See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/262159912

Validation and strategies to improve the Hydro-Estimator and NEXRAD over

Puerto Rico

Conference Paper · July 2008

CITATIONS READS

4 96

6 authors, including:

Nazario D. Ramirez-Beltran Robert Kuligowski

University of Puerto Rico at Mayagüez National Oceanic and Atmospheric Administration
43 PUBLICATIONS   166 CITATIONS    81 PUBLICATIONS   1,455 CITATIONS   

SEE PROFILE SEE PROFILE

Eric Harmsen Joan Manuel Castro Sanchez

University of Puerto Rico at Mayagüez University of Puerto Rico at Mayagüez
70 PUBLICATIONS   463 CITATIONS    20 PUBLICATIONS   59 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Geophysical Retrieval Algorithms View project

CenSSIS Education View project

All content following this page was uploaded by Nazario D. Ramirez-Beltran on 13 January 2015.

The user has requested enhancement of the downloaded file.

 12th WSEAS International Conference on SYSTEMS, Heraklion, Greece, July 22-24, 2008

Validation and Strategies to Improve the Hydro-Estimator and

NEXRAD over Puerto Rico
NAZARIO D. RAMIREZ-BELTRAN
Department of Industrial Engineering
University of Puerto Rico
P.O. Box 9030, Mayagüez, PR 00681
U.S.A

ROBERT J. KULIGOWSKI
NOAA/NESDIS Center for Satellite Applications and Research (STAR)
Camp Springs, MD 20746
U.S.A.

ERIC W. HARMSEN
Department of Agricultural and Biosystems Engineering
University of Puerto Rico
P.O. Box 9030, Mayagüez, PR 00681
U.S.A.

JOAN M. CASTRO
Department of Computer and Electrical Engineering
University of Puerto Rico
P.O. Box 9030, Mayagüez, PR 00681
U.S.A

SANDRA CRUZ-POL
Department of Computer and Electrical Engineering
University of Puerto Rico
P.O. Box 9040, Mayagüez, PR 00681
U.S.A

MELVIN J. CARDONA-SOTO
Department of Computer and Electrical Engineering
University of Puerto Rico
P.O. Box 9030, Mayagüez, PR 00681
U.S.A

ISBN: 978-960-6766-83-1 799 ISSN: 1790-2769

 12th WSEAS International Conference on SYSTEMS, Heraklion, Greece, July 22-24, 2008

Abstract - Validation of the Hydro-Estimator (HE) and the Next Generation Radar (NEXRAD) during heavy storms over
Puerto Rico (PR) is reported. The HE is a high resolution rainfall retrieval algorithm based on satellite and numerical
whether prediction model data. The accuracy of the HE and the NEXRAD rainfall estimates can be measured by
decomposing the rainfall process in sequences of discrete (rain / no rain) and continuous (rainfall rate) random variables.
Validation results are based on five heavy storms that seriously impacted human life and the economy of PR during the
period 2003 to 2005. The average discrete validation results indicate acceptable hit rate values for both the HE and
NEXRAD (0.76 vs. 0.87) and reasonable discrete bias ratios (1.04 vs. 0.73) but a very low of probability of detection of
rain for both the HE and NEXRAD (0.36 vs. 0.52). The HE shows an average overestimation whereas the NEXRAD
exhibits underestimation in the continuous validation results (continuous bias ratio of 1.14 vs 0.70 for NEXRAD), which
contributes to moderate overall errors for the HE and NEXRAD in terms of root mean squared error (2.14 mm vs. 1.66
mm) and mean absolute error (0.96 mm vs. 0.77 mm).
The HE algorithm was designed to operate over US continental areas and satisfactory results have been reported.
However, over tropical regions it was determined that warm clouds can generate substantial rainfall amounts that are not
detected by the HE algorithm. Infrared band differencing techniques are using to explore the possibility of improving the
detection of warm-cloud rain events over PR. We are also classifying clouds based on Geostationary Operational
Environmental Satellite (GOES) Imager data in a manner that will lead to improved relationships between infrared
brightness temperatures and rainfall rates.

Key-words - validation, NEXRAD, Hydro-Estimator, retrieval algorithm, rain rate, GOES, brightness temperature.

1. Introduction Data and Information Service (NESDIS) are the Hydro-

Estimation of rainfall amounts is critical for protecting Estimator (HE, [1]), GOES Multi-Spectral Rainfall
human lives and infrastructure, particularly in the case of Algorithm (GMSRA, [2]), and the Self-Calibrating
heavy rainfall that triggers flash floods or landslides. In Multivariate Precipitation Retrieval (SCaMPR, [3]). The
Puerto Rico (PR) during 2003 to 2005, five severe storms HE relies on GOES data from the infrared (IR) window
seriously impacted human lives and the economy. PR has channel (10.7 µm) with a fixed relationship to rainfall
extremely diverse terrain, and during the rainy season rates; similarly, Palmeira et al. [4] presented a self-
severe rainstorms can develop due to complex orographic consistent algorithm for rainfall estimation based on
attributes. Easterly winds come from the eastern Atlantic GOES data plus lightning data in Brazil. The GMSRA
almost all year and play an important role in bringing uses additional data from three other GOES channels and
humidity into the island and stimulating orographic rainfall updates its calibration in real time based on matches with
over the mountains of PR. Cold fronts dominate the radar rain rates. SCaMPR calibrates GOES IR parameters
weather pattern during wintertime. Tropical waves occur against passive microwave rain rates, which is an approach
during the rainy season and frequently generate large similar to Kidd et al. [5] and the Precipitation Estimation
amounts of rainfall in the Caribbean basin. These tropical from Remotely Sensed Information using Artificial Neural
waves are typically the precursor of tropical storms and Network (PERSIANN, [6]) algorithm. PERSIANN uses
hurricanes from June to November. the combined geostationary infrared and the Tropical
For these types of events, estimates of rainfall Rainfall Measuring Mission (TRMM) microwave
from instruments on geostationary platforms such as the information to estimate rainfall rate in an hourly basis at
Geostationary Operational Environmental Satellite spatial resolution of 0.25o. Another algorithm called the
(GOES) are preferred over microwave-based estimates of CPC Morphing Algorithm (CMORPH, [7]) also combines
rainfall from Low-Earth-Orbiting (LEO) platforms IR data and microwave rain rates, but uses the IR data as
because of the rapid refresh every 15 minutes over the the basis for interpolating the microwave rain rates in time
Continental United States (CONUS) and nearby regions between low-Earth orbit satellite overpasses.
and very short data latency times of GOES data relative to The HE, which will be the focus of this paper, also
low-Earth orbit data. Numerous algorithms have been uses information from numerical whether prediction
developed to estimate precipitation from GOES-based models to estimate rain rate [1]. Rainfall rates are adjusted
satellite data. The current generation of algorithms upward or downward for moist or dry environments as
produced at the National Oceanic and Atmospheric indicated by National Centers for Environmental
Administration (NOAA) National Environmental Satellite, Prediction (NCEP) North American Model (NAM) or
Global Forecast System (GFS) total column precipitable

ISBN: 978-960-6766-83-1 800 ISSN: 1790-2769

 12th WSEAS International Conference on SYSTEMS, Heraklion, Greece, July 22-24, 2008

