ArticlePDF Available

Abstract and Figures

Extreme rainfall may trigger some of the most catastrophic natural disasters, whose consequences may be exacerbated especially in places where an appropriate network of measurement instruments is not available. A combination of remotely sensed data and weather prediction model outputs can often help to obtain information with a global spatial coverage without the limitations that characterize other instruments. In order to achieve this goal, an Extreme Rainfall Detection System (ERDS – erds.ithacaweb.org) was developed and implemented with the aim of monitoring and forecasting exceptional rainfall events. The system was designed with the aim of providing information in an understandable way also for non-specialized users. The NOAA-GFS deterministic weather prediction model is used for the purpose of forecasting extreme rainfall events. Regarding the near real-time rainfall monitoring, the previous version of ERDS was using NASA TRMM TMPA 3-hourly data as input. Due to TRMM instrument shutdown, a different rainfall measurement must be used. NASA GPM IMERG early run half-hourly data proved to be the proper one. A comparison between GPM and rain gauge data allowed to defining the minimum time aggregation intervals to be used for the detection of extreme rainfall events in order to reduce the effects of the bias due to satellite data. The same comparison was also performed using GFS data instead of GPM data. A new extreme rainfall detection methodology was also developed with the aim of increasing system performances. The currently adopted methodology is based on the concept of event-identification threshold. A threshold represents the amount of precipitation needed to trigger a flood event induced by extreme rainfall. Specifically, if for a selected aggregation interval the accumulated precipitation exceeds the threshold, an alert is provided. Obtained results highlighted that the combination of new input data and new threshold methodology allowed one to increase system performances, both in terms of spatial and temporal resolution and in terms of identified events.
Content may be subject to copyright.
S1423
ГОДИШНИК НА УНИВЕРСИТЕТА ПО АРХИТЕКТУРА, СТРОИТЕЛСТВО И ГЕОДЕЗИЯ
СОФИЯ
Том
52
2019
Брой
S1
Volume
Issue
ANNUAL OF THE UNIVERSITY OF ARCHITECTURE, CIVIL ENGINEERING AND GEODESY
SOFIA
Получена: 03.06.2019 г.
Приета: 17.06.2019 г.
ERDS: AN EXTREME RAINFALL DETECTION SYSTEM BASED ON
BOTH NEAR REAL-TIME AND FORECAST RAINFALL
MEASUREMENTS
P. Mazzoglio
, F. Laio
, S. Balbo
, P. Boccardo
Keywords: early warning system, extreme events, flood monitoring, GPM, rainfall
ABSTRACT
Extreme rainfall may trigger some of the most catastrophic natural disasters, whose
consequences may be exacerbated especially in places where an appropriate network of
measurement instruments is not available. A combination of remotely sensed data and weather
prediction model outputs can often help to obtain information with a global spatial coverage
without the limitations that characterize other instruments. In order to achieve this goal, an
Extreme Rainfall Detection System (ERDS erds.ithacaweb.org) was developed and
implemented with the aim of monitoring and forecasting exceptional rainfall events. The
system was designed with the aim of providing information in an understandable way also for
non-specialized users. The NOAA-GFS deterministic weather prediction model is used for the
purpose of forecasting extreme rainfall events. Regarding the near real-time rainfall
monitoring, the previous version of ERDS was using NASA TRMM TMPA 3-hourly data as
input. Due to TRMM instrument shutdown, a different rainfall measurement must be used.
