GFS_new_usesJA4.doc - EMC - NOAA

Document Sample
GFS_new_usesJA4.doc - EMC - NOAA Powered By Docstoc
					    Applications of Agricultural Importance from Reliable Real Time Access to Operational
                                   Environmental Model Data

                                              Jordan C. Alpert

                                National Centers for Environmental Prediction
                                      Environmental Modeling Center
                                          Camp Springs, MD 20736

        Our goal is to increase smallholder productivity and incomes by reducing uncertainty for
weather variables of agricultural importance. There are many digital steps and human
connections to accomplish this mission, and they are summarized in the accompanied BTO (New
Uses for Global Weather Forecasts Study) memo. In this annex to the BTO2 we describe the
technical aspects of how the data is created and transported through digital technology from
centralized national meteorological centers to appear at the local level as a useful product to
mitigate the impact of natural hazards and environmental changes. This process begins with
forecasts from the US weather service operational global forecast system (GFS) model ensemble
(GENS). New resources, including operational delivery of the model output data to and from
high availability weather service content servers offers a reliable web access framework for
content serving NOAA’s real time data base, and aggregating the GENS forecast data set to
produce user selected weather element event probabilities. The system is controlled by
knowledge user input to access tools for rapid review of the probability of upcoming weather
events, and a tool to drive subsequent alerts to cell phones automatically. The short message
service alerts based on probability are threshold adjusted by local user sensitivity to false alarms
offering an optimal forecast for a user defined weather event, often when no other transmission
method is available.

Introduction and Motivation

       To reduce the impact of natural hazards and environmental changes, the National Centers
for Environmental Prediction (NCEP), under the National Oceanic and Atmospheric
Administration (NOAA), and the National Weather Service (NWS) provide watches and
warnings to protect life and property. Virtually all the meteorological data collected over the
globe arrives at NCEP, where environmental scientists analyze this information and generate a
wide variety of environmental guidance information. NCEP delivers national and global
weather, water, climate guidance, forecasts, warnings and analyses to a broad range of users and
partners. These products and services respond to user needs to protect life and property, and
improve economy including agriculture activity, and support the growing need for environmental

 NCEP, Environmental Modeling Center, Camp Springs, MD 20746;
  In accordance with the Terms of Reference, a World Bank team visited Turkey December 2-8, 2009, under the
Grant described in the Back-to-Office Report: Turkey - Innovation Grant: "New Uses for Global Forecasts" to
review Turkish practice and innovations in the use of weather alerts in agriculture.
        At the same time the challenges and opportunities of agriculture and rural development
can benefit from timely information and responses. For example, to increase smallholder
productivity, incomes, and to help farmers manage a range of risks, we seek to improve the
access, affordability, appliance innovation and applications to empower national and regional
capacities as pointed out in World Bank Working Paper No. 151 (2008), “Weather and Climate
Services in Europe and Central Asia, A Regional Review”. Information and communications
technologies (ICT), in particular, high-speed internet connections, content servers, and mobile
application services are transforming public service delivery of critical information and
democratizing innovation (McNamara, K. S., “Workshop on Mobile Innovations for Social and
Economic Transformation”, World Bank, September 16, 2009).

        Mobile platforms (cell phones) are emerging as the single most powerful way to extend
ICT opportunities and key services to millions of people. Poor access to information and
communication is an intangible dimension of rural poverty. Our goal is to increase smallholder
productivity and incomes by reducing uncertainty for variables of interest of agricultural
importance. In this case we use forecasts from weather service operational Global Forecast
System (GFS) global model ensemble (GENS) to alert agriculture users if and when their
selected weather events will occur. A description of how this system framework operates and
can be implemented using existing US weather service high availability servers and applications
is described.

