Displaying drought severity, date, and duration.
This image shows how an extreme drought developed at Manilla, NSW in 2018. The contours and colours indicate drought severity. They show whether the rainfall shortage was extreme, severe, or merely serious. At times without drought, they show when the rainfall was above normal.
[See Note below: “Classes of rainfall shortage”.]
Calendar months appear in order along the top, showing how the severity of drought has changed as time has passed.
Down the side, the duration of drought is shown, from one month duration at the top to thirty-six months duration at the bottom. Each month is shown on one vertical line, with a severity value plotted against each duration value.
[In other posts (like “Record 15-month Drought in 2018”) each individual month’s drought data is plotted as severity versus duration.]
The pattern of the 2018 drought
In each of the months May, June, and July 2018, the monthly rainfall was a serious or severe shortage, below the 10th percentile. As rainfall had been below normal also in March and April, each month after April saw severe and even extreme shortages that extended to durations of 3 months and longer. By August, there were extreme droughts of five-month and six-month duration, despite reasonable rain (28 mm) having fallen in August.
By September, the nine months to date were in extreme drought. That is to say, the first nine months of 2018 had a total rainfall lower than any but 1% of all nine-month periods in history.
However, in that month (September 2018), the rainfall shortage for a 12-month duration was not extreme, but merely severe. Yet the 15-month shortage for that month was also extreme. In fact, it was the lowest 15-month rainfall total in history (400 mm).
The source of the extreme 15-month shortage at this date is obvious. One year earlier, the month of September 2017 had a serious rainfall shortage, and the two previous months had little rain. From these short-term shortages in spring 2017, shortages of longer duration descend as an arc across the graph. They become mild nine-month shortages during the summer, then worsen to extreme by merging with 2018 shortages.
Manilla’s rainfall is seasonal, with two distinct modes, each dominating half the year. There is a major summer (monsoonal) mode and a minor winter (westerly) mode. The summer rainfall mode dominates from October to March and the winter rainfall mode dominates from April to September. [See Note below: “Manilla’s rainfall seasonality”.]
This graph documents a failure of winter rainfall. The 2018 months that had rainfall shortages were the months of dominance of the winter rainfall mode (April to September). Rainfall had been near or above normal in the preceding summer rainfall mode (October 2017 to March 2018), and was so again in October 2018. [See Note below: “Limitations of this analysis”]
The rainfall shortages in the previous year (2017) were also restricted to the months of the winter rainfall mode, but began very late. Shortages did not develop until the final three months (July, August, September). Only when the extreme drought of 2018 developed did those 2017 shortages have a big effect.
Note: Classes of rainfall shortage
I have adopted two classes of rainfall shortage from the classes of “Rainfall deficiency” defined by the Bureau of Meteorology in their Climate Glossary as follows:
“Serious rainfall deficiency: rainfall lies above the lowest five per cent of recorded rainfall but below the lowest ten per cent (decile range 1) for the period in question,
“Severe rainfall deficiency: rainfall is among the lowest five per cent for the period in question.”
The Manilla rainfall record allows me to go beyond these two classes. Because the record extends back 134 years, it includes more than 1200 cumulative monthly rainfall values. I can identify percentile ranks even below the 0.1th percentile.
To the Bureau’s two classes of deficiency I add a third:
“Extreme deficiency (or extreme shortage): rainfall lies below the lowest one percent for the period in question.”
Rainfall rate versus percentile rank
Classes of rainfall deficiency based on percentile rank (rarity of occurrence) are practical, while classes based on percent of normal rainfall are not. However, some features of graphs showing percentile values (such as this) may be puzzling if one point is not understood. For any class, such as “Extreme shortage”, the rainfall rate in millimetres per month is not the same for droughts of shorter and longer duration. The longer the duration considered, the higher the rainfall rate in that class of shortage.
For example, consider an extreme shortage, below the 1st percentile. A 3-month period at Manilla would be in extreme shortage only if the rainfall was below 8 mm per month (18% of normal), but a 36-month period would be in extreme shortage so long as the rainfall was below 39 mm per month (89% of normal). See the post “Short Droughts are Worst”.
Note: Manilla’s rainfall seasonality
At Manilla, rainfall is distributed through the year in a remarkably regular way. Details are in the post “A Seasonal Rainfall Model for Manilla, NSW”.
There are two distinct modes, each with a peak a few days after the solstice of summer or winter. The two modes, each forming a bell-shaped curve of normal distribution, account for almost all of the rainfall in the year. The curves have similar width in months, but the summer (monsoonal) mode brings 64% of the rain and the winter (westerly) mode brings only 36%. In simple terms, the summer mode extends through the six months October to March, and the winter mode through the six months April to September.
Note: Limitations of this analysis
Monthly rainfalls form a single population
For each month of record, the observed rainfall has been tabulated, and added to those of preceding months to form n-month totals. Then percentile ranks have been calculated using the whole data record, in which the median monthly rainfall is 44 mm. No distinction has been made between calendar months. As a result, January months, which generally have the highest rainfall, (median value 75 mm) will be ranked here as rather less drought-prone than other months.
This limitation will not affect those n-month totals that comprise whole years, such as 24 months. When I inspect the main graph here, I do not see any discrepancy between the drought trajectories for whole-year durations (12, 24, and 36 months) as compared to other durations.
Since adjusting the analysis to account for the differing rainfall in calendar months is difficult, I am letting this simpler analysis stand.
Observations are not retrospective
When a current month’s data shows (for example) an extreme rainfall shortage for a duration of 15 months, this graph (or the related line graph) will plot that 15-month duration at the current month. If the previous month did not have an extreme rainfall shortage of that duration, it will not have been recorded there. Logically, that 15-month extreme shortage must have commenced 14 months earlier. It began then, but the evidence for it has appeared only now.
The data for each current month must always be provisional concerning the drought status at durations longer than one month. I am also preparing graphs that project shortages back to the date that they began.