Spatiotemporal Pattern Networks of Heavy Rain among Automatic Weather Stations and Very-Short-Term Heavy-Rain Prediction

Algorithm 1

Pseudocode of the proposed prediction method using AWS networks of heavy-rain spatiotemporal pattern.

// MIN_SIZE            10

// INTERVAL_LENGTH     6 hours

// START_DATE        00:00 on 01/01/2003

// END_DATE         24:00 on 31/12/2012

predictUsingNetwork(START_DATE, END_DATE)

    for  each  hourly date between START_DATE and END_DATE

        if  heavy-rain event length starting at date ≥ INTERVAL_LENGTH

             predictPossibleStns(,   −  4,

                       −  2,  );

             for each  station

                 if   > POSSIBILITY_THRESHOLD

                     predict that station will satisfy heavy-rain criterion

                      at with possibility  ;

                 end if

             end for

        end if

    end for

predictPossibleStns(, , , )

    for each  station

         0.0;

    end for

     network constructed from to ;

     network constructed from to ; // actual test network

    // filter small-sized test networks for accuracy

    if  #nodes of < MIN_SIZE or #edges of < MIN_SIZE

        return;

    end if

    // get the list of nodes added between and

    for each  station

        if  node weight (stay time) of station in < that of station in

             add station to addedStnList;

        end if

    end for

    for each  network in train networks

         edge-based similarity between and ;

        if  is positive

             // cumulate the station possibility scores

             for each  directed edge (, ) incident to added station in

                 if   =

                     add to stnPossibility  of  ;

                 end if

             end for

        end if

    end for