NCEP/Climate Prediction Center ATLAS No. 5

A Precipitation Climatology for Stations in the Tropical Basin; Effects of ENSO

7. Spatial Distribution of Rainfall Percentiles

7. Spatial Distribution of Rainfall Percentiles Maps showing the spatial distribution of rainfall percentile for four regular seasons are presented in Fig. 7 for each year from 1955 to 1996. The four seasons used are Jan-Feb-Mar, Apr-May-Jun, etc., for reason of the climatology as explained above in section 3 with respect to Table 2. The maps in Fig. 7 show spatially coherent regions of rainfall deficiency (or surplus) as derived from the 66 stations used in this study. Regions with percentiles below the median (50 %ile) are shaded lightly, while those with above-median percentile values are indicated by dark shading. Contours are drawn for the 10%ile, 25%ile, 50%ile, 75%ile and 90%ile. It should be noted that there is some smoothing in the shading algorithm, so that in some cases the shading or the contours may contradict the station percentiles shown by the individual numerals on the map. The shading is intended to highlight general pockets of rainfall anomaly, while the percentile numerals show the actual rainfall conditions at individual locations more precisely. The ENSO status of the Oct-Nov-Dec and Jan-Feb-Mar periods is indicated above the maps for those periods, in keeping with the classification system discussed in the appendix and used in Table 4 and Fig. 6a. The effect of the giant 1982-83 warm ENSO episode, for example, is reflected in record or near-record dry conditions in Jan-Feb-Mar 1983 at many off-equator stations, and unusually heavy rainfall along the equator from just west of the date line eastward.

A cautionary note about Fig. 7 is in order. While large scale climatological influences such as ENSO can often readily be seen in the spatial patterns of rainfall percentile, much of the variation of rainfall percentile at individual stations may be attributed to local effects such as those related to topography and/or exposure with respect to the prevailing low-level wind flow. Thus, the microclimate measured by the rain gage may or may not well represent the rainfall generally received within its 3.5 latitude by 3.5 longitude region, or even in the island's own major watersheds. A second caveat is that at climatologically dry stations such as the three Hawaiian stations other than Hilo, sampling errors become relatively important, such that it is a matter of chance whether the only substantial rainfall event for a given season in the general locality provides a mere 2 mm of rainfall to a specific rain gage, or a plentiful 43 mm. The topography factor would be ameliorated with the use of a denser network of stations on islands having significant topography. While homogeneous rainfall records extending back to the mid-1950s do not presently exist for many more stations than those included here, shorter records are available at many locations and could be used to estimate climatologies that are more highly correlated with the hydrological welfare of the island in question. For example, 15 or more stations with records of 15 or more years may exist for Oahu island, Hawaii, even though only 1 station (Honolulu) was found to have a 40+ year record that has not recently been discontinued. The dry station sampling problem could be overcome by using rainfall surrogate measurements such as outgoing longwave radiation (OLR), which provides average cloud top heights (and hence the likely approximate precipitation amount) integrated over larger areas than those reflected in gage records. While OLR measurements are available only since the mid-1970s and are not perfectly correlated with ground-level rainfall, they represent a vast improvement over gage rainfall measurements in sparsely sampled regions such as in the tropical Pacific basin. Rainfall conditions as described by OLR have been compiled for much of the Pacific island region by Yu et al. (1997), and show large-scale anomaly patterns that generally agree with those implied by the percentile maps of Fig. 7 here.

back to the table of contents