water and mean-layer relative humidity for the lowest third 2. Data collection
of the model vertical domain. Another adjustment Puerto Rico has a rain gauge network that collects rainfall
enhances rainfall rates in regions where the convective measurements every 5, 10, 15, 30 or 45 minutes and
equilibrium level temperature is relatively high; i.e., includes 125 rain gauges with data available since January
regions where very cold cloud tops are not 2000. Since the majority of gauges collect rainfall every
thermodynamically possible but where strong updrafts and 15 minutes a computer program was designed to derive
heavy rainfall can still occur. Finally, low-level winds and HE and NEXRAD 15 minutes resolution. The rain gauge
digital topography are combined to produce enhancements data are used to perform validation of the HE and the
of rainfall rates in upslope regions and reductions in NEXRAD. The data set used for validation includes five
downslope regions, using a technique described in Vicente heavy storms that have been impacted PR: Three can be
et al. [8]. characterized as a cold front and two as tropical storms.
The HE has been the operational satellite rainfall NEXRAD data over Puerto Rico come from a
algorithm of the National Environmental Satellite, Data, WSR-88D unit located in Cayey (18.12°N, 66.08°W,
and Information Service (NESDIS) since 2002 and 886.63 m elevation). The radar frequency is 2.7 GHz and
produces rainfall estimates at the full spatial and temporal the maximum horizontal range is 462.5 km, and the radar
resolution of GOES over the CONUS and surrounding scans the entire island every 6 minutes. The NOAA
regions, including PR; real-time estimates are also National Severe Storms Laboratory (NSSL) conducted a
produced on an experimental basis for the rest of the significant effort to make possible an affordable
globe. However, validation of the Hydro-Estimator has nationwide operational capture, distribution, and archiving
generally focused on the CONUS (e.g., [1] and [9]) and of Level II NEXRAD data [10]. Unfortunately, for Puerto
has not been performed over Puerto Rico, and given the Rico the Level II data are available only until 2003 with a
differences in topography and climate of Puerto Rico significant amount of missing data in that last year [11].
relative to the CONUS, previous validation efforts may not The NWS did resume archiving level II data for PR during
necessarily be relevant to users in PR. Furthermore, the summer of 2007. On the other hand, Level III data for
validation of the HE over PR may illuminate opportunities PR are available continuously since 2000 [12], so the
to enhance the algorithm for application over PR. Level III data were selected to perform validation since the
Validation of the rainfall retrieval algorithm most recent and catastrophic floods over PR occurred after
consists of comparing the rainfall estimates with 2002. The scanning angle for reflectivity data was
corresponding observations (rain gauges in this study). selected as 0.5 degrees for this research in order to avoid
The accuracy of rainfall estimates can be measured by beam overshoot over western PR. Fig. 1 shows the
decomposing the rainfall process as sequences of discrete location of the radar and the spatial distribution of the rain
and continuous random variables; i.e., the presence or gauges.
absence of rainfall events (discrete variable) and the
amount of rainfall (continuous variable). The occurrence
of rainfall events in a given area and at a particular time
follows a Bernoulli process and consequently the
estimation accuracy of rainfall events can be conducted by
analyzing a contingency table. The typical scores that
measure the accuracy of categorical forecasts are: hit rate
(H), probability of detection (POD), false-alarm rate
(FAR), and discrete bias (DB). The continuous validation
strategy consists of comparing the amount of rainfall that
occurred at specific area in a particular time and the
continuous measurements of accuracy are: mean absolute FIG. 1. Location of rain gauges (stars) and NEXRAD
error (MAE), root mean squared error (RMSE), and (black dot on Region I) in PR. Black dots in Region III are
continuous bias (CB). small and high resolution radar that will be used to derive
The second section of this paper describes the data the bias correction factor for NEXRAD. The radii of the
collection process and source of information. The third circles are 35 km and 90 km.
section describes the conventional statistical techniques to
perform validation. The fourth section presents validation As mentioned in the Introduction, the HE uses
results during heavy storms over PR, and includes a satellite IR window (10.7 µm) data and numerical whether
comparison for rain gauges versus HE and rain gauges prediction data to estimate rainfall over the CONUS and
versus NEXRAD. The fifth section presents some PR every 15 minutes at 4 km spatial resolution, and they
strategies for algorithm improvements. The sixth section are available for the entire period of interest. In order to
presents some conclusions. ensure consistency among these data sets during the

