ENSO-CLIPER provides predictions of the primary indices that help to depict the El Nino-Southern Oscillation phenomenon. These are the four sea surface temperature (SST) indices - Nino 1+2 (0-10S, 80-90W), Nino 3 (5N-5S, 90-150W), Nino 3.4 (5N-5S, 120-170W) and Nino 4 (5N-5S, 150W-160E) - and the Southern Oscillation Index (the SOI - standardized difference of the sea level pressure anomalies for Tahiti minus Darwin). The user of the program specifies which of these five predictands one wishes to get predictions for. The model can be run both in hindcast mode (i.e. for events that have already happened) back to an initial date of 1 January 1951 or in real-time mode (i.e. using the most recent complete month for the starting point).
To run ENSO-CLIPER to get a forecast based upon current conditions, one must have an updated copy of two files from the U.S. Climate Prediction Center: "sstoi.indices" and "soi". The files are available from their web site at CPC. Both files must start with the year 1950 and are monthly values. The sstoi.indices file (containing the Nino 1+2, 3, 3.4 and 4 SSTs and anomalies) can be utilized for the model as is. The "soi" file, however, must first be edited to remove the first half of the file so that only the data that is titled the following is kept:
(STAND TAHITI - STAND DARWIN) SEA LEVEL PRESS
STANDARDIZED DATA
After compiling the program (ensoclip00.f), one needs to simply specify which ENSO index is desired and what is the date (month and year) to initialize the model. If an error occurs in starting up the program, the following message will be printed back to the screen:
Usage: ensoclip00 ensotype month forecast-year
ensotype: 1) Nino4 2) Nino3.4 3) Nino3
4) Nino 1+2 5) SOI
month: forecasting point (i.e. 1 = Jan.1 12 = Dec.1)
forecast-year: (i.e. 1996, 1997....) (must be > 1950)
For example, if one wanted get the ENSO-CLIPER prediction of Nino 3.4 SST anomalies with a 1 October 2000 initialization, the user would type:
ensoclip00 2 10 2000
The output of ENSO-CLIPER are values in 3 month average increments, first with the previous 8 seasons (8 periods of 3 month average) values, then with the predicted values for 0 to 7 seasons lead time (0 to 23 months into the future). Nomenclature utilized here refers to a 0 season lead as the immediately upcoming three months. So for the example of a 1 October initial date, a 0 season lead (0 to 2 months lead) refers to October through December conditions; a 1 season lead (3 to 5 months lead) would be for January through March of the next year; a 2 season lead (6 to 8 months lead) would be for April through June of the next year; and so on. The following is the output generated for the above example:
#ENSO-CLIPER output for Nino 3.4 # Date Observed 1998.92 -1.37 1999.17 -1.21 1999.42 -0.75 1999.67 -0.86 1999.92 -1.25 2000.17 -1.43 2000.42 -0.50 2000.67 -0.29 # Date Forecast RMSE 2000.92 -0.29 0.30 2001.17 -0.27 0.39 2001.42 -0.18 0.41 2001.67 0.26 0.55 2001.92 0.40 0.80 2002.17 0.37 0.66 2002.42 0.18 0.49 2002.67 0.64 0.60
The output is contained in a file that is named "2102000.forecast". The format for the title is "#MMYYYY.forecast" - # is ENSO type (1 is Nino 4, 2 in Nino 3.4, 3 is Nino 3, 4 is Nino 1+2, 5 is SOI) MM is month [one M if month is before October] and YYYY is year. Values provided are in degrees Celsius anomalies (or for the SOI, in standardized deviations). "RMSE" is adjusted root-mean-square- error expected for the accompanying prediction. Note that because of the flexible way that ENSO-CLIPER is written, changing base periods (i.e. 1950-79 vs. 1961-90) do not affect the performance of the model. Whatever base period is used for the input anomalies (in the sstoi.indices and soi files) for ENSO-CLIPER is also utilized for the prediction output returned. (As of the writing of this update, CPC utilizes a 1961-90 base period for the SSTs and a 1950-79 base period for SOI.) The performance of ENSO-CLIPER should still be as originally designed for at least another ten years, unless there is a dramatic change to ENSO during the first decade of the 21st Century.
For more information on the design and performance of the model, please refer to Knaff and Landsea (1997) and Landsea and Knaff (2000).