Description of NOMADS, the digital content server, and NOAA’s commitment to maintain high
availability services

        To meet the future needs of the ever-broadening user community and address the
strategic climate-water-weather issues on a national and international basis, the weather service
implemented (2009FEB) the NOAA Operational Model Archive and Distribution System
(NOMADS). NOMADS is designed to provide real-time and retrospective format independent
access to climate, ocean and weather model data. NOMADS is now delivering high availability
services as part of NOAA’s official real time data dissemination at its Web Operations Center
(WOC) server. The WOC is a web service used by organizational units in and outside NOAA,
and acts as a data repository where model and other information is posted to a secure and
scalable content server. NCEP Central Operations (NCO) has committed that data be present
and on time from their Operational super computers. NOAA WOC has committed to maintain
NOMADS servers now and into the future, as well as day to day operations with costs shared by
NOAA and NCEP base. A development commitment continues at NCEP Environmental
Modeling Center (EMC), a development division, to keep up with new data sets and create
applications. Data review groups, official committees and procedures for moving new data sets
and applications from development to operations follows the existing NCO framework for
implementations. A goal is to foster collaborations among the research and education
communities, value added content makers, and public access for science and development efforts
aimed at advancing modeling and GEO-related tasks. A case in point is this project which has
benefited from cooperation between personnel in the United States weather service and from the
World Bank resulting in presentations at a major scientific conference, and is the recipient of a
World Bank Innovation Grant, “New Uses for Global Forecasts”, awarded in 2009. Results of
the first phase of the project are reported in the accompanied BTO where it is shown how trust
and relationships at the local level have resulted in the planning of a prototype of the technology
described here in.

The Global Forecast System (GFS) and Ensemble model Forecasts

         The GFS is an operational global numerical weather prediction (NWP) model and
gridded statistical interpolation (GSI) that assimilates (~billion) conventional and non-
conventional observations each day to create an optimal initial condition and weather forecasts
from the short range through the medium range (2 weeks and more). It has a resolution of
triangular truncation T382, or about 1/3 degree over the globe, and 64 vertical levels, and is run
operationally 4 times a day by NWS. In addition, the GFS model is used to run 20 integrations,
but at a lower resolution (T126, 1x1 degree), called ensembles (GENS). Each GENS model run
produces global coverage forecasts for several hundred variables at 6-hour intervals. The
ensembles are constructed by changing the initial conditions to make a “cloud” of forecasts that
attempt to span the space of possible atmospheric realizations. Ensemble members are equally
probable forecasts – if not, we apply a suitable normalization. Probability estimates are defined
simply as the percentage of forecasts of the total sample space that satisfy a specified weather
event. We include the GFS as a control run and apply this to weather elements like surface
temperature, wind speed or precipitation at a location and future model forecast time. (See for
example Alpert and Wang, 2005, AMS 21 IIPS 17.5). Ensemble event probabilities are
calculated from values in the matrix of the binary packed real time model output which is very
large and contains many variables (Tb/day). The GENS output contains hundreds of variables
such as temperature, precipitation, wind, soil moisture and others, for each of the 20 components,
for all the 65 6-hour forecast interval times (or files, out to 2 weeks), for 26 vertical (pressure)
levels if present, on gridded global 1x1 degree latitude and longitude records with man or
machine readable metadata descriptions. These are aggregated in any of the dimensions by the
WOC content server to make the data available to users around the world from NOMADS
applications using a web URL query so users can get the values they need.

Server Applications

        Metadata descriptions uniquely describe and aggregate each file on the server and are
man or machine readable enabling collections of data in disparate (virtual and physical) places to
be simultaneously accessed with the results processed to produce a needed answer. The services
used to access the operational model data output are the Open-source Project for a Network Data
Access Protocol (OPeNDAP), implemented with the Grid Analysis and Display System
(GrADS) Data Server (GDS). This approach insures an efficient use of computer resources
because users transmit/receive only the data necessary for their tasks including metadata. In this
way the paradigm lends itself to aggregation services that act as servers of servers listing,
searching catalogs of holdings, data mining, and updating information offering vast possibilities
for the creation of value added products.

