Search Results
You are looking at 1 - 10 of 59 items for
- Author or Editor: Emmanouil N. Anagnostou x
- Refine by Access: All Content x
Abstract
An algorithm for real-time precipitation estimation that combines satellite infrared with long-range lightning network observations is developed. The emphasis is on enhancing current capabilities in continuous rainfall monitoring over large regions at high spatiotemporal resolutions and in separating precipitation type into its convective and stratiform components. Lightning information is retrieved from an experimental long-range very low frequency radio receiver network named the Sferics Timing and Ranging Network. Parameterizations for delineating the total rain area and its convective portion as well as convective and stratiform rain-rate relationships are obtained for lightning (LTG) and lightning-free (NLTG) clouds. The procedure accounts for differences in land versus ocean and for various levels of cloud system maturity. The parameters are evaluated using as reference the most definitive precipitation fields and rain classification estimates derived from the Tropical Rainfall Measuring Mission (TRMM) precipitation radar (PR). The algorithm is evaluated based on independent PR estimates and measurements from a rain gauge network in Florida. Overall, the algorithm underestimates rain area with respect to PR for LTG and NLTG clouds by about 20%, while for the rain volume there is an overestimation of ∼19% for LTG and ∼12% for NLTG clouds. Comparison of hourly estimates with rain gauges revealed an overall overestimation of 6% at 0.1° scale. At monthly scales, the biases are 2.4% and 0.27% for 1° and 2° resolutions. The significance of lightning information on rainfall estimation accuracy is investigated by applying the proposed technique without lightning information. The hypothesis made is that lightning measurement that is associated with ice aloft can provide better identification of the convective area, which could contribute to improving precipitation estimation. Indeed, comparisons with the PR showed that in rain area determination there is an overall bias reduction of 31% by using lightning information. In rain gauge comparisons, the bias reduction from incorporating lightning data is 87% for the hourly 0.1° estimates. In regards to correlation, the increase in hourly estimates varies from 0.13 to 0.03 for scales ranging from 0.1° to 1°.
Abstract
An algorithm for real-time precipitation estimation that combines satellite infrared with long-range lightning network observations is developed. The emphasis is on enhancing current capabilities in continuous rainfall monitoring over large regions at high spatiotemporal resolutions and in separating precipitation type into its convective and stratiform components. Lightning information is retrieved from an experimental long-range very low frequency radio receiver network named the Sferics Timing and Ranging Network. Parameterizations for delineating the total rain area and its convective portion as well as convective and stratiform rain-rate relationships are obtained for lightning (LTG) and lightning-free (NLTG) clouds. The procedure accounts for differences in land versus ocean and for various levels of cloud system maturity. The parameters are evaluated using as reference the most definitive precipitation fields and rain classification estimates derived from the Tropical Rainfall Measuring Mission (TRMM) precipitation radar (PR). The algorithm is evaluated based on independent PR estimates and measurements from a rain gauge network in Florida. Overall, the algorithm underestimates rain area with respect to PR for LTG and NLTG clouds by about 20%, while for the rain volume there is an overestimation of ∼19% for LTG and ∼12% for NLTG clouds. Comparison of hourly estimates with rain gauges revealed an overall overestimation of 6% at 0.1° scale. At monthly scales, the biases are 2.4% and 0.27% for 1° and 2° resolutions. The significance of lightning information on rainfall estimation accuracy is investigated by applying the proposed technique without lightning information. The hypothesis made is that lightning measurement that is associated with ice aloft can provide better identification of the convective area, which could contribute to improving precipitation estimation. Indeed, comparisons with the PR showed that in rain area determination there is an overall bias reduction of 31% by using lightning information. In rain gauge comparisons, the bias reduction from incorporating lightning data is 87% for the hourly 0.1° estimates. In regards to correlation, the increase in hourly estimates varies from 0.13 to 0.03 for scales ranging from 0.1° to 1°.