NASA GPM IMERG early run half-hourly data proved to be the proper one. A comparison
between GPM and rain gauge data allowed to define the minimum time aggregation intervals
to be used for the detection of extreme rainfall events in order to reduce the effects of the bias
due to satellite data. The same comparison was also performed using GFS data instead of GPM
data. A new extreme rainfall detection methodology was also developed with the aim of
Paola Mazzoglio, THACA Information Technology for Humanitarian Assistance, Cooperation and
Action, Torino, Italy (paola.mazzoglio@ithaca.polito.it)
Francesco Laio, Politecnico di Torino, Torino, Italy (francesco.laio@polito.it)
Simone Balbo, ITHACA Information Technology for Humanitarian Assistance, Cooperation and
Action, Torino, Italy (simone.balbo@ithaca.polito.it)
Piero Boccardo,Politecnico di Torino, Torino, Italy (piero.boccardo@polito.it)
S1424
increasing system performances. The currently adopted methodology is based on the concept
of event-identification threshold. A threshold represents the amount of precipitation needed to
trigger a flood event induced by extreme rainfall. Specifically, if for a selected aggregation
interval the accumulated precipitation exceeds the threshold, an alert is provided. Obtained
results highlighted that the combination of new input data and new threshold methodology
allowed one to increase system performances, both in terms of spatial and temporal resolution
and in terms of identified events.
1. Introduction
Several studies highlighted an increasing trend in the number of hydrological and
meteorological disasters. Early warning system based on remotely-sensed data and numerical
weather prediction models can often help in the detection and monitoring of extreme rainfall at
the global scale. These systems have an increasingly important role in the disaster risk
management, especially for the triggering of flood preparedness actions [1].
The investigation of available dataset and approaches led several institution and
research groups to develop new techniques for rainfall and/or flood monitoring. From this
perspective, ITHACA developed an Extreme Rainfall Detection System (ERDS
erds.ithacaweb.org). ERDS is a demo service for the monitoring and forecasting of exceptional
rainfall events, with a nearly global spatial coverage. ERDS is also an alert system designed to
identify hydrometeorological events (such as hurricanes, tropical storms, convective storms,
flash floods and heavy rainfall that could lead to flood events or landslides) [8]. This system
provides information both on the accumulated rainfall and on heavy rainfall alerts.
The first objective of this study is the assessment of the accuracy of two different
datasets used as input data for different aggregation intervals. The obtained results allowed the
definition of the proper aggregation intervals usable to provide information regarding the
rainfall amount and to evaluate the presence of places potentially affected by
hydrometeorological disasters. The second objective, instead, is the development of an extreme
rainfall detection methodology applicable at a global scale based on event-identification
thresholds calibrated taking into consideration the mean annual rainfall as site-specific
parameter.
One of the aspects to be taken into consideration is the non-coincidence between the
place where the alert was given and the place where the flood may occur. ERDS is a tool
developed for rainfall monitoring. The system provides an alert where the amount of rainfall is
higher than a specific threshold. The system, however, does not take into account the
morphology of the territory or information regarding basins. The flood, therefore, may occur in
the alerted cell or in nearby ones. Further studies might investigate this important aspect.
2. Input Data
Regarding the near real-time rainfall monitoring, the system takes advantage of NASA
GPM (Global Precipitation Measurement) IMERG (Integrated Multi-satellite Retrievals for
GPM) early run half-hourly data [5]. This gridded and georeferenced product is characterized
by a 0.1° × 0.1° spatial resolution, a spatial coverage between 60° N and 60° S and 4-hour
latency [6].
The rainfall forecasting is instead based on NOAA NCEP GFS (Global Forecast
System) deterministic weather prediction model [9], characterized by a spatial resolution of
S1425
0.25° × 0.25° and a spatial coverage between 9N and 90° S. GFS model runs every day at
00, 06, 12 and 18 UTC.
In this study, both datasets were compared with rain gauges measurements in order to
evaluate the relative accuracy for a set of different aggregation intervals. Fifty rain gauges
located in different climatic zones were taken into account in order to evaluate the accuracy at
a global scale (Fig. 1). Rain gauge dataset characterized by a temporal resolution of at least one
hour were downloaded from the data providers websites [10, 2]. The biased spatial distribution
of the rain gauges is induced by the difficulties encountered in the research of freely accessible
well-recognized dataset characterized by a good temporal resolution and an almost zero
percentage of missing data.
