## Info

week of the year

Figure 10.15 Measured values at Launceston, Tasmania, of the precipitation rate P, and values of the lake evaporation Eo (representing the potential evaporation rate Et), calculated from temperature values. The actual evaporation rate Ea is estimated as whichever is the less of Eo and 112 M2 mm/wk (Figure 4.10), where M is the soil's actual water content as a proportion of the soil's available-water capacity, assumed to be 100 mm of water. The soil dries out when Ea is greater than P, until M equals {Eo/ 112}05 (Linacre 1973:451). Recharge occurs when P is greater than Eo, until the soil is saturated. Thereafter, there is runoff R, as long as P exceeds Eo. The shaded areas of recharge and soil drying in the diagram must be equal because there is a long-term balance of drying and rewetting of the soil.

This happens when there is an abnormal runoff from a wide catchment area into the river—from heavy rains or the rapid melting of snow upstream. Abrupt, brief risings of the river are called 'flash floods'. Altogether, floods cause many deaths worldwide (Table 10.5), and the

Table 10.5 Numbers of people killed in natural disasters from 1967-91

Natural disaster

Number of deaths

Droughts (Section 10.7) 1,334,000 Tropical cyclones and storm surge (Chapter 13) 896,000 Storms and floods

(Section 10.6) 358,000

High winds (Chapter 14) 14,000 Extreme temperatures

(Chapter 3) 5,000

number increases with the growing population in flood-prone areas.

Floods are more likely where the ground is too impervious or already too wet to absorb more moisture. Thus the possibility of heavy rains causing flooding is estimated by monitoring or calculating the wetness of the ground of the various areas draining into the river (Section 10.5).

Flooding in a section of a river implies that the inflow upstream exceeds the outflow from the place of flooding, and both flows depend on the shape of the land. Flat areas which are notably vulnerable to flooding are called floodplains, which may be well downstream of the region where the rain falls.

The hazard is greater where there is a high variability of rainfall. So Australia is especially prone, particularly in the north (Section 10.4). The heaviest precipitation is often the consequence of depressions or tropical cyclones (Chapter 13), and causes floods if the falls are widespread. That occurred in some famous floods in Brisbane in the last century (Note 10.Q). Even earlier, seven major floods in the Hawkesbury Valley behind Sydney in the period 1795-1810 seriously threatened the food supply of the infant colony. The likelihood of such floods now is lessened by dams across the headwaters.

Years of particularly high floods in northern New South Wales are listed in Table 10.6. Flood years at Grafton coincided with those at Lismore, some 100 km away and in a quite separate valley. Almost all the forty-four floods occurred in the first half of the year when tropical cyclones are most prevalent (Chapter 13). Otherwise, there is no evidence of any regularity, and no clear correspondence with El Niño episodes (Chapter 11). The table shows that the frequency of major floods in that region has risen this century from about one to about ten each twenty years.

Extensive flooding occurs occasionally in Lake Eyre, the largest ephemeral lake in the world, draining an area of central Australia roughly 700 km across. It happened in 1916-17, 1920-21, 1949-50 and 1974-77, with minor floodings in 1907, 1940-41, 1953, 1955-59, 1984 and 1989, due to heavy rains in Queensland far away. Again, these dates do not correspond to those of El Niño or La Niña events (Note 10.M).

### 10.7 DROUGHTS

Wheat yields averaged only 0.67 tonnes per hectare during six Australian droughts between 1940 and 1968, instead of the 1.17 t/ha in the years immediately before and after. In the USA, 41 per cent of crop insurance payouts to farmers are for losses due to drought. Even more importantly, droughts are responsible for more human deaths than other natural disasters (Table 10.5).

They differ strikingly from floods. A drought can be as damaging as a flood, but affects a whole region and not just the low-lying parts. Also, droughts begin imperceptibly but usually end sharply with soaking rains, whereas floods start abruptly and have a lingering aftermath.