Abstract
Seasonal differences in the calibration of overland passive microwave rain retrieval are investigated using Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) and precipitation radar (PR). Four geographic regions from southern Africa, South Asia, the Amazon basin, and the southeastern United States are selected. Three seasons are compared for each region. Two scenarios of algorithm calibration are considered. In the first, the parameter sets are derived by calibrating the TMI algorithm with PR in each season. In the second scenario, common parameter sets are derived from the combined dataset of all three seasons. The parameter sets from both scenarios are then applied to the validation dataset of each season to determine the effect of seasonal calibration. Furthermore, calibration parameters from one season are also applied to another season, and results are compared against those derived using the season’s own parameters. Appreciable seasonal differences are observed for the U.S. region, while there are no significant differences between using individual seasonal calibration and the all-season calibration for the other regions. However, using one season’s parameter set to retrieve rainfall for another season is associated with increased uncertainty. It is also shown that the performance of the retrieval varies by season.
Abstract
Seasonal differences in the calibration of overland passive microwave rain retrieval are investigated using Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) and precipitation radar (PR). Four geographic regions from southern Africa, South Asia, the Amazon basin, and the southeastern United States are selected. Three seasons are compared for each region. Two scenarios of algorithm calibration are considered. In the first, the parameter sets are derived by calibrating the TMI algorithm with PR in each season. In the second scenario, common parameter sets are derived from the combined dataset of all three seasons. The parameter sets from both scenarios are then applied to the validation dataset of each season to determine the effect of seasonal calibration. Furthermore, calibration parameters from one season are also applied to another season, and results are compared against those derived using the season’s own parameters. Appreciable seasonal differences are observed for the U.S. region, while there are no significant differences between using individual seasonal calibration and the all-season calibration for the other regions. However, using one season’s parameter set to retrieve rainfall for another season is associated with increased uncertainty. It is also shown that the performance of the retrieval varies by season.
Abstract
A multicomponent radar-based algorithm for real-time precipitation estimation is developed. The algorithm emphasizes the combined use of weather radar observations and in situ rain gauge rainfall measurements. The temporal and spatial scales of interest are hourly to storm-total accumulations for areas of 4 km2 to approximately 16 km2. The processing steps include beam–height-effect correction, vertical integration, convective–stratiform classification, conversion from radar observables to rainfall rate, range-effect correction, and transformation of the estimated rainfall rates from polar coordinates to a Cartesian grid. Additionally, the algorithm applies advection correction to the gridded rainfall rates to minimize the temporal sampling effect and, subsequently, aggregates the corrected rainfall rates to 1-hourly, 3-hourly, and storm-total accumulations. The system applies different parameter values for convective and stratiform regimes. The calibration of the system is formulated as a global optimization problem, which is solved using the Gauss–Newton adaptive stochastic method. The algorithm is cast in a recursive formulation with parameters adjusted in real time. Evaluation of the system is based on an extensive dataset from the Melbourne, Florida, WSR-88D radar site.
Abstract
A multicomponent radar-based algorithm for real-time precipitation estimation is developed. The algorithm emphasizes the combined use of weather radar observations and in situ rain gauge rainfall measurements. The temporal and spatial scales of interest are hourly to storm-total accumulations for areas of 4 km2 to approximately 16 km2. The processing steps include beam–height-effect correction, vertical integration, convective–stratiform classification, conversion from radar observables to rainfall rate, range-effect correction, and transformation of the estimated rainfall rates from polar coordinates to a Cartesian grid. Additionally, the algorithm applies advection correction to the gridded rainfall rates to minimize the temporal sampling effect and, subsequently, aggregates the corrected rainfall rates to 1-hourly, 3-hourly, and storm-total accumulations. The system applies different parameter values for convective and stratiform regimes. The calibration of the system is formulated as a global optimization problem, which is solved using the Gauss–Newton adaptive stochastic method. The algorithm is cast in a recursive formulation with parameters adjusted in real time. Evaluation of the system is based on an extensive dataset from the Melbourne, Florida, WSR-88D radar site.
