Loading ...
Sorry, an error occurred while loading the content.

[ai-geostats] Optimal rain gauge network

Expand Messages
  • Craig von Hagen
    Hi All, I have an interesting problem to solve, I hope someone could help me. We are working on flood early warning in Somalia and we have the following
    Message 1 of 6 , Jan 3, 2006
    • 0 Attachment
      Hi All,

      I have an interesting problem to solve, I hope
      someone could help me.

      We are working on flood early warning in Somalia
      and we have the following situation.

      We have an existing network of manual rain gauges
      that we receive on a monthly basis with daily
      readings taken manually by a person in the field.
      These however can be unreliable.

      We have an option to install automatic rain gauges
      that would give us an accurate measurement of
      rainfall per day. We would like to use
      geo-statistics to then give a prediction and error
      surface and then use these surfaces to evaluate how
      accurate and reliable our existing manual network
      is.

      Is there a way to calculate the optimal network
      (number and location) for the automatic stations so
      that we get a reliable prediction surface which we
      can then use to evaluate our manual network?

      I am the most familiar with the ArcGIS GeoStatistical
      Analyst.

      Thanks and regards
      Craig


      Craig von Hagen
      FAO - GLCN/Africover/SWALIM Projects
      PO Box 30470-00100
      Nairobi, Kenya

      Tel: +254 20 444 3331
      Fax: +254 20 444 1993

      www.africover.org
      www.glcn.org; www.glcn-lccs.org
      www.faoswalim.org



      ___________________________________________________________
      To help you stay safe and secure online, we've developed the all new Yahoo! Security Centre. http://uk.security.yahoo.com
    • Rühaak, Wolfram
      Hi Craig, ... I am quite sure that there are methods to estimate an optimal station-setup (probably colleagues with more experience will answer), but I want to
      Message 2 of 6 , Jan 3, 2006
      • 0 Attachment
        Hi Craig,

        > -----Original Message-----
        > From: Craig von Hagen [mailto:craigvonhagen@...]
        > Hi All,
        >
        > I have an interesting problem to solve, I hope
        > someone could help me.
        >
        > We are working on flood early warning in Somalia
        > and we have the following situation.
        >
        > We have an existing network of manual rain gauges
        > that we receive on a monthly basis with daily
        > readings taken manually by a person in the field.
        > These however can be unreliable.
        >
        > We have an option to install automatic rain gauges
        > that would give us an accurate measurement of
        > rainfall per day. We would like to use
        > geo-statistics to then give a prediction and error
        > surface and then use these surfaces to evaluate how
        > accurate and reliable our existing manual network
        > is.
        >
        > Is there a way to calculate the optimal network
        > (number and location) for the automatic stations so
        > that we get a reliable prediction surface which we
        > can then use to evaluate our manual network?
        > [...]
        > Thanks and regards
        > Craig

        I am quite sure that there are methods to estimate an optimal station-setup (probably colleagues with more experience will answer), but I want to point out an other thing:
        rainfall is mostly not uniform distributed. The topography has a major impact, also differences of the land cover can result in differences of the (micro - meso scale) climate. There are probably also coastal effects.

        I think, that it would be necessary to take this also into account. Which makes the calculation of number and place of the new stations more difficult.

        Hope this helps a little bit, cheers Wolfram
      • Chris Hlavka
        Craig - I m not an expert with geostatistics and related software, but here are a couple of general principals to consider: 1) In statistical experimental
        Message 3 of 6 , Jan 3, 2006
        • 0 Attachment
          Craig - I'm not an expert with geostatistics and related software,
          but here are a couple of general principals to consider:

          1) In statistical experimental design design, the greatest
          improvement in estimates are achieved by adding points at the
          locations with the most uncertainty.

          2) Define your optimization carefully. Instead of minimization a
          simple sum of squares, weight locations in terms of their
          contribution to flooding. A hydrological model can be helpful in
          defining this - using soil type maps and local slopes computed from
          digital elevation data.

