithm for quantitative precipitation estimation over Switzerland

et al. 2021

View full text Add to dashboard Cite

show abstract

“…To create a database of four years of radar data, more than 10 million radar PPI scans have to be processed (5 radars, 20 elevations every 5 minutes). Because of this, it is computationally impossible to use an advanced K dp estimation method, such as a Kalman filter method (Schneebeli et al, 2014) or a Gaussian-mixture regression (Wen et al, 2019). Hence a simple method is used in this study which is also used in and is similar to Timothy et al (1999).…”

Section: Appendix A: Computation Of K Dp and A Hmentioning

confidence: 99%

RainForest: A random forest algorithm for quantitative precipitation estimation over Swizerland

Wolfensberger¹,

Gabella²,

Boscacci³

et al. 2020

Preprint

View full text Add to dashboard Cite

Abstract. Quantitative precipitation estimation (QPE) is a difficult task, particularly in complex topography, and requires the adjustment of empirical relations between radar observables and precipitation quantities, as well as methods to transform observations aloft to estimations at the ground level. In this work, we tackle this classical problem with a new twist, by training a random forest (RF) regression to learn a QPE model directly from a large database comprising four years of combined gauge and polarimetric radar observations. This algorithm is carefully fine-tuned by optimizing its hyper-parameters and then compared with MeteoSwiss' current operational non-polarimetric QPE method. The evaluation shows that the RF algorithm is able to significantly reduce the error and the bias of the predicted precipitation intensities, especially for large and solid/mixed precipitation. In weak precipitation, however, and despite a-posteriori bias correction, the RF method has a tendency to overestimate. The trained RF is then adapted to run in a quasi-operational setup providing 5 minute QPE estimates on a Cartesian grid, using a simple temporal disaggregation scheme. A series of six case-studies reveal that the RF method creates realistic precipitation fields, with no visible radar artifacts, that appear less smooth then the original non-polarimetric QPE, and offers an improved performance for five out of six events.

show abstract

“…In [37], we proposed a Gaussian mixture method (GMM) for K dp estimation, which can not only retrieve mean K dp , but also obtain the statistical uncertainty of K dp , denoted as σ(K dp ). In this study, we apply K dp to the attenuation correction and the rain rate estimation for the X-band radar at the University of Missouri (MZZU) to provide radar hydrological applications.…”

Section: Introductionmentioning

confidence: 99%

The Quality Control and Rain Rate Estimation for the X-Band Dual-Polarization Radar: A Study of Propagation of Uncertainty

2020

Self Cite

View full text Add to dashboard Cite

In this study, the specific differential phase ( K d p ) is applied to attenuation correction for radar reflectivity Z H and differential reflectivity Z D R , and then the corrected Z H , Z D R , and K d p are studied in the rain rate (R) estimation at the X-band. The statistical uncertainties of Z H , Z D R , and R are propagated from the uncertainty of K d p , leading to variability in their error characteristics. For the attenuation correction, a differential phase shift ( Φ d p )-based method is adopted, while the statistical uncertainties σ ( Z H ) and σ ( Z D R ) are related to σ ( K d p ) via the relations of K d p -specific attenuation ( A H ) and K d p -specific differential attenuation ( A D P ), respectively. For the rain rate estimation, the rain rates are retrieved by the power-law relations of R ( K d p ) , R ( Z h ) , R ( Z h , Z d r ) , and R ( Z h , Z d r , K d p ) . The statistical uncertainty σ ( R ) is propagated from Z H , Z D R , and K d p via the Taylor series expansion of the power-law relations. A composite method is then proposed to reduce the σ ( R ) effect. When compared to the existing algorithms, the composite method yields the best performance in terms of root mean square error and Pearson correlation coefficient, but shows slightly worse normalized bias than R ( K d p ) and R ( Z h , Z d r , K d p ) . The attenuation correction and rain rate estimation are evaluated by analyzing a squall line event and a prolonged rain event. It is clear that Z H , Z D R , and K d p show the storm structure consistent with the theoretical model, while the statistical uncertainties σ ( Z H ) , σ ( Z D R ) and σ ( K d p ) are increased in the transition region. The scatterplots of Z H , Z D R , and K d p agree with the self-consistency relations at X-band, indicating a fairly good performance. The rain rate estimation algorithms are also evaluated by the time-series of the prolonged rain event, yielding strong correlations between the composite method and rain gauge data. In addition, the statistical uncertainty σ ( R ) is propagated from Z H , Z D R , and K d p , showing higher uncertainty when the large gradient presents.

show abstract

A Gaussian mixture method for specific differential phase retrieval at X-band frequency

Cited by 6 publications

References 64 publications

RainForest: a random forest algorithm for quantitative precipitation estimation over Switzerland

RainForest: a random forest algorithm for quantitative precipitation estimation over Switzerland

RainForest: A random forest algorithm for quantitative precipitation estimation over Swizerland

The Quality Control and Rain Rate Estimation for the X-Band Dual-Polarization Radar: A Study of Propagation of Uncertainty

Contact Info

Product

Resources

About