Abstract
The performance of a real-time radar rainfall estimation algorithm is examined based on an extensive dataset of volume scan reflectivity and rain gauge rainfall measurements from the WSR-88D site in Melbourne, Florida. Radar rainfall estimates are evaluated based on the following radar–rain gauge statistics: mean difference (bias), normalized root-mean-square difference, and correlation coefficient. The spatiotemporal scales of interest are hourly accumulations over 4 km × 4 km grids. First, the authors demonstrate the convergence properties of the algorithm’s adaptive parameter estimation procedure and conduct sensitivity tests of the system with respect to changes in the parameter values. Second, the major components of the algorithm are compared with the operational WSR-88D Precipitation Processing Subsystem. The authors show reduction in the radar–rain gauge root-mean-square difference up to 40%, resulting from the new parameterization schemes and the real-time calibration procedure. When rainfall classification is included, the reduction is higher (up to 50%). The authors show that correction for rain field advection moderately improves estimation accuracy (up to 20%). Finally, the authors show that the algorithm can effectively remove range-dependent systematic errors in radar observations.
Abstract
The performance of a real-time radar rainfall estimation algorithm is examined based on an extensive dataset of volume scan reflectivity and rain gauge rainfall measurements from the WSR-88D site in Melbourne, Florida. Radar rainfall estimates are evaluated based on the following radar–rain gauge statistics: mean difference (bias), normalized root-mean-square difference, and correlation coefficient. The spatiotemporal scales of interest are hourly accumulations over 4 km × 4 km grids. First, the authors demonstrate the convergence properties of the algorithm’s adaptive parameter estimation procedure and conduct sensitivity tests of the system with respect to changes in the parameter values. Second, the major components of the algorithm are compared with the operational WSR-88D Precipitation Processing Subsystem. The authors show reduction in the radar–rain gauge root-mean-square difference up to 40%, resulting from the new parameterization schemes and the real-time calibration procedure. When rainfall classification is included, the reduction is higher (up to 50%). The authors show that correction for rain field advection moderately improves estimation accuracy (up to 20%). Finally, the authors show that the algorithm can effectively remove range-dependent systematic errors in radar observations.
Abstract
The study uses storm tracking information to evaluate error statistics of satellite rain estimation at different maturity stages of storm life cycles. Two satellite rain retrieval products are used for this purpose: (i) NASA’s Multisatellite Precipitation Analysis–Real Time product available at 25-km/hourly resolution (3B41-RT) and (ii) the University of California (Irvine) Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) product available at 4-km–hourly resolution. Both algorithms use geostationary satellite infrared (IR) observations calibrated to an array of passive microwave (PM) earth-orbiting satellite sensor rain retrievals. The techniques differ in terms of algorithmic structure and in the way they use the PM rainfall to calibrate the IR rain algorithms. The satellite retrievals are evaluated against rain gauge–calibrated radar rainfall estimates over the continental United States. Error statistics of hourly rain volumes are determined separately for thunderstorm and shower-type convective systems and for different storm life durations and stages of maturity. The authors show distinct differences between the two satellite retrieval error characteristics. The most notable difference is the strong storm life cycle dependence of 3B41-RT relative to the nearly independent PERSIANN behavior. Another is in the algorithm performance between thunderstorms and showers; 3B41-RT exhibits significant bias increase at longer storm life durations. PERSIANN exhibits consistently improved correlations relative to the 3B41-RT for all storm life durations and maturity stages. The findings of this study support the hypothesis that incorporating cloud type information into the retrieval (done by the PERSIANN algorithm) can help improve the satellite retrieval accuracy.