ISBN: 978-960-6766-83-1 801 ISSN: 1790-2769

 12th WSEAS International Conference on SYSTEMS, Heraklion, Greece, July 22-24, 2008
12th WSEAS International Conference on SYSTEMS
comparison, both the NEXRAD and HE rain rates were a b
aggregated in time over the corresponding 15-minute DB (4)
a c
accumulation period of the gauges.
where H is the hit rate, POD is the probability of detection,
3. Validation techniques FAR is the false-alarm rate, and DB is the discrete bias.
Validation of the rainfall retrieval algorithm consists of Hit rate is the fraction of the no estimating occasions when
comparing the rainfall estimates with observations over the
same time and space. The accuracy of rainfall estimates the categorical estimation correctly determines the
can be measured by decomposing the rainfall process into occurrence of rainfall event or nonevent. Probability of
sequences of discrete and continuous random variables, detection is the likelihood that the event would be
i.e., the presence or absence of rainfall events and the estimated, given that it occurred. The false-alarm rate is
amounts of rainfall. The occurrence of rainfall events in a the proportion of estimated rainfall events that fail to
given area and at a particular time follows a Bernoulli materialize. Bias is the ratio of the number of estimated
process and consequently the estimation accuracy of rainfall events to the number of observed events [13].
rainfall events can be conducted by analyzing contingency The continuous validation strategy consists of
tables, bivariate probability distribution of rainfall events comparing the amount of rainfall that occurred with the
[13]. Table 1 shows the classical two-way contingency estimated amount of rainfall at specific area in a particular
table. time and the continuous accuracy scores used here are:
It is assumed that the values provided by the rain
gauges are the “ground truth” while the HE and the eij yij yŷˆij i 1,, n and j 1,  , m (5)
NEXRAD provide estimated rainfall values. The variable
a in the contingency table is the number of times that the 1 n m
rain gauge identifies a rainfall event and the estimator also MAE eij (6)
correctly identifies a rainfall event at the same time and
n mi 1j 1
space. The variable d represents the number of times the
rain gauge does not observe a rainfall event and the 1 n m
estimator correctly determines that there is no rainfall RMSE ei2 (7)
event. The variable b indicates the number of times the nm i 1 j 1

rain gauge does not observe a rainfall event but the

estimator incorrectly indicates that there is a rainfall event. n m

The variable c shows the number of times that the rain yˆ ij

i 1 j 1
gauge detects a rainfall event but the estimator incorrectly CB n m
(8)
does not detect the rainfall event. yij
i 1 j 1

TABLE 1. Sample contingency table.

where y and ŷ are the observed and estimated amount of
Observed rainfall rainfall. The i and j subscripts represent time and space,
(Rain gauge) respectively. The constant n is the total number of time
Yes No intervals for a given storm, and m is the number of rain
Estimated rainfall Yes a b gauges that are collecting rain during a storm. The error e
(HE or NEXRAD) No c d is the deviation between the observed and estimated
amount of rainfall at a particular time and space and is
The typical scores that measure the accuracy of computed only when at least one of y or ŷ is greater than
categorical estimation are: zero. MAE is the mean absolute error, RMSE the root
mean squared error, and CB is the continuous bias.
a d, where no a b c d (1)
H
no 4. Validation results

a 4.1 Discrete validation

POD (2) A contingency table was computed for each rain gauge
a c during a given storm and the scores of those tables were
summarized to create contingency tables for each storm for
b the HE and NEXRAD which are shown in Tables 2a) and
FAR (3)
a b 2b) while the associated scores are given in Tables 3a) and
3b). The HE significantly underestimates the number of
ISBN:
ISBN: 978-960-6766-83-1
978-960-6766-83-1 802
802 ISSN: 1790-2769
ISSN: 1790-2769
 12th WSEAS International Conference on SYSTEMS, Heraklion, Greece, July 22-24, 2008

raining pixels in the three April-May events (DB of 0.49 to b)