Figure 1. Spatial distribution of the rain gauges used for the evaluation of the accuracy
of the input data. The reference system is WGS84
The analysis was performed using statistical performance scores and time series
analyses in the period from 15th January 2015 to 30th April 2017. The following aggregation
intervals were considered in order to evaluate the cumulated rainfall and the relative intensity:
1, 2, 3, 6, 12, 24, 48, 72, 96, 120 and 144 hours.
Bias and MAE (mean absolute error) were evaluated for both dataset, for each location
and for the previously mentioned set of aggregation intervals, as
1
1( ) ( )
n
ESTIMATOR GAUGE
t
BLAS R t R t
n



, (1)
1
1( ) ( )
n
ESTIMATOR GAUGE
t
MAE R t R t
n



, (2)
where
n
is the total number of time instants;
t
time;
ESTIMATOR
R
the average rainfall intensity measured by GPM Mission or provided by
GFS weather prediction model (expressed in mm/hr);
S1426
GAUGE
R
the average rainfall intensity measured by rain gauge in the same time
interval
t
(expressed in mm/hr).
Both parameters were evaluated both in standard conditions (taking into account only
nonzero rainfall measurements) and in heavy-rainfall conditions (considering only rainfall rate
greater than 99th percentile of the intensities distribution).
The second part of the analysis consisted of the comparison of the observed and the
estimated events. For every location and for every aggregation interval previously mentioned, a
contingency table (Tab. 1) was created. In the contingency table, the occurrences of the
following four conditions were reported:
both rain gauge and estimator data are null (case A, correct negatives);
nonzero rain gauge data and zero estimator data (case B, misses);
zero rain gauge data and non-null estimator data (case C, false alarms);
both rain gauge and estimator data are non-null (case D, hits).
Table 1. Example of a contingency table usable to evaluate the quality of the predictions
RAIN GAUGE
= 0 mm/hr
> 0 mm/hr
ESTIMAT
OR
= 0 mm/hr
Correct negatives
(A)
Misses
(B)
Estimated
non-events
> 0 mm/hr
False alarms
(C)
Hits
(D)
Estimated
events
Observed
non-events
Observed
events
Three indices were derived using these contingency tables as a basis: the false alarm
ratio, the probability of detection and the critical success index [11].
The false alarm ratio (FAR) represents the number of false alarms per number of
estimated events and is assessable on the basis of elements contained in false alarms and hits
cells. The ideal situation is characterized by approximately zero FAR.
.
FALSE ALARMS
FAR HITS FALSE ALARMS
(3)
The probability of detection (POD) is instead evaluable by combining elements
contained in the misses and hits cells. The ideal situation is characterized by a unitary POD
value.
.
HITS
POD HITS MISSES
(4)
The critical success index (CSI), unlike POD and FAR, combines the characteristics of
hits, false alarms and misses and it can be expressed in terms of POD and FAR. The ideal
situation is characterized by a unitary CSI.
.
HITS
CSI HITS FALSEALARMS MISSES

(5)
S1427
The outcomes of this analysis allowed to identify the most appropriate aggregation
interval usable to provide information about the rainfall amount and the possible occurrence of
disasters induced by heavy rainfall. The results are summarized in several boxplots (Fig. 2). In
these figures, the first quantile, the median value of the distribution and the third quantile are
identifiable. The mean values are instead summarized in a separate table (Tab. 2).