Abstract
The study uses storm tracking information to evaluate error statistics of satellite rain estimation at different maturity stages of storm life cycles. Two satellite rain retrieval products are used for this purpose: (i) NASA’s Multisatellite Precipitation Analysis–Real Time product available at 25-km/hourly resolution (3B41-RT) and (ii) the University of California (Irvine) Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) product available at 4-km–hourly resolution. Both algorithms use geostationary satellite infrared (IR) observations calibrated to an array of passive microwave (PM) earth-orbiting satellite sensor rain retrievals. The techniques differ in terms of algorithmic structure and in the way they use the PM rainfall to calibrate the IR rain algorithms. The satellite retrievals are evaluated against rain gauge–calibrated radar rainfall estimates over the continental United States. Error statistics of hourly rain volumes are determined separately for thunderstorm and shower-type convective systems and for different storm life durations and stages of maturity. The authors show distinct differences between the two satellite retrieval error characteristics. The most notable difference is the strong storm life cycle dependence of 3B41-RT relative to the nearly independent PERSIANN behavior. Another is in the algorithm performance between thunderstorms and showers; 3B41-RT exhibits significant bias increase at longer storm life durations. PERSIANN exhibits consistently improved correlations relative to the 3B41-RT for all storm life durations and maturity stages. The findings of this study support the hypothesis that incorporating cloud type information into the retrieval (done by the PERSIANN algorithm) can help improve the satellite retrieval accuracy.
Abstract
The Tropical Rainfall Measuring Mission (TRMM) satellite carries a combination of active [precipitation radar (PR)] and multichannel passive microwave [the TRMM Microwave Imager (TMI)] sensors, which advance our ability to estimate rainfall over land. Rain retrieval from the TRMM PR is associated with an unprecedented accuracy and resolution but is limited in terms of sampling because of the narrow PR swath width (215 km). TMI provides wider coverage (760 km), but its observations are associated with a more complex relationship to precipitation in comparison with PR (especially over land). The PR rain estimates are used here for calibrating an overland TMI rain algorithm. The algorithm consists of 1) multichannel-based rain screening and convective/stratiform (C/S) classification schemes, and 2) nonlinear (linear) regressions for the rain-rate retrieval of stratiform (convective) rain regimes. This study examines regional differences in the algorithm performance. Four geographic regions consisting of central Africa (AFC), the Amazon (AMZ), the U.S. southern Plains (USA), and the Ganges–Brahmaputra–Meghna River basin (GBM) in south Asia are selected. Data from three summer months of 2000 and 2001 are used for calibration; validation is done using summer 2002 data. The current algorithm is also compared with the latest [version 6 (V6)] TRMM 2A12 product in terms of rain detection, and rain-rate retrieval error statistics on the basis of PR reference rainfall. The performance of the algorithm is different for the different regions. For instance, the reduction in random error (relative to 2A12 V6) is about 24%, 36%, 57%, and 165% for USA, AFC, AMZ, and GBM, respectively. However, significant difference between global (the four regions combined) and regional calibration is observed only for the GBM region.
Abstract
The Tropical Rainfall Measuring Mission (TRMM) satellite carries a combination of active [precipitation radar (PR)] and multichannel passive microwave [the TRMM Microwave Imager (TMI)] sensors, which advance our ability to estimate rainfall over land. Rain retrieval from the TRMM PR is associated with an unprecedented accuracy and resolution but is limited in terms of sampling because of the narrow PR swath width (215 km). TMI provides wider coverage (760 km), but its observations are associated with a more complex relationship to precipitation in comparison with PR (especially over land). The PR rain estimates are used here for calibrating an overland TMI rain algorithm. The algorithm consists of 1) multichannel-based rain screening and convective/stratiform (C/S) classification schemes, and 2) nonlinear (linear) regressions for the rain-rate retrieval of stratiform (convective) rain regimes. This study examines regional differences in the algorithm performance. Four geographic regions consisting of central Africa (AFC), the Amazon (AMZ), the U.S. southern Plains (USA), and the Ganges–Brahmaputra–Meghna River basin (GBM) in south Asia are selected. Data from three summer months of 2000 and 2001 are used for calibration; validation is done using summer 2002 data. The current algorithm is also compared with the latest [version 6 (V6)] TRMM 2A12 product in terms of rain detection, and rain-rate retrieval error statistics on the basis of PR reference rainfall. The performance of the algorithm is different for the different regions. For instance, the reduction in random error (relative to 2A12 V6) is about 24%, 36%, 57%, and 165% for USA, AFC, AMZ, and GBM, respectively. However, significant difference between global (the four regions combined) and regional calibration is observed only for the GBM region.
Abstract
This paper extends the work of Dinku and Anagnostou overland rain retrieval algorithm for use with Special Sensor Microwave Imager (SSM/I) observations. In Dinku and Anagnostou, Tropical Rainfall Measuring Mission (TRMM) precipitation radar (PR) rainfall estimates were used to calibrate TRMM Microwave Imager (TMI) retrieval. Regional differences in PR-based TMI calibration were investigated by testing the algorithm over four geographic regions, consisting of Africa, northern South America (containing the Amazon basin), the continental United States, and south Asia. In this paper the performance of Dinku and Anagnostou's technique applied on SSM/I data over three of these regions (Africa, Amazon, and South Asia) is demonstrated. Two approaches are investigated for using PR rainfall products to calibrate the algorithm parameters. In the first approach, TMI channels are remapped to the spatial resolutions of the corresponding SSM/I channels; then, PR is used to calibrate the rain retrieval on the remapped TMI data. In the second approach, the PR-based TMI algorithm calibration is performed at a coarser (0.25°) resolution. To assess the quality of algorithm estimates with respect to PR, rainfall fields derived from Dinku and Anagnostou, applied to SSM/I observations (using parameters determined from both approaches), are compared with matched (within ±15 min of the satellites' overpass time difference) PR surface rain rates. Calibration data come from the wet seasons (January–March) of 2000 and 2001. To assess the quality of the estimates with respect to PR, data from a 5-month period (December–April) of 2002, 2003, and 2004 are used. In comparison with the latest version of the Goddard profiling (GPROF) algorithm rain estimates, the current algorithm shows significant improvements in terms of both bias and random error reduction. The paper also shows that rain estimation based on TMI observations is associated with lower error statistics in comparison with the corresponding SSM/I retrievals.
Abstract
This paper extends the work of Dinku and Anagnostou overland rain retrieval algorithm for use with Special Sensor Microwave Imager (SSM/I) observations. In Dinku and Anagnostou, Tropical Rainfall Measuring Mission (TRMM) precipitation radar (PR) rainfall estimates were used to calibrate TRMM Microwave Imager (TMI) retrieval. Regional differences in PR-based TMI calibration were investigated by testing the algorithm over four geographic regions, consisting of Africa, northern South America (containing the Amazon basin), the continental United States, and south Asia. In this paper the performance of Dinku and Anagnostou's technique applied on SSM/I data over three of these regions (Africa, Amazon, and South Asia) is demonstrated. Two approaches are investigated for using PR rainfall products to calibrate the algorithm parameters. In the first approach, TMI channels are remapped to the spatial resolutions of the corresponding SSM/I channels; then, PR is used to calibrate the rain retrieval on the remapped TMI data. In the second approach, the PR-based TMI algorithm calibration is performed at a coarser (0.25°) resolution. To assess the quality of algorithm estimates with respect to PR, rainfall fields derived from Dinku and Anagnostou, applied to SSM/I observations (using parameters determined from both approaches), are compared with matched (within ±15 min of the satellites' overpass time difference) PR surface rain rates. Calibration data come from the wet seasons (January–March) of 2000 and 2001. To assess the quality of the estimates with respect to PR, data from a 5-month period (December–April) of 2002, 2003, and 2004 are used. In comparison with the latest version of the Goddard profiling (GPROF) algorithm rain estimates, the current algorithm shows significant improvements in terms of both bias and random error reduction. The paper also shows that rain estimation based on TMI observations is associated with lower error statistics in comparison with the corresponding SSM/I retrievals.
Abstract
A physically based methodology to incorporate passive microwave observations in a “rain-profiling algorithm” is developed for space- or airborne radars at frequencies exhibiting attenuation. The rain-profiling algorithm deploys a formulation for reflectivity attenuation correction that is mathematically equivalent to that of Hitschfeld and Bordan. In this formulation, the reflectivity–hydrometeor content (or rainfall rate) and reflectivity–attenuation relationships are expressed as a function of one variable in the drop size distribution parameterization, namely, the multiplicative factor in a normalized gamma distribution. The multiplicative factor parameter, mean cloud water content, and one parameter describing the precipitation phase are estimated in a Bayesian framework. This involves the minimization of differences between the 10-, 19-, 37-, and 85-GHz brightness temperature values predicted by a plane-parallel multilayer radiative transfer model and those observed by space- or airborne radiometers. A variational approach is devised to perform the minimization. The methodology is first tested using data simulated using a cloud model and is subsequently applied to coincident airborne brightness temperature and radar profile observations originating in the Kwajalein Experiment of the Tropical Rainfall Measuring Mission (TRMM). Results suggest improvements in rain estimation induced by the inclusion of the brightness temperature information in the retrieval framework if consistent modeling and quantification of errors are performed. Recommendations regarding the application of the method to TRMM satellite observations are formulated based on the findings of the study.
Abstract
A physically based methodology to incorporate passive microwave observations in a “rain-profiling algorithm” is developed for space- or airborne radars at frequencies exhibiting attenuation. The rain-profiling algorithm deploys a formulation for reflectivity attenuation correction that is mathematically equivalent to that of Hitschfeld and Bordan. In this formulation, the reflectivity–hydrometeor content (or rainfall rate) and reflectivity–attenuation relationships are expressed as a function of one variable in the drop size distribution parameterization, namely, the multiplicative factor in a normalized gamma distribution. The multiplicative factor parameter, mean cloud water content, and one parameter describing the precipitation phase are estimated in a Bayesian framework. This involves the minimization of differences between the 10-, 19-, 37-, and 85-GHz brightness temperature values predicted by a plane-parallel multilayer radiative transfer model and those observed by space- or airborne radiometers. A variational approach is devised to perform the minimization. The methodology is first tested using data simulated using a cloud model and is subsequently applied to coincident airborne brightness temperature and radar profile observations originating in the Kwajalein Experiment of the Tropical Rainfall Measuring Mission (TRMM). Results suggest improvements in rain estimation induced by the inclusion of the brightness temperature information in the retrieval framework if consistent modeling and quantification of errors are performed. Recommendations regarding the application of the method to TRMM satellite observations are formulated based on the findings of the study.
Abstract
Procedures for passive microwave precipitation estimation over land are investigated based on a large database of Tropical Rainfall Measuring Mission (TRMM) observations. The procedures include components for rain area delineation, convective/stratiform (C/S) rain classification, and estimation of vertically integrated water content or surface rainfall rate. The investigated algorithms include neural network schemes for both the rain area and C/S classification and statistical algorithms for precipitation estimation. The coincident active and passive microwave observations from TRMM, with the active (TRMM precipitation radar) observations providing the reference values for the various precipitation parameters, are used for algorithm calibration and validation. The calibration and validation are based on 1 yr of data over the continental United States and a repetitive sampling strategy that make the results statistically significant. Good agreement is demonstrated with TRMM precipitation radar observations in rain delineation, and it is shown that C/S classification can considerably improve precipitation estimation. It is also shown that better performance may be achieved in estimating vertically integrated hydrometeor contents as compared with rainfall rates.
Abstract
Procedures for passive microwave precipitation estimation over land are investigated based on a large database of Tropical Rainfall Measuring Mission (TRMM) observations. The procedures include components for rain area delineation, convective/stratiform (C/S) rain classification, and estimation of vertically integrated water content or surface rainfall rate. The investigated algorithms include neural network schemes for both the rain area and C/S classification and statistical algorithms for precipitation estimation. The coincident active and passive microwave observations from TRMM, with the active (TRMM precipitation radar) observations providing the reference values for the various precipitation parameters, are used for algorithm calibration and validation. The calibration and validation are based on 1 yr of data over the continental United States and a repetitive sampling strategy that make the results statistically significant. Good agreement is demonstrated with TRMM precipitation radar observations in rain delineation, and it is shown that C/S classification can considerably improve precipitation estimation. It is also shown that better performance may be achieved in estimating vertically integrated hydrometeor contents as compared with rainfall rates.