0.52) but strongly overestimates the November-December 17 April 2003 Rain Gauge
Yes No
events (DB of 1.54 and 2.15). The physical reasons behind NEXRAD Yes 2713 951
this apparent strong seasonal variation in DB are not No 1023 6311
known at this time. Meanwhile, the NEXRAD had a
consistent dry bias (0.62-0.68) for the last four events but 19-21 May, 2003 Rain Gauge
virtually no bias (1.02) for the first; again, it is not clear at Yes No
this time what led to such a significant difference. The hit NEXRAD Yes 1177 1149
No 399 30386
rates of the HE range from 0.62 to 0.91 with an average of
0.76 and NEXRAD has a range from 0.82 to 0.95 with 11-18 November 2003 Rain Gauge
average of 0.87. Although, both HE and NEXRAD exhibit Yes No
relatively high hit rate, the HE has a lower percentage of NEXRAD Yes 8922 9967
correct rain / no rain estimates than does the NEXRAD. No 3620 62901
The probability of detection of the HE ranges from 0.14 to
5 December 2003 Rain Gauge
0.57 with an average value of 0.36, whereas, the Yes No
NEXRAD shows a range from 0.4 to 0.74 with an average NEXRAD Yes 3392 5026
of 0.52. Thus, the HE correctly detected a smaller No 1814 34462
percentage of the observed rainfall events (36%) than did
NEXRAD (52%) for these events. The false alarm rate for 20 April 2005 Rain Gauge
the HE varies between 0.39 to 0.73 with an average value Yes No
NEXRAD Yes 655 670
of 0.61, meanwhile the NEXRAD varies from 0.25 to 0.35
No 240 8583
with an average of 0.29. Thus, the false alarm rate was
actually higher for the HE (61%) than for NEXRAD
(29%). Thus, the discrete validation shows that the
TABLE 3. Discrete validation scores for the Hydro-
NEXRAD outperforms the HE in terms of correct rain / no Estimator and NEXRAD.
rain estimates.
a) 17 19-21 11-18 20
TABLE 2. Contingency tables for (a) the Hydro-Estimator 5 Dec.
Apr. May Nov. Apr. Avg.
2003
and (b) NEXRAD. 2003 2003 2003 2005
DB 0.49 0.52 1.54 2.15 0.52 1.04
a) HR 0.70 0.91 0.68 0.62 0.86 0.76
17 April 2003 Rain Gauge POD 0.30 0.14 0.55 0.57 0.23 0.36
Yes No FAR 0.39 0.72 0.64 0.73 0.56 0.61
Hydro-Estimator Yes 1105 699
No 2603 6708 b) 17 19-21 11-18 20
5 Dec.
Apr. May Nov. Apr. Avg.
2003
19-21 May, 2003 Rain Gauge 2003 2003 2003 2005
Yes No DB 1.02 0.68 0.66 0.62 0.67 0.73
Hydro-Estimator Yes 331 875 HR 0.82 0.95 0.84 0.85 0.91 0.87
No 2000 30022 POD 0.74 0.51 0.47 0.40 0.49 0.52
FAR 0.27 0.25 0.29 0.35 0.27 0.29
11-18 November, 2003 Rain Gauge
Yes No
Hydro-Estimator Yes 10430 18719
4.2 Continuous validation
No 8465 47913
The accumulated rainfall across the island was computed
5 December 2003 Rain Gauge to compare the observed and the estimated rainfall:
Yes No
Hydro-Estimator Yes 4882 13224
(9)
No 3538 23167

20 April 2005 Rain Gauge where Yi is the total rainfall recorded by all 125 rain
Yes No gauges across the island or the closest HE or radar pixels
Hydro-Estimator Yes 310 395
No 1039 8522
at the i th time.