Figure 2. Bias, MAE, FAR, POD and CSI evaluated for nonzero rainfall intensities: the dark grey
boxplot refers to GPM IMERG V05B data while the light grey boxplot refers to GFS data
S1428
Table 2. Bias, MAE, FAR, POD and CSI calculated taking into account only nonzero
rainfall intensities both for GPM IMERG V05B early run and for GFS dataset
1 hr
2 hr
3 hr
6 hr
12 hr
24 hr
48 hr
72 hr
96 hr
120 hr
144 hr
BIAS
GPM
-0.35
-0.28
-0.23
-0.17
-0.13
-0.09
-0.07
-0.06
-0.05
-0.05
-0.05
GFS
-
-
-
-
-0.59
-0.45
-0.36
-0.32
-0.30
-0.29
-0.23
MAE
GPM
1.97
1.43
1.16
0.80
0.55
0.37
0.27
0.22
0.20
0.19
0.18
GFS
-
-
-
-
0.90
0.67
0.52
0.46
0.43
0.41
0.40
FAR
GPM
0.63
0.62
0.60
0.57
0.52
0.45
0.36
0.30
0.25
0.21
0.18
GFS
-
-
-
-
0.54
0.46
0.36
0.30
0.25
0.21
0.18
POD
GPM
0.52
0.58
0.62
0.69
0.76
0.83
0.89
0.92
0.94
0.96
0.96
GFS
-
-
-
-
0.92
0.95
0.96
0.97
0.97
0.98
0.98
CSI
GPM
0.27
0.30
0.32
0.36
0.42
0.49
0.59
0.66
0.71
0.76
0.79
GFS
-
-
-
-
0.44
0.52
0.62
0.69
0.74
0.77
0.81
From this analysis, a negative value of the bias emerged for both products. As a general
rule, therefore, both the satellite-derived data and the rainfall estimation obtained through GFS
weather prediction model tend to underestimate rainfall with respect to the rain gauge. For both
products, as the aggregation interval increases, the bias tends to have a null value, allowing
more accurate information. Aggregation intervals who matched the optimal accuracy for a near
real-time application are the one greater or equal to 12 hours. Smaller time intervals have also
an unsatisfactory value of false alarm ratio (greater than 0.55). As expected, FAR shows a
decreasing trend for both datasets.
Figure 3. Bias and MAE related to heavy rainfall events. The dark grey boxplot refers to GPM
IMERG V05B data while the light grey boxplot refers to GFS data
As far as GPM data concerns, the outcomes demonstrate that a 24-hours aggregation
interval ensures a probability of detection greater than 80% and a critical success index equal
to 50%. With an aggregation interval of 72 hours, a probability of detection greater than 90%
S1429
was reached. Shorter aggregation intervals are characterized by a not acceptable probability of
detection. It is, therefore, convenient to set the minimum rainfall aggregation interval to 12
hours to be able to detect events with an acceptable accuracy.
The selected aggregation intervals were assessed taking into account several aspects,
such as the requirements of the system, the ideal latency in the provision of information or the
final use of the output. As a consequence, also similar aggregation intervals are suitable for
these purposes. It is, however, important to highlight the limitations in terms of the accuracy of
the outputs obtained using these aggregation intervals.
A further study was conducted taking into account only rainfall rate greater than the
99th percentile of the intensities distribution in order to understand the estimator performances
in heavy rainfall detection. The modest underestimation of GPM and GFS data emerged from
these results (Fig. 3 and Tab. 3).
Table 3. Bias and MAE related to heavy rainfall events evaluated both for GPM IMERG
V05B early run and for GFS data
1 hr
2 hr
3 hr
6 hr
12 hr
24 hr
48 hr
72 hr
96 hr
120 hr
144 hr
BIAS
GPM
-10.59
-9.31
-8.34
-6.54
-4.87
-3.40
-2.29
-1.84
-1.52
-1.33
-1.18
GFS
-
-
-
-
-12.80
-9.18
-6.37
-5.15
-4.37
-3.88
-3.47
MAE
GPM
11.64
10.19
9.09
7.12
5.23
3.64
2.47
2.00
1.67
1.46
1.30
GFS
-
-
-
-
12.80
9.18
6.38
5.15
4.37
3.88
3.47
3. Extreme Rainfall Detection
The main purpose of this section is to describe the extreme rainfall detection
methodology developed in order to provide near real-time alerts with an almost global spatial
coverage.