       A number of applications support downloading of repackaged WMO standard GRIB files
used by national centers (Alpert and Wang, 2007) such as NOMADS applications http or
“fast/partial ftp” access with inventory and client script, “ftp2/4u” (“GRIB filter”), to slice, dice
and area subset files delivering repackaged GRIB files and plain ftp services, but these will not
be discussed further here as we are interested in the delivery of values directly to users by web
issued queries. The GDS/OPeNDAP(DODS) application is most useful for obtaining such
values from http queries. We will also confine our discussion to the GENS data set but other
model data sets are equally applicable. NOMADS services are used to select the values of
Ensemble model run output over the ith Ensemble component, (forecast) time, vertical levels,
global horizontal location, and by variable, virtually a 6-Dimensional data cube of access across
the internet.

         An application called the “ensemble probability tool”
( and follow links to ensemble probability tool or creates a graphic display of probability
predictions of user defined weather events that is used to hone the probability threshold value for
the users chosen event. This is an optional step since a user may already be aware of the
threshold they desire, and such information from a knowledgeable users experience can be used
to set the proper false alarm rate. Results from a verification pilot to test the accuracy and
usefulness of forecast weather events is shown in Hancock et al., (2009). The report shows an
example of high value usability and relevance of NCEP products and service capability over a
wide spectrum of user and partner needs and can supplement the local observation of users. The
ensemble probability tool is a development program which calls on the NOMADS operational
server using a query similar to the one shown below to get the information it needs from a user
selected weather event request. A program used in the World Bank project “New Uses for
Global Weather Forecasts Study”, results in cell phone alerts of user selected weather events
when a selected weather event is forecast. Together these examples of client (user) side tools
can be used to optimally alert rural agriculture players to their selected weather events on cell

NOMADS and the GrADS-Data Server (GDS) OPeNDAP: How does it work?

        In summary, NOMADS participants including the NWS serve their data sets through a
client-server relationship. The data sets have machine and man readable metadata descriptions
and aggregate across any number of separate binary packed files providing the unique ability to
transpose the data across all dimensions. Display is done by the client (user). GDS combines
GrADS, a freeware client, and an OPeNDAP server to unpack, cache and exchange data from
many formats using HTML in response to user queries. This means that server data can appear
to the user or client application as a local file! OPeNDAP requests are understood (created) by
many freeware and commercial high level language clients like GrADS and MATLAB. Web
(http) URL queries to the OPeNDAP server can create value added products in addition to
scientific work and for public and federal agency needs.

       A Constrained Query

       The constrained query makes requests on the server for data and is the engine for
obtaining the data from which weather event decisions are made. Normally, computer programs
and programmers create the text of these queries and send them to the server and are hidden from
a users view, but anyone can composed and type them into any internet browser. A server query
is composed of the server address, data directory location, a command and separator characters
(.command?; the command is usually either “.info?” or “.ascii?”), variable name, and constraint
as follows, for example, for the GENS data set from the date, 20090501 and 00 Universal time
(Z) model cycle:[0:19][0:2

Notice the 5-Dimensional query for each variable (6-D data cube!), e.g, temperature (tmpprs):
ith Ensemble run            [0:19] Delivers all 20 Ensemble components at,
Forecast times              [0:21] IC and every 6-hour interval forecast to 5-days is indicated,
Vertical levels (if any)    [1:1] 975Mb indicated, and [0:0] would mean 1000mb if present,
Latitude,                   [129:129] is measured from SP (0) to NP (for a 1 degree grid)
Longitude,                  [243:243] beginning at the 0 meridian – we show Baltimore BWI

The general ordering of these square bracketed values is:


