3. Drought Risk under Scenarios of Climate Change
Key Points:
- Under all our climate change scenarios, average annual PED increases across virtually the entire country (that is, it gets drier), except for the South Island west coast, by the 2030s. Average annual PED increases even more by the 2080s.
- The risk of drought (extreme PED) increases in most eastern parts of the country - areas that are drought-prone already.
- Under all our climate change scenarios, a 1-in-20 year drought in eastern regions becomes more common in future. By the 2080s, the frequency of a current 1-in-20 year PED increases between two and more than fourfold, depending on the scenario. That is, a drought that currently occurs once in 20 years on average could become a 1-in-10 year, or even a 1-in-5 year, event in that same area.
- The areas where drought risk is projected to increase significantly include parts of North Otago, Canterbury, Marlborough, Wairarapa, Hawkes Bay, Bay of Plenty, and Northland.
- Because all the scenarios predict increased PED accumulation over the course of a year, drought periods are likely to 'expand' into spring and autumn more often than currently. In our most severe (medium-high) scenario, the drying of pasture in spring is advanced by about a month in the 2080s in dry eastern regions, compared to the current climate.
3.1 Change in Average Potential Evapotranspiration Deficit
The key result is that average annual PED increases across virtually the entire country, except on the South Island west coast, under all scenarios. As expected, the drying tendency increases with time, so is greater in the 2080s than the 2030s, and is larger under the scenarios which assume higher global temperatures (i.e., for 75% scaling than for 25% scaling). The drying tendency is also more extreme for the Hadley model than for the CSIRO projections, as expected from the much greater rainfall reductions in the east in the Hadley model (Figure 1.1). The changes are most significant in already dry eastern regions.
There are, of course, seasonal and regional differences, and quite large intensity differences, between the simulated drought occurrences of the CSIRO and Hadley models. However, the integrated effect of both precipitation and PET changes on drought occurrence is relatively uniform, which gives us confidence in our findings.
Figure 3.1 shows the most extreme result for the two models (ie, 2080s, 75% scaling). The CSIRO model projects an increase in PED of at least 90mm (average increase of about 3 weeks in length of pasture deficit) over the already dry eastern parts of New Zealand by the 2080s. In the Hadley simulation, the PED increase shows the same geographic pattern but with PED increases of over 180mm common throughout eastern regions. The Hadley model also suggests the drying effect is stronger in the North Island, which is consistent with the model's larger reduction in precipitation at lower latitudes (Figure 1.1).
Figure 3.1 Average change in annual accumulated PED (in mm) between the current climatology and projected climatology for the 2080s according to the CSIRO (left) and Hadley (right) models, scaled to the IPCC 75% global warming.
Figure 3.2 Average annual accumulated PED (in mm) at two gridpoints corresponding to Lincoln and Napier.
Figure 3.2 show line plots of PED levels at the Lincoln and Napier gridpoints. (Mapped changes for the other scenarios are provided in the Appendix (section 6.9)). For the most benign scenario of CSIRO 25% scaling, the average annual PED increases at Lincoln from 469mm to 532mm by the 2080s, an increase of more than 2 weeks in restricted pasture growth. The same holds at Napier, where average annual PED rises from the present 438mm to 511 by the 2080s. For the most extreme scenario of Hadley 75% scaling, the average annual PED by the 2080s is 629mm at Lincoln and 693mm at Napier, suggesting an additional 6 weeks or more of reduced pasture growth in an average year.
Overall, water deficits in an average year are projected to increase by between about 50 mm and 250 mm PED in the driest regions by the 2080s, depending on the climate scenario and location. To put this in context: annual averages are currently about 300-500 mm PED in these areas. In some dry areas, a 200 mm increase in average annual PED would mean that a drought of medium severity (such as the 1991/92 drought in Canterbury) could become the norm in those areas by the 2080s.
3.2 Changes in Drought Risk
3.2.1 Change in Extreme Potential Evapotranspiration Deficit
Along with the increase in average PED, there is a corresponding increase in the risk of drought (or extreme PED) in most eastern parts of the country. Figure 3.3 shows the projected changes in the 1-in-20 year PED, and the probability of PED exceeding 600mm. Time series plots are shown for all scenarios at the Lincoln and Napier gridpoints (see Appendix 6.9 for maps of whole country). The 1-in-20 year PED (upper panels of Figure 3.3) is very similar in shape to the average PED (Figure 3.2). Typically, the 1-in-20 year PED is about 300mm higher than the climatological average at both sites, and this difference remains fairly constant for all scenarios.
The lower panels of Figure 3.3 show how the probability of at least 600mm annual PED accumulation varies with the scenario. Again, the Hadley model shows more extreme changes in time (the Hadley change at the 2030s is comparable to the CSIRO change by the 2080s) and a more extreme North Island change than South Island one.
3.2.2 Change in Return Period
An alternative way of describing the changes in drought risk is to calculate what the future return period is for a PED value that currently occurs with a 1-in-20 year return period. That is, if we consider this level (1 in 20 years) to be a significant anomaly over the growing season (and it will certainly be a severe drought in the east), how much more common will it become?
