By Topic

Rain field and reflectivity vertical profile reconstruction from C-band Radar volumetric data

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

3 Author(s)
Marzano, F.S. ; Dept. of Electr. Eng., Univ. of L''Aquila, Italy ; Vulpiani, G. ; Picciotti, E.

Operating a meteorological radar is generally a challenging task when in presence of a significant beam blockage as in complex orography. Apart from enhanced ground clutter, mountainous obstructions of the radar beam can significantly reduce the radar visibility and, thus, its monitoring capabilities. Self-consistent adaptive techniques to reconstruct vertical profiles of reflectivity (VPR) and near-surface rain-rate fields from high-elevation reflectivity bins are here proposed, compared, and tested for ranges up to 60 km. The methodology is based on statistical estimators trained by a large reflectivity volumetric datasets, classified into stratiform and convective rain regimes and resampled onto a uniform Cartesian grid by means of a modified Cressman technique. For what concerns reflectivity vertical profiles, two methods, respectively named statistical nonlinear reconstruction (NSR) and neural network reconstruction (NNR), are considered. The NSR method is based on the principal component analysis, applied to the radar dataset, in order to extract significant reflectivity-profile variance. A retrieval technique, based on a nonlinear multiple regression scheme, is then used to infer near-surface reflectivity from available high-altitude echoes at a given range. The NNR is based on a three-layer artificial neural network trained by means a feedforward backpropagation algorithm. For what concerns the near-surface rain retrieval, besides a power-law reflectivity-rain-rate (ZR) approach, a three-layer neural network technique is also set up in order to estimate surface rain rate from reconstructed VPR. The proposed reconstruction techniques are here illustrated by using volumetric data acquired by the C-band Doppler single-polarization radar, operated in L'Aquila, Italy. A case study, related to a rainfall event that occurred during fall 2000, is discussed. Using a test area within 60 km from the radar site and simulating the presence of beam obstructions, a comparison of NSR and NNR with conventional area average reconstruction techniques shows that the percentage improvement of both NSR and NNR approaches is significant, both for the error bias (by 30% to more than 50%, depending on altitude) and variance (by 10% to more than 20%). A sensitivity test indic- ates that the VPR reconstruction procedure is fairly robust to missing data, especially in terms of error bias. The comparison of estimated radar rainfall with rain gauge data measurement is also illustrated. The mean field bias closer to its optimal value and an error variance much smaller is obtained when neural network techniques are applied than with conventional ZR methods for both techniques of reconstruction. With respect to the latter, the obtained improvement is more than 40% in terms of root mean square error and is comparable when estimating near-surface rain rate using either NSR or NNR methods to reconstruct the reflectivity vertical profiles. Limitations, potential, and future developments of the proposed adaptive reconstruction techniques are finally discussed.

Published in:

Geoscience and Remote Sensing, IEEE Transactions on  (Volume:42 ,  Issue: 5 )