Units and other information are represented in the metadata description file. A stride is also
possible [start:stride:finish] using colon separated values. Variables like the tmpprs are
snapshots of temperature at the model forecast time, but variables like tmin and tmax, minimum
and maximum temperature respectively, as well as precipitation, are accumulated over the
preceding 6 hour interval. Normally queries such as the one above are automatically composed
by web based programs but they can be typed into any internet browser to get the result of the
requested information. For example, typing (copy-paste-ing) the above example into any
internet browser and updating the date will result in a 20x22 number set of all 20 ensemble
forecasts at all forecast times through 5 days ( 5days on 6-hour intervals) at Baltimore (BWI) intl
airport. In addition the server result contains the location in time and space of the returned
values. The query is the way data values can be extracted from the model output. In addition we
may construct web pages with appropriate prompts for a user to enter information, such as, what
weather event should trigger an alert. The same program could make corrections to forecasts
based on local effects, and send an alert to a cell phone if certain criteria are satisfied. The
worldbank3 program is an example of this and the ens_probaility program allows users to
display forecast probabilities.

        Another example of a query that can be printed in this document (the above example
returns too many values to show) but similar to the one above is from the GENS data set:[0:1][0:4

where we have changed the location to near Bishkek (43.,75.), requested the 2 meter maximum
temperature in the last 6 hours (tmax2m) and requested the initial condition on the date
2009122400, and the 06, 12, 18 and 24 hour forecasts [0:4] for the GENS control and first
component [0:1]. Note that the 2 meter (height of standard weather station thermometers)
maximum temperature has no vertical structure so there are 4 square bracketed values in this
constraint. Typing this into any browser sends this query to the NOMADS high availability real
time server (the date should be updated) and returns the following result:

tmax2m, [2][5][1][1]
[0][0][0], 9.999E20

[0][1][0], 273.12

[0][2][0], 273.83

[0][3][0], 270.62

[0][4][0], 268.50998

[1][0][0], 9.999E20

[1][1][0], 273.03

[1][2][0], 273.83

[1][3][0], 270.72

[1][4][0], 268.22

ens, [2]
1.0, 2.0
time, [5]
733766.0, 733766.25, 733766.5, 733766.75, 733767.0
lat, [1]
lon, [1]
The first line of values returned indicates the array structure of
temperatures. The returned temperatures in this section are for the control
run ensembles beginning on the next line preceded by an indicator of which
component of the array is presented, the initial condition (f00) at the
single point, Bishkek, for the control [0] ensemble component is first.
However, the value presented is the missing value indicator, 9.999E20, since
the maximum temperature is not a dependent variable and is not calculated for
the initial condition. The variable, tmax2m, is the highest temperature over
the previous 6 hours. If we chose the temperature variable, “tmpprs”, as
described above, there would be a “snap shot” temperature result which could
be used, for example, to verify past predictions. The forecasted maximum
temperatures for this model run time are next: [0][1][0], 273.12, [0][2][0],
273.83, [0][3][0], 270.62, [0][4][0], 268.50998, presenting the maximum
temperature for 6, 12, 18 and 24 hour forecasts, the values in Kelvin
preceded by the bracketed indicators showing which ensemble component ([0]),
the forecast time ([1-4]) and the (zeroth or first and only [0]) latitude are
present in this result. The next 5 lines repeat this for the first ensemble
component forecast. The lines “ens, [2]/1.0, 2.0” indicate that we have 2
ensemble components present which are the control and first forecast
realizations. The forecast times are given as the number of days since the
year 1 (counting form zero), or there are 733766.0 days since then for
2009122400 (December 24, 2009) and each of the forecast at ¼ day apart is
shown. The latitude and longitude of the model point is returned which is
close to the Bishkek airport (42.85,74.53). The names of variables, units
and other information are contained in the man or machine readable metadata
descriptions of each data set is located at the same location as the data but
with the suffix “.info” (or no suffix, the default). Normally user (clients)
programs compose and send queries but a user can write and send their own
queries in any browser or from programs, for example, to facilitate
verification studies which is how verifications were made in the Hancock, et
al, (2009) study.