Droughts can be defined in several ways. What is called a meteorological drought occurs when there is little rain, compared with normal, in terms of the degree of variability at the place. For instance, drought may be defined as three

Table 10.6 Times of floods at Grafton and Lismore (NSW)

Grafton: years when river over 2.2 m above reference'

Lismore t

March 1890 April 1892 Feb. 1893 June 1893

May 1921

June 1945

March 1 955 April 1 962 Jan. 1967 Jan. 1974 May 1980

July 1921 Feb. 1928

March 1946

May 1 955 July 1 962 March 1967 March 1974 April 1988

June 1948

Jan. 1956 Jan. 1963 June 1967

April 1974 July 1988

June 1950 Feb. 1956

April 1963 Jan.1968 Feb.1976

April 1989 Feb. 1990

July 1950 Jan.1959 May 1963

Feb. 1971

Feb. 1954 July 1954

Feb. 1959 Nov. 1959 March 1964 July 1965 Nov. 1972

April 1990

1950,

1893

1921,1931 1945, 1948,

1954 1956

1962, 1963, 1965 1967, 1972

* The dates shown are of years when the Clarence River at Grafton rose more than 2.2 m above the reference level; on dates shown in bold the river rose beyond 6 m t The dates shown are of years when the Richmond River at Lismore (100 km north of Grafton) rose more than 10 m above the reference level t No data available consecutive months with rainfalls each within the lowest decile. This means that even a high rainfall leads to drought if it is much less than usual at that time of year, at that place; drought is not the same as aridity. However, even a modest reduction from normal leads to drought if there is usually little variation from the average. Thus a reduction of annual rainfall to 66 per cent of the average at Perth (where the variability is low) would constitute a severe drought, whereas such a reduction would be hardly noticed at Alice Springs because of the customary high variability of rainfall.

Other rainfall definitions of drought in terms of less than some certain amount within some specified period differ between countries. An agricultural drought is determined by soil-moisture conditions during critical stages in the growth of a crop. So it depends on previous runoff and evaporation (i.e. on radiation, temperature, humidity and wind) as well as rainfall. Hence parts of north-east Brazil suffered extreme meteorological drought, but not agricultural drought, during 1983. Hydrological drought reflects the drying up of streams, and therefore is governed by all the factors affecting runoff, considered in Section 10.5. This kind of drought, like agricultural drought, is revealed by water budgeting. Socio-economic drought is a time of a water shortage affected by management decisions. For instance, the water stored behind a dam may be reduced in anticipation of predicted heavy rains, giving rise to a drought if rains turn out to be merely normal. Or consider the official decision to 'declare' a drought in a region of New South Wales whenever half the sheep or cattle must be hand-fed or moved away for pasture; this depends on previous management decisions about stocking rates and kind of stock, and human judgement of the need to move it. Such drought may be induced by over-grazing, or by over-commitment of the water supply. We will focus on meteorological drought in what follows.

The best-known indication of the intensity of a drought is the American Palmer Drought Severity Index (PDSI), derived from complicated comparisons of (i) measured rainfall and estimated evaporation, with (ii) normal values, and summations of the monthly differences. The usefulness of the PDSI is limited by poor estimation of evaporation, the use of monthly averages, several arbitrary values, and the need for considerable information. It can be improved by more explicit water budgeting (Section 10.5), e.g. for estimating catchment runoff or the dryness of forests. In this case, drought intensity is described in terms of the millimetres of water needed to refill the soil (Note 10.P). There is a serious risk of a forest fire if the deficit exceeds say, 64 mm.

The occurrence and extent of droughts seem at first sight to be hopelessly irregular. For instance, Australian records show that over 20 per cent of the continent had annual rainfalls within the lowest decile on thirteen occasions between 1902 and 1982, spaced 3-14 years apart, with a median value of 5.5 years. Unfortunately, this is neither so rare as to be unimportant nor so frequent as to be accommodated within the normal routine of farming. The irregularity of occurrence is due to a combination of factors: (i) 'natural' variation, (ii) unexplained processes, (iii) weather rhythms, (iv) sea-surface temperatures nearby and (v) teleconnections, all of which operate together and are variously unpredictable. Let us consider them.

### Random Element

There is a large random element in all atmospheric processes, reflected in the variability of rainfall considered earlier (Section 10.4). This inherent property of climates inevitably produces a long series of subnormal rainfalls from time to time, just as tossing a coin leads to a long run of heads occasionally. Convective rainfall especially is associated with unpredictable randomness, as thunderstorms occur only in some parts of a region of unstable air but not in others (Note 7.G). When this effect is dominant it causes isolated patches of 'natural drought', mainly in semi-arid areas on the edges of deserts where the variability is greatest. It is quite normal in Australia.

### Unexplained Associations

Several occurrences of drought have been linked with preceding atmospheric conditions, indicating causative processes at work which are not yet understood. For instance, two consecutive wet years at Beer-Sheva in Israel are usually followed by drought. Similarly, fifteen drought years out of twenty-two in north-east Brazil were preceded by months of unusual wind patterns in the upper atmosphere of the northern hemisphere.

Persistence (Section 10.4) is a related feature of drought, perhaps due to positive feedbacks in the atmosphere prolonging a random dry period into a drought. Thus, the probability of a relatively dry summer at Alice Springs is 14.4 per cent, so that the chance of two dry summers together would be 2.1 per cent (i.e. 0.144x0.144) if the events were independent (Note 10.N). But, in fact, the chance proves to be 3.9 per cent— almost twice as much—so a dry summer this year somehow enhances the likelihood of one next year. Likewise, records of rainfalls in Africa during the period 1911-74 show more long runs of dry years than would be expected by chance.

### Rhythms

Occasionally there seems to be a rhythm of a decade or two in the occurrence of drought (Note 10.R). Tree-ring data (Note 10.D) from the American Midwest indicate gaps of about nineteen years between droughts, and a similar time has elapsed between some major droughts in New South Wales in 1809, 1828, 1857, 1866, 1885, 1895, etc. Also, there were extended dry periods in South Africa during the periods 190516, 1925-33, 1944-53, 1962-71 and 1981 onwards. Droughts occurred in the Transvaal each eighteen years or so, with about 10 per cent less rain than average. Similarly, the fluctuation of annual rainfall in Argentina shows minima about each 18.8 years, with around 20 per cent less than the mean. Also, droughts in north-east USA since 1840 indicate a periodicity of about nineteen years, like those in north-east China since AD 1500, whilst rings in the cross-

section of a cypress tree in North Carolina show a rhythm of 17.9 years.

All this might be due to the Moon, which moves in an ellipse about the Earth, with the axes of that ellipse themselves rotating each 18.6 years, the rotation being called the lunar nodal precession. This has the effect of varying the distance between Moon and Earth, and therefore the Moon's tidal pull on our atmosphere. Presumably this changes the pattern of winds and therefore of rainfalls. Alternatively, or in addition, the apparent rhythm of droughts might possibly be related to the time between sunspot maxima, i.e. to a cycle of roughly eleven or twenty-two years, though the relationship is complex and ambiguous (Note 10.S). Sadly, alleged links to the sunspot cycle or lunar precession may be based on poor analysis of selected data and in practice are too subtle to be of much help in forecasting rainfall.

### Sea-surface Temperatures

A factor which does certainly increase the chance of drought is an unusually low sea-surface temperature (SST) nearby. For instance, the dry years 1957, 1961 and 1965 in Australia were associated with sea-surface temperatures around the coast which were cooler than in the wet years 1950, 1955 and 1968. Similarly, some correlation has been found between the occurrence of drought in north-east Brazil and the adjacent Atlantic SST (Chapter 16), and between droughts in eastern Australia and the nearby Pacific SST off Queensland. Also, the rainfall between November and March along the eastern edge of South Africa depends notably on the temperature of the warm Agulhas Current and on the proximity of the Current to the coast (Chapter 11); drought is more likely when the Current is weak or absent. The top and bottom rows in Figure 10.16 show some connection between warm seas off South America and rainfalls at Santiago nearby.

This evidence that an abnormal coastal SST increases rainfall onshore is not surprising. The extra warmth means that the onshore winds can hold more water vapour (Note 4.C), and higher surface temperatures enhance convective uplift; both are factors which increase rainfall. The effect of SST on rainfall is one reason for considering the oceans in the next chapter.

### Teleconnections

Sometimes rainfall at a place increases in accord with warmer seas far away, either at the same time or months previously. For example, Figure 10.17 shows rainfall anomalies (i.e. differences from normal) in Papua New Guinea occurring about a month after anomalies of SST at the other end of the Pacific ocean, over 10,000 km away. Such a relationship is called a teleconnection between the weather at one place and the weather somewhere remote. Other examples of teleconnections involving rainfall are as follows:

1 Warm seas more than 2,000 km east of Brazil are associated with high rainfalls in December at 4°S on the Brazilian coast.

2 Relatively high SST and more westerly winds in the equatorial Atlantic are associated with above-average precipitation over coastal regions of Angola in Africa (Figure 10.18).

3 Each of the twenty-two periods of high sea-surface temperature off Peru between 1871 and 1978 (with temperatures 2-4 K above normal) was associated with about 9 per cent less monsoonal rain than usual over most of India.

There are also teleconnections between droughts in separated land areas. H.C.Russell noted in 1896 that droughts in India and Australia tended to occur in the same years. Likewise, Figure 10.16 shows a tendency towards a coincidence of droughts in Australia,

Figure 10.16 The approximate occurrences of El Niño, droughts in various places, and rainfall in Santiago (33°S). The dates of warm sea surfaces off Peru are shown in row (a), and years with a low Southern Oscillation Index (see Chapter 12) in row (b). Row (c) shows years when there were either 'strong' or 'very strong' ENSO warm episodes (Chapter 12), with considerable repercussions economically etc. Droughts in Australia are indicated in row (d), in Otago (NZ) (e), northeast Brazil (f), Midwest USA (g) and South Africa (h). Row (i) shows years when the rainfall at Santiago exceeded 500 mm, i.e. unusually wet years. (For more details, see Section 17.3 concerning this diagram.)

Figure 10.16 The approximate occurrences of El Niño, droughts in various places, and rainfall in Santiago (33°S). The dates of warm sea surfaces off Peru are shown in row (a), and years with a low Southern Oscillation Index (see Chapter 12) in row (b). Row (c) shows years when there were either 'strong' or 'very strong' ENSO warm episodes (Chapter 12), with considerable repercussions economically etc. Droughts in Australia are indicated in row (d), in Otago (NZ) (e), northeast Brazil (f), Midwest USA (g) and South Africa (h). Row (i) shows years when the rainfall at Santiago exceeded 500 mm, i.e. unusually wet years. (For more details, see Section 17.3 concerning this diagram.)

New Zealand, Brazil and Midwest USA, near the times of the sea-surface warming episodes off Peru, the El Niño (Note 10.M), and of low atmospheric pressures at Tahiti, shown by a low 'Southern Oscillation Index' (Chapter 12). (The fact of only a 'tendency' towards coincidence shows the influence of other factors which lead to droughts, as well as the difficulty of specifying the year of a drought which extends from spring in one year to autumn in the next.) On the other hand, droughts do not occur simultaneously in both East Africa and South Africa, while droughts in Chile tend to coincide with unusual flooding of the Nile. Also, droughts occur in different years in the east and west of Australia (Figure 10.19), presumably because the global pattern of winds affects the two halves of the continent differently (Chapter 12).

The causes of these various teleconnections

Figure 10.17 Parallelism of changes of rainfalls at Lae (in Papua New Guinea) and of sea-surface temperatures between 5°N-5°S and 80-180°W, in the east equatorial Pacific. The rainfalls and the temperatures shown are twelve-month 'running means' (Note 10.H).
Figure 10.18 Sea-surface temperatures SST and winds in the Atlantic at low latitudes associated with weather on the Angolan coast.

unusually wet

|1987L

|1987L

1890

% 70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 80%

fraction of the state under drought

Figure 10.19 Percentages of the areas of New South Wales and Western Australia, respectively, affected by drought, between 1888 and 1987.

are not well understood, though their scale implies a dependence on global patterns of winds (Chapter 12).