The whole study (described in detail in Mazzoglio et al [8]), was conducted using GPM
IMERG half-hourly early run data as input. Results presented in this section refer to this data.
This methodology will also be repeated using GFS data as input in order to calibrate the
threshold values that will be used for alerts forecasting.
The extreme rainfall detection methodology is based on the concept of activation
threshold: an event is identified when the rainfall exceeds a given threshold value. An “event-
identification threshold” (EIT) represents the amount of rainfall needed to trigger a flood event
induced by extreme rainfall [8]. EITs are used to define near real-time alerts about extreme
rainfall. Specifically, an alert is provided if for a selected time interval the accumulated rainfall
exceeds the EIT.
The calibration of these thresholds was performed using an empirical approach,
analyzing rainfall events that have led to hydrometeorological disasters. These threshold
values, obviously, vary over time from place to place. Extreme rainfall conditions in one place
are, in fact, very different from the one that characterizes another one. Moreover, it is
impossible to define a threshold if the aggregation interval is not defined. A longer time
interval has, in fact, a higher EIT.
To develop this extreme rainfall detection methodology, the first step was to search
databases of hydrometeorological disasters with a global spatial coverage to be used as truth
data. The adopted databases were EM-DAT (The Emergency Events Database) [3], Reliefweb
[12] and Floodlist [4].
S1430
The methodology consisted of the identification of the optimal EIT for different
aggregation intervals (12, 24, 48, 72 and 96 hours) and for every place of the world. The study
covered 85 different countries, from 12 March 2014 to 30 April 2017. For every temporal
interval and for every country, a series of simulations was performed by varying the threshold
values mask.
Threshold masks (different for every aggregation interval) were calibrated using a site-
specific parameter: the mean annual precipitation. This total rainfall amount (Fig. 4) was
calculated using 10 years of GPCC (Global Precipitation Climatology Centre) “Monitoring
product” [13] with a 1° resolution. The “Monitoring product” is a monthly global data which is
available about 2 months after the end of the month which it refers to. This product is
recommended by GPCC to be used for applications that need high-quality gridded measures of
rainfall.
Figure 4. The mean annual rainfall calculated using 10 years of GPCC monthly
“Monitoring product”. The white areas are places characterized by an absence
of measurement. The reference system is WGS84
For each aggregation interval, a series of simulations using this 1° × 1° total rainfall
amount was performed. In every simulation, thresholds equal to a percentage of the mean
annual rainfall have been adopted. Specifically, the threshold values were calculated using the
equation
..
.. TR
T T R p
(6)
where
T
represents the threshold;
..TR
represents the total rainfall (i.e. the mean annual rainfall calculated using 10 years
of GPCC data);
..TR
p
is a parameter representing the fraction of the total rainfall leading to the extreme
event identification.
The application of an upper bound and a lower bound proved to be necessary. There are
places where the recorded average annual rainfall amount is very low (below 100 mm in a
year), which would lead to very low threshold values, comparable with the satellite
measurement accuracy, with
..TR
p
values around 0.1 0.2. Analogously, in places where the
total annual rainfall is very high (above 2000 mm in a year), the EIT could be unrealistically
high because in these places rainfall events tend to occur in the form of low intensity high
frequency events [8].
S1431
Several tests were performed. For every aggregation interval, both the criterion of
minimization of missed and false alarms and the maximization of the number of identified
events were taken into account in order to choose the proper threshold values.
The best results for every aggregation interval were achieved with the parameters
summarized in Tab. 4.
