Customising the Results

The forecasts and hindcasts are expressed in the units of the Y input data, but can be converted to anomalies, standardised anomalies, standardised indices (i.e., the SPI), or if the zero-bound option is set, as percentages of average. In most cases, it is recommended that the Y input data be in absolute units rather than as any of the standardisations that CPT can perform because it is then possible to choose whichever standardisation you prefer.

The definitions of the three categories used in CPT can be customised using the Options ~ Tailoring menu item. There are three options for defining the categories, by:

  • Climatological (i.e., prior) probabilities of the outer two categories. Typically, the prior probabilities are set at 33% so that the boundaries between the categories are the terciles of the climatological distribution, which is the default definition in CPT. These probabilities can be reset, thus allowing probabilities for more extreme rainfall events, for example to be calculated. However, it should be noted that errors in estimating the thresholds for extreme events can become large, and so changes in the calculated probabilities for extremes may be inaccurate, especially when the probabilities are converted to odds or relative odds.
  • Absolute values (e.g., mm of rain). The values of the thresholds are assumed to be in the same units as the values of the Y input data. When usig absolute values, the prior probabilities of the above- and below-normal categories may no longer be equal, and so care should be taken when interpreting forecasts, and it may be helpful to look at the relative odds. If the lower threshold is set to be larger than the upper threshold, the two thresholds are swapped without warning. The absolute thresholds can also be expressed as anomalies, standardised anomalies, standardised indices or, if the zero-bound option is set, as percentages of average, depending upon the standardisation option selected.
  • Reference periods. The amount of rainfall, for example, in specific periods in the training data can be used to define the thresholds. It does not matter which period is listed first; CPT automatically assigns the lower value as defining the below-normal category, and the higher value as defining above-normal. It is possible that the period defining below-normal at one location becomes the period defining above-normal at a different location.

Regardless of how the thresholds are defined, if the probabilities for the two outer categories sum to 1.0 (i.e., the normal category is impossible), the detailed forecasts in Tools ~ Forecasts ~ Series are shown as if there were only two categories: above and below.

By default, the climatological period used to calculate the average, and to define the category boundaries, is the current WMO standard (1991 - 2020), but if the input data do not span this period, the closest dates are selected. The climatological period can be adjusted using the Options ~ Climatological Period menu item.

Last modified: