Identification and manipulation of variables
The requirements for any useful predictor
are firstly a good relationship with the selected
SPI and secondly, a reasonable correlation and lead
time (that is, months to season). For convenience,
the year in which the MEI and SPI values were
taken will be referred to as “year 0” and similarly
“year -1” for the year prior to year 0 (Hong et al.,
2001). Time scales for the accumulation of rainfall
in this study were defined in terms of the dry period
after the rainy season from November to April.
Therefore, six time scales of cumulative monthly
rainfall data at each station were transformed into
1-, 2-, 3-, 4-, 5-, and 6-month SPI values with mean
= 0 and standard deviation = 1. Those computed
SPI values were denoted