Table 4. Threshold values currently adopted in ERDS for the near real-time alerts
provision based on NASA GPM IMERG early run data
Aggregation interval
(hours)
T.R.
p
(%)
Lower bound
(mm)
Upper bound
(mm)
12
6
100
150
24
8
120
210
48
12
140
240
72
15
170
260
96
16
190
280
Before publishing on ERDS website, the alerts produced on areas entirely occupied by
sea or ocean are discarded. In order to accomplish this operation, a mask containing, in each
cell, the water coverage of the area was used. This mask is freely available on NASA’s website
[7]. Alerts provided on cells characterized by water coverage equal to 100% were discarded.
ERDS, in fact, is a tool developed for the provision of alerts on populated areas.
4. Conclusion
To sum up, the current version of ERDS automatically downloads the most recent GPM
IMERG early run data and GFS data and cumulates them according to specific time intervals.
More importantly, ERDS generates rainfall alerts based on GPM data where and when the
rainfall amount is higher than a specific set of event-identification thresholds.
The current version of ERDS is able to provide alerts using GPM as input data every
hour with 4-hour latency and a 0.1° × 0.1° resolution in the latitude range between 60° N and
60° S. The ERDS data are uploaded every hour because GPM IMERG data, despite the 30-
minute resolution, is made public in pairs of two.
ERDS is an alert system designed to identify hydrometeorological events. However,
there are some types of phenomena that put a strain on this alert system. Specifically, ERDS
may fail in the identification of convective storms characterized by a high spatial and temporal
variability and discontinuity. They can indeed be transparent to the satellite (rainfall could
affect an area smaller than the size of a single cell of GPM data) or their intensity may be
underestimated. ERDS may also fail in the provision of a timely alert in the case of intense
flash floods affecting very small basins. ERDS, in fact, is characterized by a delay of about 4
hours (due to original GPM data delay plus the time required by the data download, processing
and alert evaluation in the ERDS system). If the event is very short, very intense and can cause
a flood within 4 hours, the alert will be provided too late. Conversely, the system showed a
good performance regarding the identification of hydrometeorological disasters like hurricanes,
cyclones, tropical storms, heavy rainfall that might lead to flood events and flash floods
characterized by durations greater than the ERDS latency [8].
S1432
New improvements are still required to increase the overall accuracy of early warning
systems based on remotely sensed rainfall measurements and weather prediction model
outputs. The system performances are deeply influenced by the input data resolution. The
system is working at the global scale with a spatial resolution of 0.1° × 0.1°. These
characteristics could lead to a wrong picture of rainfall events that vary greatly on a small scale
and over time. A local-scale validation is advisable. Further studies aimed at developing a
temporal/spatial downscaling of the GPM data could help to provide more accurate and reliable
outputs. Moreover, one of the major problems with this kind of application is the error that may
be present in the rainfall measurement. The system has no control over any underestimation,
overestimation or random errors. Furthermore, no remedy is in place with reference to any kind
of temporary interruptions in the provision of data.
With this type of application, the model calibration and performance evaluation
continue to be challenging problems in large data-scarce regions or in areas where only a few
hydrometeorological events were recorded. The currently adopted thresholds may be
influenced by this problem.
Acknowledgements
We express our gratitude to Paolo Pasquali, who helped us in the development of the
web app.
REFERENCES
1. Alfieri, L.; Cohen, S.; Galantowicz, J.; Schumann, G.J.P.; Trigg, M.A.; Zsoter, E.;
Prudhomme, C.; Kruczkiewicz, A.; Coughlan de Perez, E.; Flamig, Z.; et al. A Global Network
for Operational Flood Risk Reduction. Environ. Sci. Policy 2018, 84, 149158.
doi.org/10.1016/j.envsci.2018.03.014.
2. Deutscher Wetterdienst Archiv der Klimadaten Deutschland Stundenwerte.
Available online: https://rcc.dwd.de/DE/leistungen/klimadatendeutschland/
klarchivstunden.html%5Enn=561158 (accessed on 20 April 2019).
3. EM-DAT: The Emergency Events Database Université Catholique de Louvain
(UCL) CRED, D. Guha-Sapir. Brussels, Belgium. Available online: www.emdat.be
(accessed on 20 April 2019).
4. Floodlist. http://floodlist.com (accessed on 20 April 2019).
5. Huffman, G.J. GPM IMERG Early Precipitation L3 Half Hourly 0.1 Degree x 0.1
Degree V05; Goddard Earth Sciences Data and Information Services Center (GES DISC):
Greenbelt, MD, USA, 2015.
6. Huffman, G.J.; Bolvin, D.T.; Braithwaite, D.; Hsu, K.; Joyce, R.; Kidd, C.; Nelkin,
E.J.; Sorooshian, S.; Tan, J.; Xie, P. NASA Global PrecipitationMeasurement (GPM)
IntegratedMulti-satellitE Retrievals for GPM (IMERG) Algorithm Theoretical Basis Document
(ATBD) Version 5.2. 2018. Available online: https://pmm.nasa.gov/sites/default/files/
document_files/IMERG_ATBD_V5.2.pdf (accessed on 20 April 2019).
7. IMERG Land / Sea mask. Available online: https://pmm.nasa.gov/sites/default/files/
downloads/surfrac0.1.PPS.gz (accessed on 20 April 2019).
S1433
8. Mazzoglio, P.; Laio, F.; Balbo, S.; Boccardo, P.; Disabato, F. Improving an Extreme
Rainfall Detection System with GPM IMERG data. Remote Sens. 2019, 11, 677.
doi.org/10.3390/rs11060677.
9. National Centers for Environmental Prediction/National Weather Service/NOAA/
U.S. Department of Commerce, 2015: NCEP GFS 0.25 Degree Global Forecast Grids
Historical Archive. Research Data Archive at the National Center for Atmospheric Research,
Computational and Information Systems Laboratory, Boulder, CO. Available online:
https://doi.org/10.5065/D65D8PWK (accessed on 20 April 2019).
10. NOAA Local Climatological Data. Available online: www.ncdc.noaa.gov/cdo-
web/datatools/lcd (accessed on 20 April 2019).
11. Nurmi, P. Recommendations on the Verification of Local Weather Forecasts.
ECMWF Technical Memorandum n° 430. 2003. Available online: www.ecmwf.int/en/elibrary/
11401-recommendations-verification-local-weather-forecasts (accessed on 20 April 2019).
12. Reliefweb. https://reliefweb.int (accessed on 20 April 2019).
13. Schneider, U.; Becker, A.; Finger, P.; Meyer-Christoffer, A.; Rudolf, B.; Ziese, M.
GPCC Monitoring Product: Near Real-Time Monthly Land-Surface Precipitation from Rain-
Gauges based on SYNOP and CLIMAT data. 2011, doi:10.5676/DWD_GPCC/
MP_M_V4_100.
Chapter
The Extreme Rainfall Detection System (ERDS) is an early warning system (EWS) developed for the monitoring and forecasting of rainfall events on a global scale. Within ERDS the near real-time rainfall monitoring is performed using the Global Precipitation Measurement data, while rainfall forecasts are provided by the Global Forecast System model. Rainfall depths determined on the basis of these data are then compared with a set of rainfall thresholds to evaluate the presence of heavy rainfall events: in places where the rainfall depth is higher than a rainfall threshold, an alert of a severe rainfall event is issued. The information provided by ERDS is accessible through a WebGIS application (http://erds.ithacaweb.org) in the form of maps of rainfall depths and related alerts to provide immediate and intuitive information also for nonspecialized users. This chapter is intended to describe the input data and the extreme rainfall detection methodology currently implemented in ERDS. Furthermore, several case studies (2019 Queensland flood event, 2017 Atlantic hurricane season, and 2017 Eastern Pacific hurricane season) are included to highlight the strengths and weaknesses of this EWS based on global-scale rainfall datasets.
A Global Network for Operational Flood Risk Reduction
  • L Alfieri
  • S Cohen
  • J Galantowicz
  • G J P Schumann
  • M A Trigg
  • E Zsoter
  • C Prudhomme
  • A Kruczkiewicz
  • E Coughlan De Perez
  • Z Flamig
Alfieri, L.; Cohen, S.; Galantowicz, J.; Schumann, G.J.P.; Trigg, M.A.; Zsoter, E.; Prudhomme, C.; Kruczkiewicz, A.; Coughlan de Perez, E.; Flamig, Z.; et al. A Global Network for Operational Flood Risk Reduction. Environ. Sci. Policy 2018, 84, 149-158. doi.org/10.1016/j.envsci.2018.03.014.
  • Degree
  • Goddard
Degree V05; Goddard Earth Sciences Data and Information Services Center (GES DISC): Greenbelt, MD, USA, 2015.
NASA Global PrecipitationMeasurement (GPM) IntegratedMulti-satellitE Retrievals for GPM (IMERG) Algorithm Theoretical Basis Document (ATBD) Version 5.2
  • G J Huffman
  • D T Bolvin
  • D Braithwaite
  • K Hsu
  • R Joyce
  • C Kidd
  • E J Nelkin
  • S Sorooshian
  • J Tan
  • P Xie
Huffman, G.J.; Bolvin, D.T.; Braithwaite, D.; Hsu, K.; Joyce, R.; Kidd, C.; Nelkin, E.J.; Sorooshian, S.; Tan, J.; Xie, P. NASA Global PrecipitationMeasurement (GPM) IntegratedMulti-satellitE Retrievals for GPM (IMERG) Algorithm Theoretical Basis Document (ATBD) Version 5.2. 2018. Available online: https://pmm.nasa.gov/sites/default/files/ document_files/IMERG_ATBD_V5.2.pdf (accessed on 20 April 2019).
National Centers for Environmental Prediction/National Weather Service/NOAA/ U.S. Department of Commerce, 2015: NCEP GFS 0.25 Degree Global Forecast Grids Historical Archive. Research Data Archive at the National Center for Atmospheric Research
  • P Mazzoglio
  • F Laio
  • S Balbo
  • P Boccardo
  • F Disabato
Mazzoglio, P.; Laio, F.; Balbo, S.; Boccardo, P.; Disabato, F. Improving an Extreme Rainfall Detection System with GPM IMERG data. Remote Sens. 2019, 11, 677. doi.org/10.3390/rs11060677. 9. National Centers for Environmental Prediction/National Weather Service/NOAA/ U.S. Department of Commerce, 2015: NCEP GFS 0.25 Degree Global Forecast Grids Historical Archive. Research Data Archive at the National Center for Atmospheric Research, Computational and Information Systems Laboratory, Boulder, CO. Available online: https://doi.org/10.5065/D65D8PWK (accessed on 20 April 2019).
Available online: www
  • Climatological Noaa Local
  • Data
NOAA Local Climatological Data. Available online: www.ncdc.noaa.gov/cdoweb/datatools/lcd (accessed on 20 April 2019).
GPCC Monitoring Product: Near Real-Time Monthly Land-Surface Precipitation from Rain-Gauges based on SYNOP and CLIMAT data
  • U Schneider
  • A Becker
  • P Finger
  • A Meyer-Christoffer
  • B Rudolf
  • M Ziese
Schneider, U.; Becker, A.; Finger, P.; Meyer-Christoffer, A.; Rudolf, B.; Ziese, M. GPCC Monitoring Product: Near Real-Time Monthly Land-Surface Precipitation from Rain-Gauges based on SYNOP and CLIMAT data. 2011, doi:10.5676/DWD_GPCC/ MP_M_V4_100.
GPM IMERG Early Precipitation L3 Half Hourly 0.1 Degree x 0
  • G J Huffman
Huffman, G.J. GPM IMERG Early Precipitation L3 Half Hourly 0.1 Degree x 0.1