Tables 4a) and 4b) show the continuous validation

scores for HE and NEXRAD, respectively. The
continuous bias of the HE is even more variable than the

ISBN: 978-960-6766-83-1 803 ISSN: 1790-2769

 12th WSEAS International Conference on SYSTEMS, Heraklion, Greece, July 22-24, 2008

DB, with values ranging from 0.16-0.26 for the April-May correspond with convective cloud tops that are above the
storms and 1.68-2.42 for the November-December events. tropopause (i.e. overshooting tops), ([14 and [15]).
The lower CB relative to the DB for the April-May storms Convective clouds with positive differences indicate the
suggests that the HE is underestimating the conditional possibility of warm-top convection. We will also explore
rainfall rates in addition to the spatial extent of the rainfall, the use of the reflected portion of the 3.9-µm GOES band
while the opposite is happening for the November- 2 during the daytime to indicate the presence large cloud-
December events. The NEXRAD has nearly no top particles that suggest rain in warm-top clouds [2].
continuous bias for two storms and a strong dry bias for
three (0.41-0.68), albeit with no apparent seasonal pattern
like the HE. As a result, both the mean absolute error and RG and HE Cumulated Rainfall
90
root mean squared error of the HE are also higher than that 80
Rain Gauges
Hydroestimator
of NEXRAD.
70

60
Table 4. Continuous validation scores for the Hydro-
Estimator and NEXRAD. 50

mm
a) 17 19-21 11-18 5 20 Avg. 40

Apr. May Nov. Dec. Apr. 30

2003 2003 2003 2003 2005
CB 0.26 0.23 1.68 2.42 0.16 0.95 20

MAE (mm) 1.33 0.74 1.10 0.86 0.79 0.96 10

RMSE (mm) 2.73 2.10 2.24 1.93 1.71 2.14 0

318.0104 318.2604 318.5104 318.7604
time (hours UTC)

b) 17 19-21 11-18 5 20 Avg.

Apr. May, Nov. Dec. Apr.
FIG. 2. Comparison between observed and estimated
2003 2003 2003 2003 2005 accumulated rainfall (Nov. 14, 2006).
CB 1.02 0.68 0.41 0.42 1.01 0.71
MAE (mm) 1.02 0.66 0.85 0.53 0.80 0.77
RMSE (mm) 1.91 1.79 1.78 1.15 1.68 1.66

5. Algorithm improvements
5.1 Rainfall detection
As stated previously, the HE uses GOES brightness
temperatures (Tb) from channel 4 (10.7 µm) to
discriminate raining from non-raining events [1]. During
the validation exercise we noted that there are some warm-
top convective events that are not detected by the HE. The
HE generally produces little or no rainfall for brightness
temperatures exceeding 235K; however, there are
numerous events in PR where the HE largely failed to FIG. 3. GOES-12 brightness temperature from
detect significant rainfall. For instance, Fig. 2 shows the
channel 4 (Nov. 14, 2006).
observed accumulated rainfall for all gauges located in PR
(red line) and the accumulated rainfall by the
5.2 Improving rain rate estimates
corresponding HE pixels (blue line) on November 14,
The rainfall retrieval procedure of the HE is also mainly
2006. The horizontal axes shows the time every 15
based on the relationship between the brightness
minutes and the vertical axis exhibits the accumulated
temperature (10.7 µm) and observed rain rate. Estimation
rainfall in mm. Fig. 3 shows the distribution of brightness
of the amount of rainfall may be improved by classifying
temperatures over the GOES pixels corresponding to
the brightness temperature patterns (BTP) with the
gauge locations during this storm and there are few pixels
below 235 K; a comparison with Fig. 2 indicates that the corresponding rain formation processes. The following
channels will be used to classify the BTP with the
poor detection by the HE was at least in part because it
corresponding rain process. Channel 1 (0.65 µm) will be
was not calibrated to produce rainfall from relatively warm
used to classify the events according to the cloud optical
clouds. In order to improve the detection skill of the HE,
thickness. The reflected portion of channel 2 (3.9 µm)
we plan to examine the differences in brightness
temperature between 10.7 µm and the water vapor band during the daytime will be used as an indirect
measurement of the cloud drop size distribution,
(6.5 µm in GOES-12). Positive values of the WV-infrared
thermodynamic phase, and particle shape [16]. Channel 4
window temperature difference have been shown to
ISBN: 978-960-6766-83-1 804 ISSN: 1790-2769
 12th WSEAS International Conference on SYSTEMS, Heraklion, Greece, July 22-24, 2008

(10.7 µm) will be used to classify the rainfall events the HE for PR; specifically, the algorithm proposed by
according to temperature. The infrared band differencing Ramirez-Beltran et al. [17] will be implemented to
will also be used to develop the classification algorithm. improve the HE rainfall detection and the equation that
The difference between the 10.7-µm brightness relates brightness temperatures with rain rates.
temperature and 3.9-µm will be useful to determine
whether a cloud top is composed of liquid water or ice. As 7 Acknowledgements.
stated previously, the IR-WV difference (6.5–10.7 µm) is
usually negative; however, convective clouds with positive This research has been supported by NOAA-CREST grant
differences have likely already begun to precipitate, number NA17AE1625, the NSF-ERC-CASA with a grant
especially in tropical atmospheres that support warm top Number 0313747, NOAA-NWS grant number
convection. The 13.3–10.7 µm differencing technique is NA06NWS468001, and also by the University of Puerto
used to characterize and delineate cumulus clouds. This Rico at Mayagüez. The authors appreciate and recognize
research will focus on convective clouds, and the funding support from these institutions.
consequently, the factors to be consider for the
classification of BTP and rain types are: area, depth, References
duration, and updraft velocity.
A variable selection algorithm will be used to [1] Scofield, R.A., and R.J. Kuligowski, 2003: Status and
identify the variables that best explain the rainfall outlook of operational satellite precipitation
variability, and the selected variables will be used to algorithms for extreme-precipitation events. Wea.
develop the training patterns for a self-organized artificial Forecasting, 18, 1037-1051.
neural network [17]. A self-organized artificial neural [2] Ba, M. B., and A. Gruber, 2000: GOES Multispectral
network will be used to identify a set of homogenous Rainfall Algorithm (GMSRA). J. Appl. Meteor., 40,
groups that reveal similarities within the member of a 1500-1514.
class, but different among the classes. The Kohonen [3] Kuligowski, R.J., 2002: A Self-Calibrating real-time
learning rule will be used to determining the optimal GOES rainfall algorithm for short-term rainfall
weights of the artificial neural network ([18], [19], and estimates. J. Hydrometeor., 3, 112-130.
[20]). A successful application was reported [21] to [4] Palmeira, F. L. B., C. A. Morales, G. B. França, and L.
identify the spatial variability of soil to select the Landau, 2004: Rainfall estimation using satellite
appropriate model to estimate soil moisture.
data for Paraíba do Sul Basin (Brazil). The XX th
ISPRS International Society for Photogrammetry
6 Summary and conclusions and Remote Sensing Congress, Istanbul, Turkey -
The HE is a high resolution satellite rainfall retrieval Commission 7.
algorithm run operationally by NOAA/NESDIS that [5] Kidd, C., Kniveton, D.R., Todd, M.C., Bellerby, T.J.,
provides estimates of rainfall every 15 minutes at 4-km 2003: Satellite Rainfall Estimation Using
resolution over the CONUS and nearby areas including Combined Passive Microwave and Infrared
PR. (Global estimates are also produced in real time on an Algorithms. J. Hydrometeor., 4, 1088-1104.
experimental basis.) The rain rates are primarily derived [6] Sorooshian, S., K.-L. Hsu, X. Gao, H. V. Gupta, B.
from GOES 10.7-µm brightness temperatures and then Imam, and D. Braithwaite, 2000: An evaluation of
adjusted using parameters derived from a numerical PERSIANN system satellite-based estimates of
weather prediction model. The HE estimates should be tropical rainfall. Bull. Amer. Meteor. Soc., 81,
especially useful over regions of complex topography such 2035-2046.
as western PR because of the difficulties associated with [7] Joyce, R.J., J. E. Janowiak, P.A. Arkin, and P. Xie,
radar in those regions such as beam block. However, for 2004: CMORPH: A method that produces global
the very small sample of heavy rainfall events examined in precipitation estimates from passive microwave
this paper, NEXRAD clearly outperforms the HE, perhaps and infrared data at high spatial and temporal
in part because of most of the rainfall events were located resolution. J. Hydrometeor., 5, 487-503.
in the central and eastern parts of the island where the [8] Vicente, G. A., J. C. Davenport, and R. A. Scofield,
radar data will be most reliable. Specifically, the HE 2001: The role of orographic and parallax
underestimates both the number of rainfall events and the corrections on real time high resolution satellite
amounts of rainfall, whereas NEXRAD is nearly unbiased rainfall rate distribution. Int. J. Remote Sens., 23,
in these respects. The HE algorithm does exhibit a 221-230.
satisfactory hit rate, but a very low probability of detection [9] Ebert, E. E., J. E. Janowiak, and C. Kidd, 2007:
and a large false alarm rate that is surprisingly higher than Comparison of near-real-time precipitation
that of NEXRAD despite the dry bias of the HE. A estimates from satellite observations and
research effort is undergone to improve the performance of
ISBN: 978-960-6766-83-1 805 ISSN: 1790-2769
 12th WSEAS International Conference on SYSTEMS, Heraklion, Greece, July 22-24, 2008

