Statistical Downscaling

Direct output from General Circulation Model (GCM) simulations are largely inadequate for regional scale impact analysis due to their relatively coarse spatial resolution. Important sub-grid regional scale differences in climate may be filtered out by the coarse output resolution. In order to overcome these difficulties and to provide useful local/regional scale climate scenarios for impacts modelling, a number of techniques have been developed which utilise direct output from a GCM, but also take account of observed regional variations in climate. One such method is statistical downscaling. Statistical downscaling seeks to establish empirical relationships between observed mesoscale variables, such as, pressure, vorticity, geopotential heights, and an observed climate time series, such as, temperature, precipitation, radiation. These derived relationships between upper air and surface variables can then be forced by upper air variables output from a GCM, providing future climate scenarios that contain a regional signal.

Data was compiled from a thirty-year period of surface observations of precipitation, temperature and radiation from Met Eireann's and the British Atmospheric Data Center databases. This amounted to 570 stations for precipitation and 70 stations for maximum/minimum temperature. Data for incident radiation and sunshine hours were also acquired for as many locations as possible. Upper air data for the 1961-1990 period was obtained from the NCEP/NCAR Re-analysis project for the spatial domain around Ireland and regridded to the GCM output grid resolution (2.5o latitude by 3.75o longitude). The GCM utilised in this study to provide forcing to the derived empirical relationships was the Hadley Climate Model HadCM3. Results from the Coupled Model Inter-comparison Project suggested that HadCM3 was as effective in simulating mean monthly observed temperature and precipitation patterns as other leading models. Like most, however, HadCM3 was less effective in simulating observed precipitation, particularly polewards of 55oN. This implies less confidence can be attached to precipitation scenarios for these areas.

Daily output for the grid cell specific to Ireland was extracted from the GCM. The particular run concerned (HadCM3GGa1) was based on historical increases in individual greenhouse gases from 1860-1990 and then partly on the emission scenario IS95a. This involved a 1% per annum rise in radiative forcing but no consideration of tropospheric ozone. The end product was a 'middle of the road' scenario which produces global temperature increases of approximately 3.5oC by 2100. More recent runs of HadCM3 have employed the new SRES scenarios A2a and B2a. These have produced slightly more and slightly less warming respectively than GGa1, though the level of interannual 'noise' is such that the three scenarios produce almost indistinguishable trends until mid century and SRESA2a and GGa1 are relatively indistinguishable until the 2090s.

The resulting datasets were then used to generate four thirty year periods of climate. Two representing current climate using the NCEP and HadCM3 predictors for the 1961-1990 period and two representing modeled future climate for the 2041-2070 and 2061-2090 periods. The difference between the GCM future scenarios and GCM current [ie. GCM (2041-2070) minus GCM (1961-1990)] was then applied to the actual observed variable of interest to provide a climate scenario for a particular site. This differencing of downscaled GCM output is an important step in the methodology as it is suggested that GCMs better predict relative changes over absolute ones.

Statistical downscaling was carried out for approximately 250 precipitation stations, 65 maximum/minimum temperature stations and approximately 35 radiation/sun hour stations. Validation was performed using an independent dataset from 1991-97.
In order to derive climatological values for the areas intervening between the station locations, a regression model was used whereby each of the climatic variables was predicted according to the following equation:

Climatic Variable = a +bx + cy + dx2 +ey2 + fxy + gz

where a is a constant,
b - g are co-efficients derived from the regression,
x is the row number of the grid cell, y is the column number, and
z is the elevation.

Results

Current mean January figures are predicted to increase by 1.5oC mid century with a further increase of 0.5oC-1.0oC by 2075. By 2055, the extreme south and south west coasts are predicted to have a mean January temperature of 7.5-8.0oC. By mid century winters in Northern Ireland and in the north Midlands will be similar to those of Cork/Kerry during the 1961-90 period. Since temperature is a primary meteorological parameter, secondary parameters such as frost frequency and growing season length and efficiency can be expected to undergo considerable changes over this time interval.
July temperatures show an overall increase of 2.5oC by 2055 and a further increase of 1.0oC by 2075. Maximum July temperatures in the order of 22.5oC could be expected with areas in the central Midlands experiencing maximum July temperatures of 24.5oC.

The current geographical distribution of precipitation is largely dominated by orography. The upland regions on the west coast acting as a barrier to rain laden air masses flowing in of the Atlantic. There is a marked west-east gradient in precipitation as a consequence. This pattern is largely replicated in the 2055 and 2075 scenarios with the exception of the summer and early autumn months. Overall increases in precipitation are predicted for the winter months of December- February. On average these amount to 11%. The greatest increases are suggested for the north west where increases of approximately 20% are suggested by mid century. Little change is suggested as occurring on the east coast and in the eastern part of the Central Plain. These decreases in rainfall become more marked during the summer and early autumn months across eastern and central Ireland. Nationally, these are of the order of 25% with decreases of over 40% in some parts of the south-east suggested. Such decreases, if realised, would clearly have profound implications for agriculture and water resource management. Reductions on the south-east coast of similar magnitudes are also seen in the recently released UKCIPS02 scenarios generated from an RCM for the 2050 and 2080 periods with the Medium-High and High emissions scenarios.

Impact Assessments

Agriculture
The scenarios produced were used to drive crop simulation models for a range of present and potential future crops. The simulation results show that the expected climate changes will have a major impact on Irish agriculture which though significant cannot be regarded as potentially catastrophic.

For livestock production, the expectation of more frequent summer droughts may be expected to introduce the need for significant supplementation of grazed grass. At the same time, increased production of Maize is expected. This may allow livestock systems to be less rigidly geared to the grass crop. Maize silage may replace grass silage so that the land currently reserved from grazing for grass silage production would become available for grazing. Barley is another potentially important source of energy for supplemental feeding of livestock. The expected increases in cereal grain production may be expected to reduce the cost of feed Barley. However, the extra costs associated with irrigation may offset this if it proves necessary thereby bringing the economic viability into question, especially if Barley is in competition with Maize as a forage crop. Soybean is an important supplemental source of livestock protein and is currently imported. Soybean has the potential to replace Maize as the marginal crop in Irish agriculture.

In the eastern half of the country irrigation will become important for all crops. This will have a major impact on the economics, machinery requirement and labour demand in both tillage and livestock systems. In recent years it was estimated that irrigation in dairying in the drought-prone southeast is justified economically only if water is available without charge and without the construction of farm reservoirs. With the projected scenarios, a much greater area of agricultural land will be affected by drought loss, and the quantities of water involved to compensate by irrigation will be large. Given that agriculture may have to compete for scarce summer water extraction with other users, the consequent economic effects may make crops with good potential uneconomical.

Water Resources
A physical process-based hydrological model (HYSIM) was used to simulate effective runoff across a 10 x 10 km grid under future climatic conditions. This suggested that there will be a widespread reduction in annual runoff that will be most marked in the east and south-east of country. A slight increase may be observed over a limited area in the north-west. Winter runoff is predicted to increase in the west of the country, especially under the 2061-90 scenario where an increase in winter runoff is predicted for over 60% of the land area. The greatest increases are predicted to occur in the north west. All areas will experience a decrease in summer runoff, with the greatest reductions in the east of the country. It is likely that the frequency and duration of low flows will increase in many areas.

The magnitude and frequency of individual flood events will probably increase in the western half of the country. Seasonal flooding may occur over a larger area and persist for longer periods of time. Long term deficits in soil moisture, aquifers, lakes and reservoirs are likely to develop.

Natural Ecosystems and Biodiversity
Changes in climate zones were projected to have implications for natural ecosystems and biodiversity. The projected increases in temperature combined with a longer growing season were found to be potentially most detrimental to Arctic or Boreal relicts and mountain species. In contrast, species which are at their northern or northwestern limits may move northwards and possibly extend their range. Such changes are likely to result in significant alterations to habitat conditions though movement of habitats in Ireland will be restricted by non-climatic considerations. Montane heaths are suggested as being particularly sensitive to climate change while peatlands are expected to suffer considerably from summer drying.

Other Impacts

Other areas of potential impacts were also examined such as forestry and coastal impacts.

Conclusions

Climate changes over the next half century can be anticipated and their regional dimensions can be projected using statistical downscaling techniques. While considerable uncertainty remains, especially with respect to precipitation changes, forward planning is now required to accommodate climate change in Ireland. In key areas such as agriculture, water resources and the natural environment, climate change impacts are likely to be considerable and significant adjustment of present management practices will be entailed to ensure a sustainable future.