Abstract
This paper presents development of a statistical procedure for estimation of ensemble rainfall fields from a combination of ground radar observations and in situ rain gauge measurements. The uncertainty framework characterizes radar-rainfall estimation algorithm limitation accounting for rain gauge sampling uncertainty. The procedure is applied on a multicomponent rainfall estimation algorithm, which utilizes a rain-path attenuation correction technique, a power-law reflectivity-to-rainfall (Z–R) relationship, and a parameter to differentiate between convective (C) and stratiform (S) regimes in the Z–R conversion. Uncertainty is explicitly accounted for by evaluating the algorithm’s parameter set posterior probability density function (known as parameters’ equifinality) on the basis of the Generalized Likelihood Uncertainty Estimation (GLUE) framework. The study is facilitated by NASA’s C-band Doppler radar [named the Tropical Ocean Global Atmosphere (TOGA)] observations and four dense rain gauge clusters available from the Tropical Rainfall Measuring Mission (TRMM)-Large-Scale Biosphere–Atmosphere (LBA) experiment, conducted between January and February of 1999 in Southwest Amazon. Statistics are proposed for jointly evaluating the wideness of radar retrieval uncertainty limits [uncertainty ratio (UR)] and the percentage of observations that fall within those error bounds [exceedance ratio (ER)]. Results show that the parameter range selected in GLUE could characterize the radar-rainfall estimation uncertainty. Combined assessment of UR and ER for a varying range of parameters’ equifinality provides an objective basis for comparing rain retrieval algorithms and determining uncertainty bounds. Ensemble radar-rainfall fields derived on the basis of this procedure can be used to statistically assess satellite rain retrieval algorithms and derive ensemble hydrologic predictions driven by radar-rainfall input (e.g., runoff and soil moisture).
Abstract
This paper presents development of a statistical procedure for estimation of ensemble rainfall fields from a combination of ground radar observations and in situ rain gauge measurements. The uncertainty framework characterizes radar-rainfall estimation algorithm limitation accounting for rain gauge sampling uncertainty. The procedure is applied on a multicomponent rainfall estimation algorithm, which utilizes a rain-path attenuation correction technique, a power-law reflectivity-to-rainfall (Z–R) relationship, and a parameter to differentiate between convective (C) and stratiform (S) regimes in the Z–R conversion. Uncertainty is explicitly accounted for by evaluating the algorithm’s parameter set posterior probability density function (known as parameters’ equifinality) on the basis of the Generalized Likelihood Uncertainty Estimation (GLUE) framework. The study is facilitated by NASA’s C-band Doppler radar [named the Tropical Ocean Global Atmosphere (TOGA)] observations and four dense rain gauge clusters available from the Tropical Rainfall Measuring Mission (TRMM)-Large-Scale Biosphere–Atmosphere (LBA) experiment, conducted between January and February of 1999 in Southwest Amazon. Statistics are proposed for jointly evaluating the wideness of radar retrieval uncertainty limits [uncertainty ratio (UR)] and the percentage of observations that fall within those error bounds [exceedance ratio (ER)]. Results show that the parameter range selected in GLUE could characterize the radar-rainfall estimation uncertainty. Combined assessment of UR and ER for a varying range of parameters’ equifinality provides an objective basis for comparing rain retrieval algorithms and determining uncertainty bounds. Ensemble radar-rainfall fields derived on the basis of this procedure can be used to statistically assess satellite rain retrieval algorithms and derive ensemble hydrologic predictions driven by radar-rainfall input (e.g., runoff and soil moisture).