          Chris

          >Hi All,
          >
          >I have an interesting problem to solve, I hope
          >someone could help me.
          >
          >We are working on flood early warning in Somalia
          >and we have the following situation.
          >
          >We have an existing network of manual rain gauges
          >that we receive on a monthly basis with daily
          >readings taken manually by a person in the field.
          >These however can be unreliable.
          >
          >We have an option to install automatic rain gauges
          >that would give us an accurate measurement of
          >rainfall per day. We would like to use
          >geo-statistics to then give a prediction and error
          >surface and then use these surfaces to evaluate how
          >accurate and reliable our existing manual network
          >is.
          >
          >Is there a way to calculate the optimal network
          >(number and location) for the automatic stations so
          >that we get a reliable prediction surface which we
          >can then use to evaluate our manual network?
          >
          >I am the most familiar with the ArcGIS GeoStatistical
          >Analyst.
          >
          >Thanks and regards
          >Craig
          >
          >
          >Craig von Hagen
          >FAO - GLCN/Africover/SWALIM Projects
          >PO Box 30470-00100
          >Nairobi, Kenya
          >
          >Tel: +254 20 444 3331
          >Fax: +254 20 444 1993
          >
          >www.africover.org
          >www.glcn.org; www.glcn-lccs.org
          >www.faoswalim.org
          >
          >
          >
          >___________________________________________________________
          >To help you stay safe and secure online, we've developed the all new
          >Yahoo! Security Centre. http://uk.security.yahoo.com
          >
          >
          >* By using the ai-geostats mailing list you agree to follow its rules
          >( see http://www.ai-geostats.org/help_ai-geostats.htm )
          >
          >* To unsubscribe to ai-geostats, send the following in the subject
          >or in the body (plain text format) of an email message to
          >sympa@...
          >
          >Signoff ai-geostats


          --
          ***************************************
          Chris Hlavka
          NASA/Ames Research Center 242-4
          Moffett Field, CA 94035-1000
          (650)604-3328 FAX 604-4680
          Christine.A.Hlavka@...
          ***************************************
        • Gerald van den Boogaart
          Dear Craig von Hagen, If you install a network of automated gauges, which is dense enough to make a map to check the manual gauges, sure enough it makes the
          Message 4 of 6 , Jan 4, 2006
          • 0 Attachment
            Dear Craig von Hagen,

            If you install a network of automated gauges, which is dense enough to make a
            map to check the manual gauges, sure enough it makes the manual network
            superfluous.

            And such network would not really help evaluating the operators:
            * Anyway if the measurement of a manual gauge is underestimated systematically
            by the interpolated surface I would first suspect a difference in
            microclimate to be the origin of that and not a bad operator.

            * If the variation of the measurement is underestimated by the kriging error I
            would suspect an additional measurment error in the manual measurements or a
            ill specified variogramm before blameing the operator.

            If you would like to check a manual network you might consider two options:

            1) Install mobile automatic gauges next to manual ones and check by simple
            comparison (however you need to consider wether the operator should be
            allowed to know that he is controlled currently). Than move the mobile gauge
            to the next station.

            2) To check single operators you can run cross-validation with the existing
            manual network: estimate a prediction and a kriging error from all stations
            but the one to be checked and compare. You need to add the nugget effect of
            the semivariogram to the kriging error to get the variance of
            Prediction-Measurement.

            Best regards,
            Gerald v.d. Boogaart




            Am Dienstag, 3. Januar 2006 14:53 schrieb Craig von Hagen:
            > Hi All,
            >
            > I have an interesting problem to solve, I hope
            > someone could help me.
            >
            > We are working on flood early warning in Somalia
            > and we have the following situation.
            >
            > We have an existing network of manual rain gauges
            > that we receive on a monthly basis with daily
            > readings taken manually by a person in the field.
            > These however can be unreliable.
            >
            > We have an option to install automatic rain gauges
            > that would give us an accurate measurement of
            > rainfall per day. We would like to use
            > geo-statistics to then give a prediction and error
            > surface and then use these surfaces to evaluate how
            > accurate and reliable our existing manual network
            > is.
            >
            > Is there a way to calculate the optimal network
            > (number and location) for the automatic stations so
            > that we get a reliable prediction surface which we
            > can then use to evaluate our manual network?
            >
            > I am the most familiar with the ArcGIS GeoStatistical
            > Analyst.
            >
            > Thanks and regards
            > Craig
            >
            >
            > Craig von Hagen
            > FAO - GLCN/Africover/SWALIM Projects
            > PO Box 30470-00100
            > Nairobi, Kenya
            >
            > Tel: +254 20 444 3331
            > Fax: +254 20 444 1993
            >
            > www.africover.org
            > www.glcn.org; www.glcn-lccs.org
            > www.faoswalim.org
            >
            >
            >
            > ___________________________________________________________
            > To help you stay safe and secure online, we've developed the all new Yahoo!
            > Security Centre. http://uk.security.yahoo.com