numerical models. Bull. Amer. Meteor. Soc., 88,

47-64.
[10] Kelleher, K.E., K. K. Droegemeier, J. J. Levit, C.
Sinclair, D. E. Jahn, S. D. Hill, L. Mueller, G.
Qualley, T. D. Crum, S. D. Smith, S. A. Del
Greco, S. Lakshmivarahan, L. Miller, M.
Ramamurthy, B. Domenic, and D. W. Fulker,
2007: A real-time delivery system for NEXRAD
Level II data via the internet. Bull. Amer. Meteor.
Soc., 88, 1045-1057.
[11] NCDC, 2005a: National Climatic Data Center: Data
Documentation for DSI – 6500 – NEXRAD Level
II. NCDC, Asheville, NC.
[12] NCDC, 2005b: National Climatic Data Center: Data
Documentation for DSI – 7000 – NEXRAD Level
III. NCDC, Asheville, NC.
[13] Wilks, D.S., 1995: Statistical Methods in the
Atmospheric Sciences: An Introduction.
Academic Press, San Diego, 467 pp.
[14] Ackerman, S. A., 1996: Global satellite observations
of negative brightness temperature differences
between 11 and 6.7 mm. J. Atmos. Sci., 53, 2803-
2812.
[15] Schmetz, J., H. P. Roesli and W. P. Menzel, 1997:
Third International Winds Workshop (Meeting
summary). Bull. Amer. Meteor. Soc., 78, 893 – 896.
[16] Arking, A., and J. D. Childs, 1985: Retrieval of cloud
cover parameters from multispectral satellite images.
J. Climate Appl. Meteor., 24,
[17] Ramirez-Beltran, N.D., W.K.M. Lau, A. Winter, J.M.
Castro, and N.R. Escalante, 2007: Empirical
probability models to predict precipitation levels
over Puerto Rico stations. Mon. Wea. Rev. 135,
877-890.
[18] Hagan, T.H., Demuth, H.B., and Beal, M., 1996:
Neural Network Design, PWS Publishing Co.,
Boston.
[19] Ramirez-Beltran, N.D. and Montes, J.A. 2002:
Neural networks to model dynamic systems with
time delays. IIE Transactions, 34, 313-327.
[20] Ramirez-Beltran, N.D., and A. Veneros 2004: Upper
air information and neural networks to estimate
hurricane intensity. Preprints, 26th Conference on
Hurricanes and Tropical Meteorology. Miami FL.
[21] Ramírez Beltran, N.D, J. M. Castro., E. Harmsen,
and R. Vasquez. 2008: Stochastic transfer
function models and neural networks to estimate
soil moisture. J. of the Amer. Water Resources
Assoc. in press.

ISBN: 978-960-6766-83-1 806 ISSN: 1790-2769

View publication stats