NZSM Online

Get TurboNote+ desktop sticky notes

Interclue makes your browsing smarter, faster, more informative

SciTech Daily Review

Webcentre Ltd: Web solutions, Smart software, Quality graphics

Feature

Of Droughts and Durations

When is a 100-year drought not a 100-year drought?

A.M. Fowler

Near the height of Auckland's so-called water "crisis" in mid-1994, statements were being made to the effect that the region was in the midst of a 100-year drought. This claim rather surprised some local climatologists and prompted a number of investigations into the drought's significance. The results from one such study raise some rather interesting questions about just how rare events of such magnitude really are.

The return period of a drought is the average interval between events of the same or greater severity, for a specific duration. Thus, over the next 100 years and for a given duration, our best guess would be that we will experience one 100-year-return-period event, but more than one -- or none -- would not be unusual. Duration is critical and is often a cause of confusion about the return period concept because the same drought is likely to have different return periods for different durations. A 100-year drought for a duration of two years may well be a 50-year or 200-year event for a duration of three years.

Closer inspection of the 100-year drought claim reveals that it is based on analysis of rainfall in the Hunua Ranges, dating from 1928. In fact, rainfall collected by the Manganese Mine gauge was the lowest in 66 years of record, for a duration of 18 months. Clearly, the drought was a significant event but the question of duration needs to be kept firmly in mind. The drought was less significant for both shorter and longer durations.

The Hunua Ranges are the main water source area for Auckland, hence the focus on rainfall records available for that region. Droughts, though, are regional-scale phenomena and much useful information can often be gleaned from analysis of longer-term records available for more distant sites. This is especially so when assessing the significance of relatively rare events, as in this case. Albert Park (used here) is the longest rainfall record available in the Auckland region.

Interestingly, inspection of the monthly rainfall data from Albert Park (from January 1910 to December 1995) indicates drier periods than the 1994 drought for all five durations plotted. Particularly noteworthy is the 1913-15 drought, which stands out clearly as the most severe drought on record. Also clearly evident is the significance of duration when assessing drought severity. For example, minimum rainfall totals associated with the 1994 drought for durations of one and two years are not as low as those recorded during two other droughts. For durations of three, four and five years, there are respectively five, three and four more significant droughts in the period 1910-1990 than that experienced in 1994.

Contrary to the Albert Park results presented here, rainfall records in the Hunua Ranges indicate that the 1994 drought was more significant than that in 1983, for durations around 18 months. However it is reasonable -- and, in my opinion, prudent -- to assume that the 1913-15 drought was regionally the most significant event since 1910. It follows that, at worst, the 1994 drought is only the second largest event in 86 years. If durations of three or more years are most relevant, as indicated by the late 1980s bulk water supply study, then a return period in the vicinity of 30 years is suggested.

Although the Albert Park rainguage has been in its present location only since late 1909, rainfall data nominally assigned to "Albert Park" goes back to 1853 at three other nearby sites. Between 1853 and 1868, rainfall was recorded at Albert Barracks, quite close to the current guage location. In 1868, the recording site was moved about 2.5 km to the Auckland Domain, where it remained until 1883. Between 1883 and 1909, the recording site was at the "Old Museum" location. It moved a few hundred meters to its present location in1909.

Rainfall observations from the three earlier sites are considered "rather unsatisfactory", especially between April 1883 and August 1909 when the rainguage was located on a roof! Problems are confirmed by a simple "mass curve analysis" of the monthly rainfall data. Accumulated rainfall from the first month of observation (January 1853) can be plotted against the relevant data. If site changes had no major impact on the recorded rainfall we would expect the cumulative totals to plot against time as close to a straight line. The fact there are major slope changes (especially at about 1909) indicates that the 1853-1995 daily rainfall time series should not be treated as a single homogenous record.

In the case of the "Old Museum" record, under-measurement of rainfall would be expected as a result from the rooftop position of the rainguage. Increased exposure typically leads to under-measurement, as rainfall is increasingly swept over the top of the raingauge. A 20% upward adjustment of the rainfall data from the museum site is considered necessary to correct for under-measurement relative to a traditional ground site.

Dealing with the data from the Auckland Barracks and Domain sites is more problematic. Again, the slopes of the mass curve analysis are less than for the 1909-1995 period as a whole, suggesting possible measurement relative to the latter period. The difference is most pronounced for the Domain record and noteworthy given that mean annual rainfall maps show an increase in rainfall south of Albert Park. On the other hand, there are indications of periods since 1909 when the slope is comparable to these earlier periods (since 1980 and especially 1909-1915). It is possible then that these earlier records are in fact indicative of a period of lower rainfall of the order of 10-17% rather than a homogeneity problem.

In the context of water resource planning, the problem caused by uncertainties about the quality of the pre-1909 rainfall record is clearly demonstrated when you look at the graph below. Here the two-year-duration plot shown in the earlier graph has been extended back to 1853, using the additional data detailed above. The solid line is for the data as recorded, while the dashed line is a revised plot for the April 1883 to August 1909 period following scaling up of the Museum Roof data by 20%. The upper thin horizontal line is the mean rainfall total for two years, calculated over the period 1920-1990 (2502mm).

The lower line corresponds to the minimum rainfall over a two-year period recorded during the 1994 drought (1885mm). Using the adjusted data for the period 1883-1909, the graph shows six droughts in the second half of the nineteenth century more significant than the 1994 drought. This would make the 1994 drought only the ninth most significant in 143 years (less than a 20-year event). However, if the pre-1883 data were also scaled up to give the same mean rainfall as 1909-1995, then only one further drought (ca. 1887) would be larger than the 1994 drought. Added to the two twentieth-century droughts (ca. 1915, 1983), this would suggest a 30-year to 40-year event in 1994. Moreover, if one accepts that the 1994 drought was in fact more significant in the Hunua Ranges than the 1983 drought, then perhaps only the 1914/15 drought is more significant, which makes a 100-year event for a duration of two years more plausible.

Water resource planners are faced with a dilemma. Should they treat the late nineteenth century record as reasonable, indicative of a somewhat drier climate regime than experienced most of this century? Or should they treat the early rainfall data as suspect and either exclude it from consideration or adjust it upwards to conform with the pattern exhibited by more recent records?

A precautionary approach suggests the former, ensuring that planning is not based on what may be an atypically wet few decades. A rather more sophisticated analysis than that presented here is required to answer such questions. But, quite clearly, the result of any such analysis will be highly sensitive to the approach adopted and will have a major impact on any assessment of the significance of past and future droughts.

A M Fowler is in the Department of Geography at the University of Auckland.