            --
            -------------------------------------------------
            Prof. Dr. K. Gerald v.d. Boogaart
            Professor als Juniorprofessor fuer Statistik
            http://www.math-inf.uni-greifswald.de/statistik/

            office: Franz-Mehring-Str. 48, 1.Etage rechts
            e-mail: Gerald.Boogaart@...
            phone: 00+49 (0)3834/86-4621
            fax: 00+49 (0)89-1488-293932 (Faxmail)
            fax: 00+49 (0)3834/86-4615 (Institut)

            paper-mail:
            Ernst-Moritz-Arndt-Universitaet Greifswald
            Institut f�r Mathematik und Informatik
            Jahnstr. 15a
            17487 Greifswald
            Germany
            --------------------------------------------------
          • Craig von Hagen
            Hi Gerald, Thanks very much. However maybe it was not clear in my first email. We are not just interested in checking the gauge readers but also in: 1.
            Message 5 of 6 , Jan 9, 2006
            • 0 Attachment
              Hi Gerald,

              Thanks very much. However maybe it was not clear in my
              first email. We are not just interested in checking
              the gauge readers but also in:

              1. Connecting Somalia to the Global Observing
              System
              2. Feeding into the GTS to allow for validation
              of RFE data (and corrections used in the RFE 2.0
              algorithm).
              3. Using the data collected to provide data on
              storm duration and intensity (at present we only get
              24 hour readings)
              4. Real time flood forecasting and drought
              monitoring.

              The use of the rain gauges for checking/validating
              manual readings is an added benefit.

              I am also looking into the WMO methods for
              establishing a gauge network.

              So far responses received from the list for optimal
              gauge location inlcude: stratification by elevation -
              assumption being rain increases with height (although
              not really the case here as we have rainfall on the
              coast).
              - wieghting locations in terms of their contribution
              to flooding
              - adding gauges at locations where their is most
              uncertainty
              - taking into account topogrpahy, slope and coastal
              effects.

              Further inputs are welcome...

              Regards
              Craig


              --- Gerald van den Boogaart
              <boogaart@...> wrote:

              > Dear Craig von Hagen,
              >
              > If you install a network of automated gauges, which
              > is dense enough to make a
              > map to check the manual gauges, sure enough it makes
              > the manual network
              > superfluous.
              >
              > And such network would not really help evaluating
              > the operators:
              > * Anyway if the measurement of a manual gauge is
              > underestimated systematically
              > by the interpolated surface I would first suspect a
              > difference in
              > microclimate to be the origin of that and not a bad
              > operator.
              >
              > * If the variation of the measurement is
              > underestimated by the kriging error I
              > would suspect an additional measurment error in the
              > manual measurements or a
              > ill specified variogramm before blameing the
              > operator.
              >
              > If you would like to check a manual network you
              > might consider two options:
              >
              > 1) Install mobile automatic gauges next to manual
              > ones and check by simple
              > comparison (however you need to consider wether the
              > operator should be
              > allowed to know that he is controlled currently).
              > Than move the mobile gauge
              > to the next station.
              >
              > 2) To check single operators you can run
              > cross-validation with the existing
              > manual network: estimate a prediction and a kriging
              > error from all stations
              > but the one to be checked and compare. You need to
              > add the nugget effect of
              > the semivariogram to the kriging error to get the
              > variance of
              > Prediction-Measurement.
              >
              > Best regards,
              > Gerald v.d. Boogaart
              >
              >
              >
              >
              > Am Dienstag, 3. Januar 2006 14:53 schrieb Craig von
              > Hagen:
              > > Hi All,
              > >
              > > I have an interesting problem to solve, I hope
              > > someone could help me.
              > >
              > > We are working on flood early warning in Somalia
              > > and we have the following situation.
              > >
              > > We have an existing network of manual rain gauges
              > > that we receive on a monthly basis with daily
              > > readings taken manually by a person in the field.
              > > These however can be unreliable.
              > >
              > > We have an option to install automatic rain gauges
              > > that would give us an accurate measurement of
              > > rainfall per day. We would like to use
              > > geo-statistics to then give a prediction and error
              > > surface and then use these surfaces to evaluate
              > how
              > > accurate and reliable our existing manual network
              > > is.
              > >
              > > Is there a way to calculate the optimal network
              > > (number and location) for the automatic stations
              > so
              > > that we get a reliable prediction surface which we
              > > can then use to evaluate our manual network?
              > >
              > > I am the most familiar with the ArcGIS
              > GeoStatistical
              > > Analyst.
              > >
              > > Thanks and regards
              > > Craig
              > >
              > >
              > > Craig von Hagen
              > > FAO - GLCN/Africover/SWALIM Projects
              > > PO Box 30470-00100
              > > Nairobi, Kenya
              > >
              > > Tel: +254 20 444 3331
              > > Fax: +254 20 444 1993
              > >
              > > www.africover.org
              > > www.glcn.org; www.glcn-lccs.org
              > > www.faoswalim.org
              > >
              > >
              > >
              > >
              >
              ___________________________________________________________
              > > To help you stay safe and secure online, we've
              > developed the all new Yahoo!
              > > Security Centre. http://uk.security.yahoo.com
              >
              > --
              > -------------------------------------------------
              > Prof. Dr. K. Gerald v.d. Boogaart
              > Professor als Juniorprofessor fuer Statistik
              > http://www.math-inf.uni-greifswald.de/statistik/
              >
              > office: Franz-Mehring-Str. 48, 1.Etage rechts
              > e-mail: Gerald.Boogaart@...
              > phone: 00+49 (0)3834/86-4621
              > fax: 00+49 (0)89-1488-293932 (Faxmail)
              > fax: 00+49 (0)3834/86-4615 (Institut)
              >
              > paper-mail:
              > Ernst-Moritz-Arndt-Universitaet Greifswald
              > Institut für Mathematik und Informatik
              > Jahnstr. 15a
              > 17487 Greifswald
              > Germany
              > --------------------------------------------------
              >
              > > * By using the ai-geostats mailing list you agree
              to
              > follow its rules
              > ( see
              > http://www.ai-geostats.org/help_ai-geostats.htm )
              >
              > * To unsubscribe to ai-geostats, send the following
              > in the subject or in the body (plain text format) of
              > an email message to sympa@...
              >
              > Signoff ai-geostats






              ___________________________________________________________
              Yahoo! Messenger - NEW crystal clear PC to PC calling worldwide with voicemail http://uk.messenger.yahoo.com
            • Thomas M. Parris
              Craig, You may wish to take a look at what FEWS-NET has been doing in Southern Africa. It look very similar to what you have in mind. See
              Message 6 of 6 , Jan 9, 2006
              • 0 Attachment
                Craig,

                You may wish to take a look at what FEWS-NET has been doing in Southern
                Africa. It look very similar to what you have in mind. See
                http://www.sadc-hazards.net/. Particularly,
                http://gisdata.usgs.net/sa_floods/aspmap/. I know there are several
                technical papers that provide additional detail. One such paper is at
                http://www.isprs.org/commission1/proceedings02/paper/00030.pdf. You might
                want to get in touch with the authors to see if there are any others.

                With best regards,
                Tom Parris
                ------------------------------------------------------------
                Thomas M. Parris
                Director of Sustainability Programs
                ISciences, LLC
                61 Main Street, Suite 200
                Burlington, VT 05401 USA

                Tel: +802-864-2999 http://www.isciences.com/
                Fax: +617-344-2580 http://www.terraviva.net/
                Email: parris@...
                -----------------------------------------------------------
              Your message has been successfully submitted and would be delivered to recipients shortly.