The results are shown in Figure 3.4 for the four scenarios of the 2080s, where the current 1-in-20 year drought becomes more common everywhere that is not shaded grey. At the boundary between dark blue and yellow, the future return period is 10 years; here, a current 1-in-20 year dry event becomes twice as likely in the future scenario. At the yellow-brown boundary, the event becomes four times as likely (every 5 years on average, instead of every 20).
The two upper panels show PED levels with a 1-in-20 year return period (5% chance of occurrence in any one year), and the two lower panels show probabilities of annual PED exceeding 600mm. For context, the worst droughts in the recent historical record had 851mm PED at Lincoln (in 1988/89), and 799mm at Napier (1997/98).
Under the 'low-medium' scenario (25% scaling, CSIRO model) by the 2080s, severe droughts are projected to occur at least twice as often as currently in the following areas: inland and northern parts of Otago; eastern parts of Canterbury and Marlborough; part of the Wairarapa; parts of Hawkes Bay; parts of the Bay of Plenty; and parts of Northland.
Under the 'medium-high' scenario (75% scaling, Hadley model), by the 2080s, severe droughts are projected to occur more than four times as often in the following regions: eastern parts of the North Otago, Canterbury and Marlborough; much of the Wairarapa, Bay of Plenty, and Coromandel; most of Gisborne; much of Northland. For many of the other eastern regions, the frequency of severe drought is projected to at least double by the 2080s under this scenario.
Figure 3.4: Future return periods (years) of current climate 1-in-20 year PED events, for four scenarios: CSIRO 2080s 25% and 75% scaling (upper panels) and Hadley 2080s 25% and 75% scaling (lower panels).
Grey areas indicate regions of very low drought risk (where return period can't be estimated) and/or regions where drought risk decreases.
The table below summarises changes in severe drought risk by the 2080s for characteristic locations in some currently drought-prone locations. The other two scenarios for the 2080s not given in Table 3.1 (ie, CSIRO 75% and Hadley 25%) are intermediate between the two scenarios shown.
Table 3.1: Present drought risk, and future changes in 2080s for CSIRO 25% and Hadley 75% scenarios, at selected locations.
The first three columns of the table provide information on how dry the current and future 1-in-20 year droughts could be. The last two columns indicate how often a drought that currently occurs only about once in 20 years could occur in future.
| Location | Present climate PED (mm) for 1-in-20 year drought | 2080s CSIRO 25% scenario PED(mm) for 1-in-20 year drought | 2080s Hadley 75% scenario PED(mm) for 1-in-20 year drought | 2080s CSIRO 25% scenario. Average return interval (years) for current 1 in 20 yr drought | 2080s Hadley 75% scenario. Average return interval (years) for current 1 in 20 yr drought |
|---|---|---|---|---|---|
Ranfurly (N. Otago) |
645 |
700 |
725 |
8.5 |
6.5 |
Darfield (E.Canterbury) |
465 |
515 |
650 |
10.5 |
3.5 |
Blenheim (E. Marlborough) |
895 |
955 |
1035 |
12.0 |
7.0 |
Napier (Hawkes Bay) |
740 |
820 |
1010 |
9.5 |
2.5 |
Whangarei (Northland) |
415 |
465 |
580 |
8.0 |
3.0 |
Similar maps of return period changes by the 2030s are given in the Appendix (section 6.9). It needs to be recognised that natural variations in climate on the decadal time-scale can also affect drought incidence. The importance of natural variations, relative to those caused by anthropogenic change, decreases the further ahead in time one goes.
3.2.3 Overseas studies of changes in drought under global warming
There have been a large number of international studies of changes in water availability under global warming. We have been unable to find a study that is truly comparable to ours in terms of using PED as a quantitative indicator of changes in drought risk, and calculating return period changes for drought. Nonetheless, overseas studies have found comparable changes in water resources (runoff or some other measure). They are discussed in more detail in the Appendix (section 6.12).
3.2.4 Change in Seasonality of Future Drought
Figure 3.5 illustrates another aspect of drought change at the Lincoln and Napier gridpoints. The figure shows monthlyPED accumulation for the Hadley 75% scenario. This scenario is the most extreme; all other scenarios show the same direction of change but are less pronounced.
Figure 3.5 Monthly accumulation of PED (in mm) at the Lincoln (left) and Napier (right) gridpoints, comparing historical distributions with those projected by Hadley model with IPCC 75% scaling.
Figure 3.5 indicates how the July-June PED accumulation is built up over the year. The greatest increments to PED currently occur in December and January at the two sites illustrated, and this is predicted to continue to be the case in the future. However, under future warming there is increased accumulation throughout the year, and we can therefore infer that drought periods will tend to 'expand' into the spring and autumn months more often than currently. The average drying tendency (PED increment) during the spring and early summer will tend to occur about one month earlier by the 2080s under the Hadley 75% scenario (ie, the current 'dryness' at the end of November would in future occur at the end of October under this scenario).