Describe the ens_probability Script (

         A web based client (user) application that can be used to provide the probability of a
weather event and threshold information to the user is the EnsProb web page. The program
obtains values from the Global Ensemble Forecast information matrix, as specified by the user,
and returns the information to the user as a probability display to allow the user to determine a
proper threshold of their user defined weather event. This program script begins by opening a
web page so the user can choose his location from a station list, or enters latitude and longitude
and define the weather event to the program. At present the program offers weather events
concerning high or low temperature, precipitation or wind speed. Other variables such as snow
amount, soil moisture, or categorical forecasts can be added. For example, a user can define
frost (not necessarily 0C) as the minimum temperature (tmin2m) below a critical value. Behind
the scenes, the program script constructs text queries, like the one shown above, sends the
requests to the server, and parses the results returned from the server. The results from the server
contain the forecasts for all the ensembles every 6 hours out to 7 days. These are organized to
make a graphical display of the requested forecast probabilities as each returned value is tested
against the users requested weather event. The hits are recorded and divided by the total number
of ensemble components to get the probabilities. The results are displayed on a bar graph with
each 6 hour interval represented on the abscissa out to 7 weeks and probability in percent shown
on the ordinate. A user can examine the bar graph, as shown in Fig 1 below, and determine how
often they desire to be notified. By raising the probability threshold to 75%, as shown by the red
line in the figure, only a few events would trigger the event. If notification is required for
smaller chances of the event, then lower values can be selected. Over a period of time, or
checking with archived data which is also available, one can hone in the best value for a
particular weather event. A knowledgeable user would do this but in many cases the threshold is
known by locals through their own experience.

       The information that users entered into the program web pages is also updated with
keywords and values and returned in the URL address at the top of the web page window when
the program is successful. This address can be resent in any browser and the results would be
repeated but one can alter the keyword values so a different date, variable or other information
can be modified. One does not need to rerun the program and enter values into web pages as
many options are available through directly altering the keyword values in the given URL. The
URL is sent by means of a non-interactive web download program like “wget” or cURL.

                        1) Ensemble probability tool: A web based client (user)
                        application that can be used to provide threshold information to
                        the user. The program obtains the Global Ensemble Forecast
                        information matrix from the server, and returns the information to
                        the user as a display (below) to allow the user to determine a
                        proper threshold of their user defined weather event.
 3) The application
   delivers ensprob
   program “code”           User Selected Threshold…
   in the form of the                                                  2) User determines their
   returned URL            as per their tolerance for false alarms
                                                                        threshold (to tolerate
   address, like the                                                    false alarms) for an
   one above, for                                                       alert, as the Worldbank
   the user to                                                          application sends an
   repeat the                                                           alert to email and cell
   action by copy-                                                      phone text-message
   pasting it into                                                      when the threshold is
   any browser or...                                                    met.

4) The user can re-issue (send) the “check for an alert URL” automatically from a
 scheduler like cron using a non-interactive web download command like “wget” with
 the returned URL from a successful session.
 Fig 1

How to get the information to a cell phone: Description of the Program

        To put together a user created weather event and threshold to set an alert to a cell phone,
we have created a web based client application program were knowledge users enter responses to
define the probability of a weather event that they choose, similar to ensemble probability tool
described above. The program, “Worldbank application” (, alerts the
user (via email and/or cell phone text message) if and when the event will occur in the future
within the next 7 days from when issued. The program prompts the user to adjust or customize
their threshold value to tolerate (the level of) false alarms for their mobile alert or an optimal
value can be obtained from past experience. The default is given at 0.5, meaning half of the
ensemble runs need to meet the users weather event criteria. An example of the web page
interface is shown below in Fig 2.


                                           Ensprob World Bank Application

                                                    Like the Ensprob tool, the user
                                                 chooses the location from the
                                                 station list, or enters Lat/Lon and
                                                 defines the weather event.

                     A user selects the threshold of probability by their
                 experience of false alarms for an alert, and the application sends
                 an alert to email and/or cell phone text-message using providers

  Fig 2


A user enters the location from the station list. The script allows maximum or minimum
temperature, precipitation accumulation, or wind speed but other variables could be programmed
to define a weather event. An email address, if entered, will keep a record of the session and the
email address of the cell phone provider is also added. For example, a Verizon cell phone entry
is (Other wireless provider SMS email addresses can be found at will send the alert
message to the cell phone number specified. The threshold value indicating what criteria is used
to satisfy the given weather event can also be entered.

        The horizontal resolution of the resulting probability calculation is the same as that
defined by the GENS forecasts, 1x1 degree which is too course for direct application to a
particular farm. This is mitigated a number of ways. To some extent having a probabilistic
forecast and threshold selected from local experience mitigates local differences. In addition, the
application program can be improved to take account of local information, especially bias due to
the locations on or near mountains or water bodies. The accuracy of global ensemble probability
forecasts have been studied for global and time averages but verification has not been shown for
particular locations (see Hancock et al., 2009). This study shows some locations that are verified
and reports on the accuracy and delivery of the forecast results in a real case, as well as the
feasibility of implementing this mobile application to promote agricultural innovation and value
in a given context.

         As an example of using the application to make an alert, one can answer the prompts
presented by the application as shown in Fig 2, and the user “Make an Alert of a Weather Event
that You Define”. In this case we have defined our weather event as the maximum temperature
in a 6 hour period over the next week being greater than 2ºC or 275ºK. This was done on
29DEC2009. The application makes a single query to the NWS official NOMADS server to
obtain the maximum surface temperature (tmax2m) over each 6-hour period for all 20 ensemble
runs and 29 forecast times (20X29) numbers which are parsed and organized by the application.
This is done at the single point close to Bishkek, KZ airport (Lat: 42.85 N, Lon:74.53 E
rounded to nearest degree). The threshold was set at 75% meaning that for an alert to take place
the weather event (6-hour high temperature greater than 2ºC) has to occur with a threshold of at
least .75 or that ¾ of the ensemble runs are forecasting 6-hour maximum temperature greater
than 2ºC. This setting was designed to produce at least a single result over the next week as an
illustration of the application. Application diagnostics show that there is no or little chance for
this to occur until the 6 hours preceding January 3 00Z when there is 100% agreement of the
ensemble components as shown below in Fig 3.

Diagnostic Printout From the Application
Event probability: 0% , fcsttime is 2009 dec 29 18Z
Event probability: 0% , fcsttime is 2009 dec 30 00Z
Event probability: 0% , fcsttime is 2009 dec 30 06Z
Event probability: 0% , fcsttime is 2009 dec 30 12Z
Event probability: 0% , fcsttime is 2009 dec 30 18Z
Event probability: 0% , fcsttime is 2009 dec 31 00Z
Event probability: 0% , fcsttime is 2009 dec 31 06Z
Event probability: 0% , fcsttime is 2009 dec 31 12Z
Event probability: 0% , fcsttime is 2009 dec 31 18Z
Event probability: 0% , fcsttime is 2010 jan 01 00Z
Event probability: 0% , fcsttime is 2010 jan 01 06Z
Event probability: 0% , fcsttime is 2010 jan 01 12Z
Event probability: 0% , fcsttime is 2010 jan 01 18Z
Event probability: 30% , fcsttime is 2010 jan 02 00Z
Event probability: 0% , fcsttime is 2010 jan 02 06Z
Event probability: 0% , fcsttime is 2010 jan 02 12Z
Event probability: 30% , fcsttime is 2010 jan 02 18Z
Event probability: 100% , fcsttime is 2010 jan 03 00Z
Event probability: 5% , fcsttime is 2010 jan 03 06Z
Event probability: 0% , fcsttime is 2010 jan 03 12Z
Event probability: 55% , fcsttime is 2010 jan 03 18Z
Event probability: 90% , fcsttime is 2010 jan 04 00Z
Event probability: 0% , fcsttime is 2010 jan 04 06Z
Event probability: 0% , fcsttime is 2010 jan 04 12Z
Event probability: 90% , fcsttime is 2010 jan 04 18Z
Event probability: 100% , fcsttime is 2010 jan 05 00Z
Event probability: 30% , fcsttime is 2010 jan 05 06Z
Event probability: 0% , fcsttime is 2010 jan 05 12Z
Fig 3
       The email that was sent for this somewhat contrived (so it would give a response)
weather event is shown in Fig 5 below:

Original Message --------
Subject:      ALERT WEATHER EVENT: Temperature, highest TEMP: gt 2 (275.1 K)
Date: Tue, 29 Dec 2009 17:03:30 -0500
From: Apache <>

 Temperature, highest TEMP: gt 2 (275.1 K) > 75%, chance @ ft=
 2010 jan 03 00Z
108 hr fcst
2010 jan 04 00Z
132 hr fcst
2010 jan 04 18Z
150 hr fcst
2010 jan 05 00Z
156 hr fcst

Fig 5
Note that all the forecast times that exceeded 75% are shown. The SMS cell
phone messages are restricted to 150 characters so only the first one or two
forecast times would be shown. We can attempt to use the GFS data to check
on the verification of this event and it is instructive to show how other
models can be queried to get data for more detailed and useful applications.
That is, the temperature at a past forecast time can be found from the global
GFS analysis of 03JAN2010 00Z which verifies the 108 hour forecast (4 ½ day)
made from the initial time 29DEC2009 12Z. The value of the analysis at
(near) the Bishkek airport can be obtained with a query to the deterministic
GFS model (on a half degree lat/lon grid), since at the time of this writing
it has already happened, as[0:0][

where the query has included the closest 4 points to the Bishkek airport. The result from this
query for the surface (tmp2m) temperature is:

tmp2m, [1][2][2]
[0][0], 277.94998, 276.47
[0][1], 277.07, 275.47

time, [1]
lat, [2]
43.0, 43.5
lon, [2]
74.5, 75.0
Fig 4

The latitude and longitude of the nearest 4 points are shown and the values
at each of the four points 277.94998, 276.47 / 277.07, 275.47 ºK do satisfy
the chosen weather event of greater than 275K predicted by the GENS
probability forecast. One would have to repeat this process numerous times
to have statistical significance. The time represented here is 733776 or
03JAN2012 00Z. The application is presently set to deliver a number of
diagnostic print output to the knowledgeable user as shown in Fig 3 but this
can be turned off. The application will write an
augmented URL on the final web page which contains all the users input from
the successful session and may be re-submitted at any time to rerun the
application and re-check for the weather event for the next or future model
forecast cycle.   The URL produced by the application was copy-pasted from
the web page and is shown below:

Note that there is no date request as the application will take the latest
available model cycle run and use that data to test for the weather event of the next 7 days.
Therefore, the URL, so modified, can be placed in a cron scheduler by using a cgi-bin script to
provide automatically repeated user interaction with the program, that is, to check for the users
weather event and alert the user if it will occur in the next 7 days. The time granularity, thus, is a
maximum of 4 times per day since the GENS model and output is updated at that rate. The URL
is sent by means of a non-interactive web download (open source) program like “wget” or

Summary and Ways Forward

The technical aspects described above and used with the programs were tested successfully in a
proof of concept mode or prototype. That is, a common cell phone was used to transmit example
alerts at all the locations visited by the project in phase. In addition to the program scripts,
EnsProd and programs an addition script to show any deterministic
variable in the model output (listed in the metadata description) as available for user selection.
In addition a display of a Global meteogram script depicting variable values versus forecast time
in addition to the probability displays. Finally, we point out that a knowledgeable user can set up
their own computer and install their own server and script. The software is open source and the
source code, for example, of can be obtained at

References: Hancock et al., (2009) AGU Fall meeting poster U21A-0007 and paper IN11E-05)
using Global Forecast Ensemble (GENS) as a tool for alerting users through mobile platforms of
weather events

Shared By: