(i)  the P index showed a weak positive correlation with summer EWR suggesting that westerly weather types are associated with higher rainfall than other directional weather types. There was  no significant relationship in the other three seasons.

(ii)  the S index showed a weak positive correlation with winter EWR indicating that southerly weather types in this season are associated with higher rainfall.  There was no significant relationship in the other three seasons.

(iii)  the C index was positively and significantly correlated with EWR in all four seasons indicating that cyclonic weather types are associated with high rainfall and anticyclonic types with low rainfall throughout the year.  All four relationships were highly significant with 43-63% of the variability accounted for.

Murray & Benwell (1970) demonstrated that multiple correlations between monthly PSC indices and monthly CET were generally greater than those between single indices and CET whereas multiple correlations between monthly PSC indices and monthly EWR did not differ significantly from simple correlation coefficients between the C index and EWR. These conclusions for individual months are consistent with the correlations shown in Table 3 for seasons.

The various Figures presented above in Method 1 are discussed with the other results later on in the study.
 

The strong correlations between PSC indices, temperature and rainfall (Table 3) provide a methodology for testing the hypothesis that fluctuations in the number of days with wind coming from warm/cold directions accounts for the higher temperatures in the UK from 1989 to 2002 and the long term warming trend in CET since 1881 and also that fluctuations in weather types explain EWR variations from 1881 to 2002. This methodology (Murray & Benwell, 1970)  involves constructing multiple regression models which predict CET and EWR from historical PSC indices and using them to predict CET & EWR from 1971-2002. These predictions can then be compared with observed CET and EWR to assess whether an increase of warm weather types account for the 1989-2002 warm period (and the long term warming trends) or whether there has been a warming of weather types themselves.
 

Testing for a warming of weather types using multiple regression models

UKCIP02  temperature predictions (30 year averages) for the UK  later this century are dramatic. For example, in the High Emissions scenario by the 2020s (2011-2030), the average  rise in annual  temperature in England is predicted to be 0.5 to 1.5 C relative to the 1961-90 average rising to 3.0 to 4.5 C by the 2080s (2071-2100).  The comparative figures for the Low Emissions scenario are 0.5 to 1.0 C and 1.5 to 2.5 C.  In the 343 CET record, while annual temperatures have varied from 6.8 C to 10.6 C, a range of  3.8 C,  30 year averages have only varied from 8.5 C to 9.8 C,  a range of  just 1.3 C.  If correct, these predictions imply that future correlations between PSC indices and CET will be different from those in Table 3 as future weather types would  have higher temperatures.

Multiple regression models between seasonal PSC indices (five combinations) and seasonal CET were developed using the data from the control period 1881-1970 and these models were used to predict seasonal temperatures from 1971 to 2002.  If  weather types are becoming warmer which logically they should if higher atmospheric concentrations of CO₂ are warming the atmosphere, the multiple regression models should progressively under-estimate CET from 1971-2002.

To assist in the interpretation of the results, Table 4 details the seasonal means of CET  for the control period and  for 1971 to 2002.  Main features to note are the higher means from 1971-2002 and the greater variability of winter temperature relative to the other three seasons. The UKCIP02 predictions mentioned below are derived from studying the UKCIP02 maps, focussing on the central England region and noting the full range of temperature anomalies from Low Emissions scenario to High Emissions scenario for both the 2020s and 2080s time periods. Ninety-five per cent  confidence intervals are presented using seasonal CET data from 1961-90 to test whether or not data from 1989 to 2002 show a warming which could not be expected by chance. The period 1961-90 is the baseline period used by climate change scientists and all predictions use it as the reference.
 

Table 4  Comparison of  seasonal CET means and standard deviations (sd) from 1881-1970 with 1971-2002.

Table 5    Summary of  results for winter  CET

Winter -  1961-90 average CET:  4.1 C                                                average CET 1989-2002:  5.0 C
                1961-90  CET  95% confidence interval (CI):  1.3 to 6.9 C    No. of years 1989-2002 above 95% CI:  0

UKCIP02  average temperatures:    2011-2040:   4.6 to 5.1 C               2071-2100:  5.6 to 7.6 C

Equation 1    winter CET = 2.802 + ( 0.01788 x winter P index) + (0.02344 x winter S index);

Winter (Table 5): temperatures in this season have risen 1 C  from 4.0 C in 1881-1970 to 5.0 C in 1989-2002 although none of these recent mild winters exceeded the 1961-90 95% confidence interval. The best multiple regression model (equation 1) for this season included the P and S indices  and accounted for 51.8 % of the variation in winter CET . Figure 13 shows the predicted values for both the control period and for 1971 to 2002 and compares them with observed winter CET  from 1881 to 2002.  The observed winter CET shows large annual variations as suggested by the higher standard deviations in Table 4 and also some long term trends with a warming bias from 1881 to around 1925 followed by a period of cooling  to 1970 and culminating in the warming up to the present day. The yellow line in Figure 13 shows observed winter CET minus predicted winter CET (i.e differences between observed and model predictions) which can be studied to assess two issues. First, of course, is the ability of the model to predict winter CET on an annual  basis. Second, a warming trend in the atmosphere logically should result in a tendency of the model to underestimate winter CET with time giving increasingly positive values. A visual inspection of Figure 13 shows that differences between observed and predicted winter CET have not steadily increased with time but have averaged around zero from about 1910 to 1987. Latterly, differences have been somewhat positive but are not anymore positive than differences at various times earlier in the record.

Conclusion: the 1989 to 2002 winter CET  average  is  5.0 C, 0.9 C above the 1961-90  average and is therefore consistent with the UKCIP02  predictions but it's somewhat high value can be explained  by the large variability of winter temperature. Figure 13 demonstrates that annual variations in weather types credibly explain this warmth rather than a warming of weather types.

Table 6    Summary of  results for spring CET

Spring - 1961-90 average CET:  8.3 C                                           average CET 1989-2002:  9.1  C
              1961-90  95% confidence interval(CI):  7.1 to 9.5 C          No. of years 1989-2002 above 95% CI:  5

UKCIP02  average temperatures:   2011-2040:   8.8 to 9.3 C         2071-2100:  9.8 to 11.8 C

Equation 2    spring CET = 7.9058 + (0.00715 x spring P index) + (0.02152 x spring S index) + (-0.01314 x spring C index);

Spring (Table 6): temperatures in this season have risen 0.8 C  from 8.3 C in 1881-1970 to 9.1 C in 1989-2002 and five of these recent warm springs exceeded the 1961-90 95% confidence interval. A regression model (equation 2) including all three PSC indices explained the most variation (45%) in spring CET.  Observed spring temperatures (Figure 14) from 1881 to 2002 were lowest at the beginning of the period, were high in the 1940s and 1950s, were lower from the 1960s to the mid 1980s and warmed in the 1990s to similar levels as those of the mid twentieth century. The predictions were clearly high during the cold  period while during the two aforementioned warm periods, the predictions were low suggesting that spring weather types have warmed and cooled several times from 1881-2002. The predictions from 1988-2002 are on average 0.9 C lower than the observed values but spring temperatures have fluctuated several times since 1881 and the observed values in the 1990s are no higher than those observed in the 1940s & 1950s. The difference (yellow line) between observed and predicted spring temperatures has grown noticeably more positive since 1975 which does indicate a clear warming of spring weather types.

Conclusion: the 1989 to 2002 spring CET  average  is  9.1 C, 0.9 C above the 1961-90  average and  five years between 1989 and 2002 were above the 95% confidence interval for 1961-90 which is remarkable.  Recent spring temperatures are therefore not only very high but also consistent with the UKCIP02  predictions  and Figure14 certainly shows some  warming of  spring weather types since 1975.
 

Table 7    Summary of  results for summer  CET

Summer - 1961-90 average: 15.3 C                                                    average CET 1989-2002:  16.0  C
                  95% confidence interval(CI):  14.0  to 17.0 C                     No. of years 1989-2002 above 95% CI:   1

UKCIP02  average temperatures:     2011-2040:  16.3 to 16.8 C        2071-2100:  17.3 to 20.3  C

Equation 3   summer CET=15.0974+(summer P index*-0.00302)+(summer S index x 0.01579)+(summer C index x -0.01066);

Summer (Table 7):  temperatures in this season have risen 0.8 C  from 15.2 C from 1881-1970 to 16.0  C in 1989-2002 and one of these recent warm summers exceeded the 1961-90 95% confidence interval (1995). In this season, the best multiple regression model (equation 3) included all three PSC indices and  accounted for 34.8 % of the variation in summer CET.  Observed summer temperatures (Figure 15)  have not varied very much but were higher from the 1930s to 1950s  and have been high in the 1990s. Four anomalously warm years have occurred since 1970: 1975; 1976; 1983 & 1995. The differences (observed minus predicted values)  were negative during the early cold period, were positive during the 1930s to 1940s  and have been positive 1989 but not to a greater extent than during the 1930s to 1940s.

Conclusion: the 1989 to 2002 summer CET  average  is  16.0 C, 0.7 C above the 1961-90  average and one year  (17.4 C in 1995) was above the 95% confidence interval for 1961-90 and a further two have exceeded  this threshold since 1970: 1976 & 1983.  These few exceptional summers  are perhaps indicative of  a warming trend  but the differences in  Figure 15 do not show  a clear warming trend suggestive of a warming of weather types.
 

Table 8    Summary of  results for autumn  CET

Autumn - 1961-90 average: 10.3 C;                                            average CET 1989-2002:  10.6  C
                 95% confidence interval(CI):  9.2 to 11.4  C                No. of years 1989-2002 above 95% CI:  1

UKCIP02 average temperatures:  2011-2040:  10.8 to 11.8 C     2071-2100:  12.3 to 15.3  C

Equation 4   autumn CET = 9.381+(autumn P index*0.00395)+(autumn S index x 0.02496)+(autumn C index x -0.00122);

Autumn (Table 8): temperatures in this season have risen 0.8 C  from 9.8 C from 1881-1970 to 10.6  C in 1989-2002 and one of these recent mild autumns exceeded the 1961-90 95% confidence interval (1995). A  multiple regression model (equation 4) including all three PSC indices gave the best result of all four seasons accounting for 60.5 % of the variation in autumn CET.  Figure 16 shows a long term warming trend in observed autumn temperatures: the 1990s were 1.2 C warmer than the 1880s. The predictions show a fairly close correspondence with observed temperatures until around 1980 when a marked and increasing tendency for predictions to be less than observed developed: on average 0.65 C below the observed from 1980 to 2002.

Conclusion: the 1989 to 2002 autumn  CET  average  is  10.6 C and one autumn (11.43 C in 1995) was above the 95% confidence interval for 1961-90 although two other years (1999 & 2001) were borderline with 11.40 C.  Figure 16 clearly shows that  weather types in autumn have increased  in temperature since around 1980 and such a trend  is what might be expected if higher concentrations of CO₂ are warming the atmosphere.
 

Testing for a change in rainfall intensity using multiple regression models

UKCIP02 predict annual rainfall 10% lower  than the 1961-90 average in England by the 2080s. However, much greater changes are predicted for individual seasons: while drier conditions are predicted for spring, summer and autumn with up to 50% reduction in summer by the 2080s under the High Emissions scenario, winters are expected to become up to 30% wetter by the 2080s under the High Emissions scenario. A number of publications have documented recent changes in rainfall (e.g Osborn & Hulme, 2002, Changing intensity of rainfall over Britain ): higher rainfall in winter, less in summer and an increase in the intensity of precipitation, especially in winter. A higher intensity of rainfall may expected because higher global temperatures should lead to greater evaporation from oceans, land  and water droplets in the atmosphere and therefore both a moister (more water vapour)  and warmer  atmosphere with a theoretical potential to give more intense rainfall in some regions.  The same methodology to that used for CET was employed to test for changes in rainfall as measured by the EWR. 

Multiple regression models were developed between seasonal PSC indices and seasonal EWR for the control period (1881-1970) and these were used to predict seasonal  EWR from 1971 to 2002. If an increase in rainfall intensity has recently occurred owing to a moister and warmer atmosphere, higher monthly totals of  EWR should logically result. Consequently, these models should progressively under-estimate EWR from 1971 to 2002 because seasonal rainfall totals would be larger than if there was no increase in rainfall intensity.

To assist in the interpretation of the results, Table 9 details the seasonal means of  EWR  for the control period and 1971 to 2001.  Main features to note are the high standard deviations in all seasons indicating very large annual variations and the trends to wetter winters, springs and autumns since 1970 and the trend to drier summers. The UKCIP02 predictions mentioned below are derived from studying the UKCIP02 maps, focussing on the central England region and noting the full range of rainfall anomalies from Low Emissions scenario to High Emissions scenario for both the 2020s and 2080s time periods. Ninety-five per cent confidence intervals are presented using seasonal data from 1961-90 to test whether data from 1989  to 2002 show a change in rainfall  which could not be expected by chance.

Table 9   Comparison of mean seasonal EWR and standard deviations (sd) from 1881-1970 with 1971-2001.

Table 10    Summary of  results for winter  EWR

Winter -  1961-90 average:   251 mm                                         average EWR  1989-2002:  269 mm
                95% confidence interval(CI):  110 to 392 mm              No. of years 1989-2002 above 95% CI:  2

UKCIP02  mean rainfall:  2011-2040:   251 to 276 mm        2071-2100:  276 to 326 mm

Equation 5   winter EWR = 255.37+(winter P index*0.1305)+(winter S index*0.575)+(winter C index*1.885);

Winter (Table 10): rainfall  in this season has  risen  23 mm (8%)  from 246  in 1881-1970 to 269 mm  in 1989-2002 and two of these recent wet  winters exceeded the 1961-90 95% confidence interval (1990 & 1995). Although Table 3 shows that C index in all seasons is highly correlated with EWR, a multiple regression model (equation 5) including all three PSC indices accounted for the most variation in winter EWR (68.4%). Observed values of winter EWR from 1881-2002 (Figure 17) show no long term trend but large year to year variations. The predicted values show a very close correspondence  with the observed values throughout with only 1990, 1994 and 1995 being grossly under-estimated.. These winters  were all exceptionally wet but 1915 & 1916 were both equally wet. The higher average from 1989-2002 owes to the three very wet winters (1990; 1994 & 1995) rather than consistently wet winters in this period. The differences did not show a trend to support the theory that a warmer and moister atmosphere is resulting in heavier rainfall unless it is argued that the three very wet winter are the result of this effect. As all other recent winters have had average or below average rainfall and were well predicted on the basis of weather types, such an argument is not persuasive.

Conclusion: the 1989 to 2002 winter  EWR  average  is 269 mm  and two winters  (1990 &  1995) were above the 95% confidence interval for 1961-90 which is significant. Figure 17 shows neither a long term trend to increased winter rainfall  nor a tendency to underestimate winter EWR. The average higher rainfall of winters since 1989 is largely the result of  three very wet winters and annual variations in winter rainfall are credibly explained by fluctuations in weather types.

Table 11    Summary of  results for spring EWR

Spring -   1961-90 average:  201 mm                                            average EWR  1989-2002:  191 mm
                95% confidence interval(CI):  99 to 303 mm                   No. of years 1991-2002 below 95% CI:  0

UKCIP02  mean rainfall:  2011-2040:  181 to 201  mm                  2071-2100:   161 to 181 mm

Equation 6   spring EWR = 207.55+(spring S index*0.269)+(spring C index*1.252);

Spring (Table 11): rainfall  in this season has  risen  7 mm from 184  in 1881-1970 to 191 mm  in 1989-2002 defying the expected trend to drier springs. However, the 1989-2002 average is 10mm less than the 1961-90 average supporting the expected drier trend. The best multiple regression model included the S & C index (equation 6) and accounted for 50% of the variation in spring EWR.  Figure 18 shows no long term trend in observed spring EWR 1881 to 2002 and there is generally a very good correspondence between the predicted and observed values. The differences did not show a trend to support the theory that a warmer and moister atmosphere is resulting in heavier rainfall.

Conclusion: the 1989 to 2002 spring  EWR  average  is 191 mm  which is a little higher than the 1881-1970 average but a little lower than the 1961-90 average. Figure 18 shows neither a long term trend to decreased spring  rainfall  nor a tendency to underestimate spring  EWR. As in winter, annual variations in spring rainfall  are credibly explained  by variations  in weather  types.
 

Table 12   Summary of  results for summer EWR

Summer -   1961-90 average:  205 mm                                                average EWR  1989-2002:  194 mm
                    95% confidence interval(CI):  100 to 310  mm                   No. of years 1991-2002 below 95% CI:  1

UKCIP02    mean rainfall:  2011-2040:  165 to 205  mm                       2071-2100:   102 to 165 mm

Equation 7   summer EWR = 258.15+(summer P index*-0.305)+(summer C index*1.721);

Summer (Table 12): rainfall  in this season has  fallen  30 mm from 224 mm  in 1881-1970 to 194 mm  in 1989-2002 and one summer, 1995 had rainfall below the 1961-90 95% confidence interval. A multiple regression model (equation 7) including the P and C indices accounted for most (63.4%) variation in summer EWR.  Summer rainfall  from 1881 to 1960 (Figure 19) showed no long term trends but subsequently summer rainfall has been lower with some notably dry summers: 1975; 1976; 1983; 1984 & 1995. Throughout 1881-2002, predicted values show a close correspondence with observed values and the lower rainfall since the 1960s is credibly  explained by variations in weather types with no suggestion that the models are tending to underestimate summer EWR since the 1960s. This means the reduction in summer rainfall owes to changes to drier weather types (i.e anticyclonic types).

Conclusion: the 1989 to 2002 summer  EWR  average  is 194 mm  which is lower  than both the 1881-1970 average and the 1961-90 average. This is  consistent with the UKCIP02 predictions which predict up to 50% less summer rainfall under the High Emissions scenario by the 2080s and the five notable dry summers of recent years also support UKCIP02: 1975; 1976; 1983; 1984 & 1995.  Figure 19 shows that this reduced summer rainfall was credibly explained by a change to more anticyclonic weather types.

Table 13   Summary of  results for autumn EWR

Autumn -   1961-90 average:  258 mm                                             average EWR  1989-2002:  295 mm
                   95% confidence interval(CI):  115  to 401  mm               No. of years 1991-2002 below 95% CI:  0

UKCIP02   mean rainfall:  2011-2040:  233 to 258  mm                    2071-2100:   208  to 258 mm

Equation 8       autumn EWR = 308.10+(autumn P index*-0.177)+(autumn S index*-0.352)+(autumn C index*1.832);

Autumn (Table 13): rainfall in this season has risen 30 mm from 265 mm in 1881-1970 to 295 mm in 1989-2002 defying the predicted trend to drier autumns. Indeed, autumn 2000 was so exceptionally wet that it exceeded the 95% 1961-90 confidence interval. The best multiple regression model (equation 8) included all three PSC indices and accounted for 60.5% of variations in autumn EWR. Observed values (Figure 20) were low from 1887 to 1922 but there is no obvious  long term trend indicating that the high 1989-2002 average largely owes to the exceptionally wet autumn of 2000. There is a very close correspondence between observed and predicted values indicating that annual differences in  weather types credibly explain variations in autumn rainfall.

Conclusion:  the 1989 to 2002 autumn EWR average is 295 mm which is higher than both the 1881-1970 average and the 1961-90 average but this is largely accounted for by the exceptionally wet autumn 2000. There is in fact no trend to wetter or drier autumns. As in all other seasons, annual variations in autumn rainfall are credibly explained  by variations  in weather  types.

Classification of individual months and seasons

Murray & Benwell (1970) calculated PSC indices using the original Lamb types and determined quintile boundaries for months and seasons using indices for the 100 year period 1869-1968. A month's or season's classification can therefore be summarised as P_xS_xC_x where x is a quintile. Quintile 3 is therefore average, quintiles 4 & 5 above average and quintiles 1 & 2 below average.  Experience has shown that the PSC indices are very useful as parameters for specifying large scale anomalous circulations which cause long periods of unseasonable weather in the British Isles (Murray & Benwell, 1970).

Monthly and seasonal quintile boundaries were calculated using the PSC indices from 1881-1970 (see appendix for quintile boundaries) so that individual month's or season's  PSC indices can be allocated to quintiles. To illustrate the advantages of PSC indices in classifying anomalous weather in the UK,  two cold winters (1962/63 and 1978/79) and two hot summers (1976 & 1995) of the twentieth century are discussed.

Winter 1962/63   The classification of this winter which was the coldest of the twentieth century was P₁S₁C_1.   P quintile 1 means that westerly winds were unusually rare and easterly winds dominant. The low quintile for the S index  indicates that northerly winds were much more frequent than normal while the low quintile for the C index indicates that anticyclonic weather types were very dominant.  This combination of indices indicates that anticyclones were often over or to the north west, north or north east of the UK bringing frequent north, north easterly and easterly winds.  The CET anomaly from the 1961-90 average was -4.4 C and the percentage of average winter  EWR was 57% which is what would be expected from a winter with dominant northerly and easterly winds and very anticyclonic conditions.

Winter 1978/79   The classification of this winter  was P₁S₄C₄ which indicates that westerly winds were again unusually rare and easterly winds dominant but the quintile 4 of the S index indicates that southerly winds were more common than usual and the quintile 4 of the C index indicates cyclonic weather types were dominant.  This combination of indices indicates that  winds from southerly, south easterly and easterly points prevailed and the weather was often cyclonic with the low pressure near to or  just south or south west of the UK..  The CET anomaly from the 1961-90 average was -2.5 C and the percentage of average  winter  EWR was 135% which is what would be expected from a winter with dominant easterly winds and  cyclonic conditions.

Summer 1976  The classification of this summer is P₁S₄C₁ which indicates an absence of westerly winds and dominance of southerly winds together with very anticyclonic conditions. This combination of indices  mean warm airstreams off the continent and long hours of sunshine resulted in temperatures higher than average. The CET anomaly from the 1961-90 average was +2.4 C and the percentage of average of  summer  EWR was 38% which are consistent with the very anticyclonic conditions of this summer.

Summer 1995  The classification of this summer is P₁S₃C₁ which indicates a predominance of easterly winds and very anticyclonic conditions. As in 1976, warm winds off the continent and long hours of sunshine resulted in a hot summer. The CET anomaly from the 1961-90 average was +2.0 C and the percentage of average of  summer  EWR was 37% which are consistent with the very anticyclonic conditions of this summer.

In the following section, this method of classification will sometimes be used.
 

The 1989-2002 warm period in more detail

Winters  The main features of the half monthly PSC indices from 1989-2002  was that Februaries had much higher P values (Figure 10), late Februaries had higher C values and the first half of Decembers had much lower C values (Figure 12). The consequences of these features are that in Februaries 1989-2002  temperatures  were 1.5 C above the 1961-90 average and February rainfall was 123% of the 1961-90 average. The comparative figures for Decembers and Januaries were 0.0 C and 104%, and 1.0 C & 98% respectively resulting in milder and wetter winters overall.  Figure 13 showed that variations in PS indices from 1989 to 2002 largely explained the higher than average temperatures of  most of these winters (0.9 C above 1961-90 average; Table 5). Indeed, the difference between the observed and predicted values was less than 0.7 C in 57% of  of the years 1881 to 2002 which is remarkable considering the relatively high variability of winter temperature (Table 4).

Figure 21 shows the PSC indices of the winters 1989-2002 allocated to quintiles using the boundaries calculated for the 1881-1970 period. Only 16% of the 42 PSC indices (3 indices x 14 years) were average (i.e quintile 3) indicating that the weather patterns were  highly anomalous during this period.  Fifty per cent of the 14 years had above average P indices (i.e quintiles 4 & 5), 36% above average S indices and 43% above average C indices. The comparative figures for below average PSC indices (i.e quintiles 1 & 2) were 36%, 36% and 50%. Clearly there was no overall consistency in the weather types of this period  Table 3 shows that the P index in winter is highly correlated with winter CET from 1881 to 2001 and Table 14 compares winter PSC indices in quintiles 1989-2002 with winter CET in terciles where both quintile and tercile boundaries were calculated using data 1881-1970. In Table 14, a diagonal distribution of positive numbers would suggest that the index in question was very influential in determining winter temperature. It is apparent from Table 14 that the contingency table for the P index has this diagonal distribution of positive values and therefore it can be concluded that the high temperatures  of the winters 1989-2002 were associated with high P indexes i.e the anomalous warmth owed to much higher than average number of days with westerly winds.

Table 14   Contingency tables comparing PSC indices in quintiles with CET in terciles for the winters 1989-2002

                                            -   denotes zero

Winter rainfall 1989-2002  was somewhat above average and  two winters exceeded the 1961-90  95% confidence interval (Table 10) and UKCIP02 do predict an increase in winter rainfall of up to 30% by the 2080s under the High Emissions scenario. However, Figure 17 shows that in the context of the 1881-2002 period, the 1989-2002 period is not obviously wetter. Indeed, some of the winters were very dry (1989: 167 mm; 1992: 132 mm; & 1997: 172 mm),  the higher mean mainly owes to three  very wet winters (1990: 416 mm; 1994: 371 mm & 1995: 407 mm) two of which exceeded the 95% confidence interval. Figure 22 shows that annual variations in winter rainfall are highly correlated with winter C indices as indicated by the high correlation coefficient (r=0.794) shown in Table 3.  Figure 6 shows that the 1990s did not have a high mean C index confirming that these winters did not have consistently high C indices.  Figure 21  shows only one winter from 1989-2002 with  a C index of quintile 3 whereas six years had above average quintiles and seven years below average quintiles emphasising the remarkable variability in winter cyclonicity and rainfall during this period.

Osborn & Hulme (2002) examined daily precipitation data from 1961-2001 from 100 stations across the UK and found an increase  in total winter precipitation across nearly all the UK, an increase in wet days (>0.4mm) in western regions and a increase in the mean amount of rain on each wet day across the whole UK. They also categorised rainfall events and found that for the heaviest rainfall category, the contribution to total winter  rainfall increased significantly from low values prior to 1975 to very high values around 1990 although they have decreased somewhat since. Their conclusion was that in relation to the entire twentieth century, there is an indication that recent winter increases in heavy precipitation are unusual but the sparseness of the longer observational records leads to reduced confidence in the extended results.

As to the causes of these winter rainfall increases, Osborn & Hulme (2002) point out that these changes are consistent with model simulations like UKCIP02 but also that they may simply be part of the multi-decadal climate fluctuation, the North Atlantic Oscillation (NAO) which has been in its positive phase recently. In addition, they argued that the NAO may itself be influenced by anthropogenic climate change.  Figure 22 clearly shows that annual variations in winter EWR are highly correlated with the winter C index  so it logically follows that there must be an increase in winter cyclonicity since 1960 to account for the findings of Osborn & Hulme (2002). However, neither winter EWR nor winter C indices show an increase since 1960 (Figure 22) and Figure 6 surprisingly shows winters becoming increasingly anticyclonic in recent decades. This conundrum is solved by reference to the suggested cause of the increased rainfall put forward by Osborn & Hulme (2002), the NAO. While there is no significant correlation between winter NAO and winter C index 1881-2001 (r = -0.025), there is a very strong correlation between winter NAO and winter P index (r = 0.799) and high P indices have been a feature of winters 1989-2002 with 64% of  these years having average or above average P quintiles (Figure 21 & Table 14). Table 3 shows that there is no significant correlation between winter EWR and winter P index which accounts for the lack of an increase in winter EWR since 1960. Osborne & Hulme (2002) analysed rainfall data from the whole UK and not just central England and found that the greatest increases in both total precipitation, wet days and mean rainfall on each wet day occurred in the more northern and western parts of the UK which is consistent with the increase in winter P index (i.e greater number of days with westerly winds) since 1960 (Figure 6). Murray & Benwell (1970) showed that  the P index is significantly correlated with rainfall in Scotland in all months whereas this index was only weakly correlated with EWR in February, September and November which accounts for the dramatic increases in winter rainfall in Scotland while winter rainfall in England has been much more variable from 1989 to 2002 with some very wet winters and some very dry ones. As to whether or not the recent increase in winter NAO/winter P index is natural, this can be statistically assessed with reference to the 95% confidence interval of the winter P index constructed with data from 1881 to 1970 and to note whether winters in recent decades have winter P indices above the 95% confidence interval. The mean winter P index 1881 to 1970 is 37.4, the standard deviation is 44.2, giving a 95% confidence interval of -49.1 to 123.9. There is no evidence that the high winter P indices during the 1980s and 1990s are unnatural with only one year (1989) exceeding the 95% confidence interval  (Table 15).

Table 15  Number of years each decade with winter P  index above the 95% confidence interval

Winter Conclusion:  the notably mild winters of 1989-2002 are credibly explained by an increase in the P index (i.e westerly weather types) which is within natural variation and there is no long term trend indicating a warming of weather types in winter although  1998 to 2002 were underestimated by the model (Figure 13). The three very wet winters contrary to expectation did not have exceptionally  high C indexes but had a combination of both high P & C indexes (1990: P₅S₅C₄ 1994: P₃S₄C₅ & 1995: P₅S₃C₄)  but there was no long term trend to wetter winters from 1881 to 2002 as there were some very dry winters from 1989 to 2002  (Figure 17).
 

Springs  The main features of the half monthly PSC indices from 1989-2002 were that Marches have high P values and Mays low P values (Figure 10) while early Marches and the whole of Aprils were more cyclonic (Figure 12). The consequence of these features is that in Marches 1989-2002 temperatures  were 1.4 C above the 1961-90 average but March rainfall was just 83% of the 1961-90 average in spite of the cyclonic nature of early Marches. The comparative figures for Aprils and Mays were 0.5 C and 121%, and 0.7 C & 83% respectively resulting in milder and drier springs overall.  Figure 14 shows that variations in PSC indices from 1989 to 2002 cannot explain the higher than average temperatures of  most of these springs (0.8 C above 1961-90 average; Table 6),and there is a clear tendency for the model to increasingly underestimate spring CET from the mid 1970s onwards underlining the fact that a factor other than annual fluctuations in weather types is controlling spring temperatures in recent years.

Figure 23 shows the PSC indices of the springs 1989-2002 allocated to quintiles using the boundaries calculated for the 1881-1970 period. Forty per cent of  the 42 PSC indices (3 indices x 14 years)  were average (quintile 3) in spring which compares with only 16% in winter indicating that spring weather patterns were much closer to average compared with the highly anomalous  winter weather patterns from 1989-2002.  Thirty six  per cent of the 14 years had above average P indices (i.e quintiles 4 & 5), 14% above average S indices and 36% above average C indices. The comparative figures for below average PSC indices (i.e quintiles 1 & 2) were 36%, 29% and 36%. Clearly there was no overall consistency in the weather types of this period. Table 3 shows that all three PSC indices  in spring are correlated with spring CET from 1881 to 2001 and Table 15 compares spring  PSC indices in quintiles 1989-2002 with spring  CET in terciles where both quintile and tercile boundaries were calculated using data 1881-1970. It is apparent from Table 16  that there were no clear associations between CET in terciles and PSC indices in quintiles, probably because 12 of the 14 years were CET  tercile 3. The rainfall in springs 1989-2002 was not noteworthy.

Spring Conclusion: five of the 14 springs 1989-2002 were above the 1961-90 95% confidence interval which demonstrates that spring temperatures have been significantly higher than in the 1961-90 period. The model based on spring PSC indices consistently underestimated spring temperatures from 1989-2002 and Table 15 also shows that PSC indices do not  account for this warmth. There is strong evidence of a warming of spring weather types since the mid 1970s indicative of some sort of climate change which cannot be explained on the basis of a change in the incidence of cold/warm weather types.
 

Table 16  Contingency tables comparing PSC indices in quintiles with CET in terciles for the springs 1989-2002

                                            -   denotes zero

Summers   The main features of the half monthly PSC indices from 1989-2002  were that Junes to early Julys had low S indices (northerly winds higher than average) while late Julys to Augusts had high S indices (southerly winds higher than average) (Figure 11) and  early Junes had high C indices (cyclonic days higher than average) and late Julys to Augusts low C indices (anticyclonic days higher than average) (Figure 12).  The consequences of these features is that in Junes 1989-2002  temperatures  (0.1 C above the 1961-90 average) and rainfall (101% of 1961-90 average)  were average in contrast to Julys and  Augusts  when temperatures were higher and rainfall was lower than average (comparative figures for July and August  were 0.7 C and 93%, and 1.0 C & 91% respectively) resulting in warmer and drier summers overall.  Figure 15 shows that variations in PSC indices from 1989 to 2002 credibly explain the higher than average temperatures of  most of these summers (0.7 C above 1961-90 average; Table 6) and there is no clear trend in the observed minus predicted differences to suggest a warming of weather types in summer.

Figure 24  shows the PSC indices of the summers 1989-2002 allocated to quintiles using the boundaries calculated for the 1881-1970 period. Only 29% of  the 42 PSC indices (3 indices x 14 years)  were average (quintile 3) in summer  indicating that weather patterns in these summers were highly anomalous.  Fifty  per cent of the 14 years had above average P indices (i.e quintiles 4 & 5), 36% above average S indices and 22% above average C indices. The comparative figures for below average PSC indices (i.e quintiles 1 & 2) were 36%, 22% and 50%.  The two statistics relating to 50% of years 1989 to 2002 are now examined: 50% with above average P indices and 50% with below average C indices. Table 3 shows that both the P and C indices  in summer are negatively correlated with summer CET and  that the correlation for the C index was much stronger than that for the P index. The implication of the higher than average P indices in 50% of summers 1989-2002 is that temperatures would be lower than average but this was clearly more than cancelled out by the implication of the lower than average C indices (more anticyclonic than average)  in 50% of these summers which meant these summers were 0.7 C above the 1961-90 average (Table 7). Table 3 shows that all three PSC indices  in summer  are correlated with summer CET from 1881 to 2001 and this is supported by Table 17 which compares summer  PSC indices in quintiles 1989-2002 with summer CET in terciles where both quintile and tercile boundaries were calculated using data 1881-1970. All three coolest summers (CET tercile 2) had high P indices, average or below average S indices and average or above average C indices while the majority of  summers (tercile 3) had average or above average S indices and average or below average  P and C indices. Clearly, the warmth of these 1989-2002 summers owed to high than average incidence of warm weather types.

Table 17   Contingency tables comparing PSC indices in quintiles with CET in terciles for the summers 1989-2002

                                             -   denotes zero

UKCIP02 predict up to a 50% reduction in summer rainfall by the 2080s under the High Emissions scenario which is a truly alarming statistic as it would result in central England receiving average summer rainfall of only 102 to 165 mm (Table 12). The 1989-2002 average summer EWR is 194 mm which is 5% less than the 1961-90 average and 13% less than the 1881-1970 which strongly suggests a trend to drier summers. Figure 19 does clearly show both a lower average rainfall since 1970 and the occurrence of  four very dry summers: 1975; 1983; 1984 & 1995 but it also shows that this reduction in summer rainfall is the result of a reduction of cyclonic weather types and an increase in anticyclonic weather types. The anticyclonic nature of the 1970s, 1980s and 1990s is  also apparent in Figure 8 and Figure 25 shows the very close correspondence between annual summer C index and summer rainfall  as suggested by the correlation coefficient (r=0.772; Table 3).

Osborn & Hulme (2002) examined daily precipitation data from 1961-2001 from 100 stations across the UK and found a reduction in total summer precipitation, a reduction in wet days (>0.4mm) and a reduction in the mean amount of rain on each wet day across the UK except the west of Scotland. They also categorised rainfall events and found that for the heaviest rainfall category, the contribution to total summer rainfall decreased from high values prior to 1973 to lower values since with a slight recovery during recent summers. Their conclusion was that these recent decreases in summer rainfall may not be unusual in the context of the whole twentieth century although the sparseness of the longer observational records leads to reduced confidence in the extended results. As to the causes of  these recent changes in summer rainfall, Osborn & Hulme (2002) did not offer any suggestions. Figure 25 which demonstrates that annual variations in summer EWR are highly correlated with the  summer C index strongly  suggests that decreases in summer rainfall since 1960 are the result of  fewer cyclonic Lamb type days and more anticyclonic  Lamb type days. This could account for all the rainfall reductions noted by Osborn & Hulme (2002) as cyclonic Lamb type days (which are days when a depression centre is located over or very near to the UK) often produce  heavier rainfall  than directional cyclonic days when the centre of the depression is further away from the UK.  This means  that the  more anticyclonic summers of  recent years have lower precipitation totals, fewer wet days (>0.4mm) and lower rainfall totals on those wet days.  As to whether or not this change to more anticyclonic summers is the result of climate change caused by higher concentrations of CO₂ in the atmosphere, the UKCIP02 scenarios (Table 12) do predict drier summers. However, Figure 8 shows that during the period 1881-1909, summers were more anticyclonic than during the 1980s and 1990s and Figure 25 shows that summer rainfall totals at this time were often lower than average with some notable dry summers (1885: 148 mm; 1887: 112 mm  & 1899: 133 mm) which are comparable with two of the four very dry summers of recent times (1983: 107 mm & 1984: 128 mm) but not with the two driest summers (1976 &  1995: both  76 mm). This increase in anticyclonicity since 1960 can be assessed statistically using 95% confidence intervals constructed using summer C indices 1881 to 1970. With a mean of -16.3 and standard deviation of  26.7, the 95% confidence interval for summer C index is -72.5 to 40.0.  Table 18 compares  the number of years in each decade 1880s to 1990s which had summer C index below the 1881-1970 95% confidence interval. There are no years since 1960 with summer C index below the 95% confidence interval so the change to more anticyclonic summers can be considered natural.

Table 18  Number of years each decade with summer C index below the 95% confidence interval

Summer Conclusion:  the warmer summers 1989-2002  are credibly explained by an increase in warm weather types, in particular an increase in anticyclonic and southerly weather types and there is no clear trend indicating a warming of weather types in summer. The decrease in summer rainfall 1989-2002 and the longer term decrease since 1960 can simply be explained by the increase in anticyclonicity which is comparable to a long period of anticyclonicity which prevailed from 1881-1909 which also contained some very dry summers.

Autumns:  The main features of the half monthly PSC indices from 1989-2002 are that Septembers had low P indices (Figure 10), late Octobers had high S indices (Figure 11), but the most notable difference is that all six half months had above average C indices (Figure 12). The consequences of these features is that in Septembers 1989-2002  temperatures  were 0.4 C above the 1961-90 average and rainfall was 107% of the 1961-90 average. The comparative figures for October and November were 0.1 C and 123%,  and 0.6 C and 112% respectively giving milder and wetter autumns overall. Figure 16 shows that from 1881 to 1977, variations in autumn temperatures were credibly explained by the PSC model but since then there has been a very obvious tendency for the models to underestimate autumn CET. This seems to be a long term increasing trend which strongly suggests autumn weather types are warming. Figure 9  shows that mean decadal values of autumn P indices have been falling sharply since the 1960s but there is no significant correlation between autumn CET and autumn P index (Table 3) or indeed between monthly CET and monthly P index in the autumn (Sep: r = 0.089;  Oct: r = 0.115;  Nov:  -0.111).

Figure 26  shows the PSC indices of  the autumns 1989-2002 allocated to quintiles using the boundaries calculated for the 1881-1970 period. Thirty six per cent  of  the 42 PSC indices (3 indices x 14 years)  were average (quintile 3) in autumn  indicating that weather patterns in these autumns were somewhat anomalous.  Only  22% of the 14 years had above average P indices (i.e quintiles 4 & 5) compared with  50% of  years for S indices and C indices. The comparative figures for below average PSC indices (i.e quintiles 1 & 2) were 43%, 22% and 7%.  Table 19  compares autumn  PSC indices in quintiles 1989-2002 with autumn CET in terciles where both quintile and tercile boundaries were calculated using data 1881-1970 but as 12 of the 14 years were autumn CET tercile three, Table 19 is not very helpful in accounting for the warms autumns from 1989 to 2002.  One characteristic which Table 19 does show which has already been mentioned is the high C indexes in this period with all but one of the 14 years having average or above average C quintiles. Table 3 shows that there is no significant correlation between autumn C index and autumn CET and there is also no significant correlation between monthly CET and monthly C index in the autumn (Sep: r = -0.041;  Oct: r =  0.097;  Nov:  r = 0.04).  So it can be said with confidence that variations in weather types cannot account for the warming of autumn CET as indicated by  Figure 16 so there has to be another factor which is driving this warming. There is a highly significant positive correlation between autumn S index and autumn CET  (Table 3) but there is no long term trend in decadal averages of S index (Figure 9) that could account for the warming trend.
 

Table 19   Contingency tables comparing PSC indices in quintiles with CET in terciles for the autumn 1989-2002

                                             -   denotes zero

With respect to autumn rainfall (Figure 20), autumns were noticeably drier from 1881 to 1920 and this not surprisingly corresponds with greater anticyclonicity at this time (Figure 9).  UKCIP02 predict drier autumns (Table 13) but autumn rainfall 1989-2002 was above average. In the context of rainfall from 1920 onwards, rainfall from 1989 to 2002 looks close to average (Figure 20)  with the higher mean largely owing to the exceptionally wet autumn of  2000 (P₄S₄C₅)  which was well predicted by the model showing that weather types account for the exceptionally high rainfall of this autumn. Obviously the associated flooding during autumn 2000 was  highly disruptive and economically damaging  but it appears to be a one off event and not part of a trend to wetter autumns. Osborn & Hulme (2002)  also did not find a coherent trend to wetter autumns in their study of UK wide  rainfall.

Autumn Conclusion:  the warmer autumns 1989-2002  can only  be partially explained by variations in weather types as there is a strong  trend indicating a long term warming of weather types. Although autumn 2000 was unquestionably an exceptional and worrying rainfall event, there is no evidence of  increasing rainfall in autumn.
 

DISCUSSION

Climate Change in the UK

The introduction to the UK Climate Impacts program begins "The Earth's climate has been relatively stable since the end of the last ice age (about 10,000 years ago) but is now changing".  Some evidence is provided to support the statement that climate is now changing. Climate change scenarios specific to the UK have been developed by the Tyndall Centre for Climate Change Research at the University of East Anglia and the Hadley Centre for Climate Prediction and Research at the Meteorological Office. These UKCIP02 scenarios unequivocally state "the UK will become warmer".  It is the scientific consensus of  these meteorological organisations that the UK will warm and apparently cooling is not an option for the UK climate during the twenty first century.

Temperature Changes
The CET record from 1659 (annual data: Figure 1; seasonal data relative to 1961-90 averages: winter & summer  Figure 27;  spring & autumn  Figure 28) certainly lends support to the theory that the UK climate is changing and that it is warming but clearly change is not a recent characteristic of the CET, in fact it has changed throughout the entire 332 year record and the current rise in the 30 year running mean is far less dramatic than the rise from 1700 to 1735. Table 20 confirms the long term warming trend and shows for example that the 30 year running mean rose from 8.5 C in 1700 to 9.7 C in 2000. Regression models where the four start of century temperature measurements are regressed on year (i.e time) may be used to predict the CET in 2100  and these predictions are also presented in Table 20. These predictions give an indication of CET in 100 years time assuming the UK continues to warm at the same rate as over the last 332 years. UKCIP02 are predictions of  30 year averages and so are comparable with the 30 year running mean prediction shown in Table 20. For the 2080s (2071-2100),  temperatures in England under the Low Emissions Scenario are expected to be in the range 11.0 to 12.0 C while the comparative figures for the High Emissions Scenario is 12.5 to 14.0 C,  all much higher than the 10.0 C predicted for 2100  in Table 20. Even the lowest range for the Low Emissions Scenario is double the warming rate of the last 300 years in which  Europe has emerged from the Little Ice Age.
 

Table 20   Comparison of observed values of  annual Central England Temperature in the first year of each century and the predicted values for 2100 based on three regression models (temperature regressed on year) all of which were significant.

On the basis of the 332 year CET record, there is no reason to argue with the UKCIP02 claims that temperatures over the next 100 years will rise (mean increase in 30 year running mean per century is 0.4 C), but the extreme rises predicted by UKCIP02  appear incredible. A method of assessing whether or not such  rises are attainable is to calculate  the rate of annual change in 30 year running means of average annual temperature (current year's 30 year running mean  minus previous year's running mean) in recent years and compare them with the average annual rate of increase required to achieve these predictions in 2100.   Table 21 compares such data from five UK sites and one in Ireland and shows that on the basis of  the mean  increase in 30 year average per year 1990-2002 for the six sites (0.022 C per year), a rise of  2 C would be expected but the  rates in a few years have exceeded the rates required to exceed the more extreme predictions. So clearly, the low to mid ranges of the UKCIP02  scenarios are attainable on the basis of recent rates of warming. However, these high rates of increase from 1990-2002 are less than the rates which prevailed for a while in the eighteenth century for a much longer period (mean 1703-1739: 0.026 C per year).

Table 21   Comparison of rates of increase in 30 year running means of annual temperature from five sites in the UK and one in Ireland
(Data source:  Goddard Institute)

*  for example, to exceed a 1 C rise from 2000 to 2100, the average increase in the 30 year average per year (calculated by subtracting previous year's 30 year running mean from current year's running mean) must exceed 0.01 C per year

**  30 year average only available for 12 years compared with 13 years data from all the other sites.

However, on the basis of the 332 year CET record, temperatures do not increase relentlessly over periods of 100 years but the long term increase is interrupted at times by periods of decreasing temperatures. Indeed,  there are six occasions in the CET when the 30 year running mean decreased for 20 years or more  (Figure 1)  with the most recent occasion being the 1960s to 1980s showing that temperatures  have fluctuated above and below the long term average albeit with a long term warming trend. Understanding the factor which accounts for these fluctuations is very relevant to the global warming/climate change debate about future UK climate.

The North Atlantic Osillation (NAO) is acknowledged as a factor that controls temperatures in Europe, particularly in winter (Rodwell, Rowell & Folland, 1999;  NAO; data) and currently there is a collaborative research project between European meteorological organisations investigating the predictability of the NAO (Predicate). The correlation between December to March NAO and CET in the subsequent year is significant at the 0.1% level (r = 0.410) indicating that winters with high NAO result  in years with above average CET. This relationship shown in Figure 29 does not account for a high proportion of  the annual variations in CET but it is a major contributory factor. The relationship is stronger (r = 0.536) and more persuasive  if a sample of meteorological stations across Europe is obtained and a European temperature average constructed (Figure 30). The position of the UK on the western fringe of Europe where the warming Atlantic influence is so dominant, even in negative NAO winters, probably accounts for the poorer relationship between NAO and temperature in central England than in  Europe. Table 22 shows the meteorological stations used to construct the European mean and the annual and seasonal temperature trends can be viewed by clicking on the hyperlinks. 
 

Table 22  Ten climatological stations around Europe: graphs showing anomalies from the 1961-90 average annual temperature and annual and seasonal values of mean temperature.  (Data Source: Goddard Institute.)

Judging from Figure 30, European temperatures will fall should winter NAO turn negative and winter NAO has indeed been slightly negative since 2001 but but both 2001 and 2002 have been remarkably warm years in Europe. The NAO in winter 2002/2003 has been predominantly and at times highly negative (data) and winter temperatures have been lower in Europe than in recent years  perhaps heralding a new phase of negative NAO. This article expresses the view that this may be the case. If  a new phase of  negative NAO has commenced  lower temperatures in Europe for the next few decades is a possibility because the NAO is a multi-decadal oscillation. UKCIP02 do not appear to acknowledge this as a possibility judging from their  predictions. Climate model simulations suggest  and  some scientists believe that global warming may favour the positive phase of the NAO (Osborn & Hulme, 2002).

A valid argument against the high  rates of increase in 30 year running means (Table 21) being sustained is that if the 14 year 1989-2002 period is split in half and the temperature of the two seven year periods compared, there is only a difference of  0.1 C (mean annual CET: 1989-1995:  10.11 C;  1996-2002:  10.21 C, monthly anomalies in Figure 2) which is an indication that temperatures are not relentlessly increasing year after year but rather have settled at a higher level dictated by the positive phase of the NAO. So logically, even if the climate model simulations are correct in their controversial predictions that global warming would mean higher NAO (i.e mild westerly weather type winters in most years), the 30 year running mean will eventually stop rising when it reaches the 1989-2002 mean of  10.2 C.  The high rates of  temperature increase from 1989 to 2002 are the result of the change to the positive phase in the NAO in 1989 which will not apply to any future decade of the twenty first century unless a new phase of negative NAO (which would cause cooling) intervenes and ends. Also, this study found no evidence of a warming of winter weather types which again underlines the dependence of the high rates of  increase in 30 year running means from 1989 to 2002 on the commencement of the positive phase of winter NAO in 1989.

With respect to the NAO, it is relevant to try to assess  its contribution to the total warming of the 1989 to 2002 warm period as Table 21 shows that temperatures have been  rising unusually rapidly during this time. Marches in addition to January and Februaries  have had high P indices (Figure 10) and Marches 1989 to 2002 had temperatures 1.4 C above average showing that some of the spring warming since 1989 can also be attributed to the positive NAO.  There is also an association between  the  P index in winter and the C index in the following summer.  Although the correlation between winter P index and summer C index is not significant (r = -0.157), Table 23 shows that the 47 winters 1881 to 2002 with above average P indices (quintiles 4 or 5)  were followed by summers of which 70% had average or below average C indexes (quintiles 1, 2 or 3). Figure 12 shows the anticyclonic nature of the late summers 1989 to 2002 and Table 3 shows that temperatures in summer are highly correlated with the C index: anticyclonic conditions favouring high summer temperatures. As a result of the higher summer temperatures of 1989 to 2002 also being associated with the high P indices of the winters, it is probably the case that well over half the warming 1989 to 2002 can be attributed to the positive phase of the NAO  and when a new phase of the negative NAO commences, much of  this warming will be reversed.
 

Table 23    Distribution of  summer C quintiles following winters 1881 to 2002 with high P quintiles (4 or 5).

If  CET seasonal data from 1881 to 2002 are examined,  it is apparent that the long term warming trend in  annual temperatures largely owes to warming in spring and autumn (Figure 28). While from 1989 to 2002, both winter and summer temperatures were higher than average, there is no apparent warming trend from 1881 to 1988 in these two seasons (Figure 27).  This is consistent with the findings of this study that there is a warming of weather types in spring and autumn but not in winter and summer. Brown (2002) also used the objective LDWTs and daily values of  CET rather than seasonal values of  CET to assess whether or not there has been a warming of weather types since 1881 and concluded that there had been, mainly in spring and autumn as some weather types in winter and summer  had cooled. The agreement between this study and that of Brown (2002) using two quite different approaches that weather types in spring and autumn are warming offers confidence that climate change not accounted for by a change in the incidence of warm/cold weather types is occurring in these two seasons. In both studies, these warming trends appear to be long term trends that apply to the full 1881 to 2002 period and not just a recent phenomenon (the differences (yellow line) in Figures 14 & 16 clearly show this). A warming of weather types (e.g weather associated with westerly winds) is logically what would be expected if higher concentrations of CO₂ are warming  the atmosphere (so-called anthropogenic warming). UKCIP02 predict warming in all seasons but "there may be greater warming in summer and autumn than in winter and spring".  The pattern of warming of  CET during the twentieth century (Table 1) is greatest warming in autumn and least in summer so with respect to summers, there is disagreement between temperature observations and the climate predictions made by UKCIP02 using atmosphere-ocean model simulations of global climate.  The findings in this study and that of Brown (2002) of a warming of  weather types in spring and autumn but not in winter and summer is somewhat  inconsistent with anthropogenic warming as the higher concentrations of atmospheric CO₂ should have a warming effect all year round if they are having this warming effect at all.  These observations together with others like the lack of  warming at high latitudes where anthropogenic warming should be greatest (Predictions Fall Foul of Reality) and the modest warming of  satellite temperatures of the  lower atmosphere since 1979 most of which is explained by the strong El Nino of 1998 (Figure 31) collectively constitute a body of scientific evidence that is at variance with the global warming predictions made by atmosphere-ocean model simulations of global climate and advocated by  the IPCC and others as proof of  anthropogenic global warming. Certainly the UK climate is warming but it has been since instrumental records began in 1659 and with respect to the global surface temperature record,  the warming trend it shows judging by the CET is probably broadly correct but as the accuracy of  the surface record is questionable, the rise it portrays may well be somewhat exaggerated.  Given these long term warming trends that existed long before anthropogenic  CO₂ became an issue, the global warming sceptic argument that the observed climate changes of the 1990s and indeed the whole twentieth century may be entirely natural is reasonable.

Europe is one of a few discrete regions which the satellite record of the lower atmosphere shows to have warmed from 1979 to 1996  (Jones, Osborn, Wigley, Kelly & Santer, 1997)  which is consistent with the warming in the UK from 1989 to 2002. A relevant question to the anthropogenic climate change debate is whether or not the warming of 1989 to 2002 is unusual in the context of historical weather data. Figure 1 shows that although CET is at it's highest level in 332 years,  this is merely the most recent part of a long established warming trend. Although the current rate of warming is higher than average, a long period of higher rates prevailed in the early eighteenth century so overall, the CET  indicates that the 1989-2002 period is unusual but not unprecedented in terms of  rate of  temperature change. The LDWTs and the PSC indices which are derived from them, provide a more detailed insight into historical climate variations than that portrayed by CET and EWR. Murray & Lewis (1966) examined monthly and annual PSC means for five year periods commencing from 1873 to 1964 and found marked differences between five year periods for all three indices. These results are supported by Figures 5, 6, 7, 8 & 9  which compare mean decadal values of  PSC indices, annually and in each season. Therefore large differences between decades in the synoptic weather patterns around the UK are normal so the fact that synoptic weather patterns around the UK from 1989 to 2002  have been different to those of preceding decades is not necessarily noteworthy or indeed evidence of anthropogenic climate change. This issue is discussed further with reference to winters and summers 1989 to 2002 as these two seasons, as measured by  PSC quintiles had the most anomalous synoptic weather patterns.

For winters 1989 to 2002, only 16% of the 42 PSC indices (3 indices x 14 years) were average (i.e quintile 3) and Table 14 shows that high P quintiles (50% of years with quintiles 4 or 5)  were characteristic of these winters and that this factor accounted for the high temperatures of many these winters. The correlation between the NAO and the P index is very high  (r = 0.799) supporting the argument that the positive phase of the NAO which has prevailed from 1989 to 2000 is the overriding factor which accounts for the anomalous synoptic weather patterns of  winters 1989 to 2002. This argument can be verified by selecting an earlier period with positive NAO winters and examining the PSC quintiles of these winters. The period 1903 to 1916 had positive winter NAO in each year (mean NAO = 1.2) (Figure 30) and the incidence of PSC quintiles in these fourteen winters is compared with those of 1989-2002 in Table 24. Clearly these two periods have very similar PSC quintiles indicating that the synoptic weather patterns of these two winter periods were similar (Above Average X²₍₂₎ = 0.10;  Average: X²₍₂₎ = 0.09;  Below Average: X²₍₂₎ = 0.15;  all three Chi-squared tests were not significant) which verifies the argument  that the anomalous winter weather patterns of 1989 to 2002 can be credibly explained by the recent positive phase of the NAO.
 

Table 24  Comparison of  the incidence of PSC quintiles in winters 1903-16 with quintiles winters 1989-2002

For summers 1989 to 2002,  only 29% of  the 42 PSC indices (3 indices x 14 years) were average (i.e quintile 3) and Table 17 shows that high S quintiles (36% of years with quintiles 4 or 5)  and low C quintiles (50% of years with quintiles 1 or 2) were characteristic of these summers and that both these factors accounted for the higher temperatures of many these summers. It has already been noted that the period 1881-1909 also enjoyed drier (Figure 19) and more anticyclonic (Figure 8) summers than average although only three of these summers exceeded the mean temperature of the 1989-2002 summers (16.0 C). The incidence of PSC quintiles in summers 1881-1909 is compared with those of summers 1989-2002 in Table 25.  Clearly these two periods have very similar PSC quintiles indicating that the synoptic weather patterns of these two summer periods were similar (Above Average X²₍₂₎ = 1.44;  Average: X²₍₂₎ = 1.63;  Below Average: X²₍₂₎ = 0.06;  all three Chi-squared tests were not significant)  and showing  that the anticyclonic summer conditions which have prevailed since the 1970s are not unprecedented in the last 120 years.
 

Table 25  Comparison of  the incidence of PSC quintiles in summers 1881-1909 with quintiles summers 1989-2002

Rainfall Changes

The EWR record from 1766 (Figure 3)  also lends support to the theory that the UK climate is changing but as with CET, change is a characteristic of the entire 236 year record and not just a recent phenomenon. Unlike CET, there is no long term trend in annual rainfall and while it is true that the last few years have been much wetter than average, these years are not without precedent in the entire series. UKCIP02 states that "winters will become wetter and summers may become drier throughout the UK".  This study found that the winter EWR showed no long term trend to wetter winters (Figure 17) as the few very wet winters of recent years were balanced by some very dry winters. The dramatic increases in total winter rainfall noted by Osborn & Hulme (2002) were mainly confined to northern and western UK which correlates with the recent positive phase of the NAO whereas a UK wide increase in heavy rainfall days (> 15mm) was  found. Presumably, the few very wet winters in  England largely account for the increase in heavy rainfall days here as total rainfall in some regions only increased modestly and in some regions, the number of rain days decreased.

Osborn & Hulme (2002) present the argument (also expressed by others) that globally averaged precipitation increases with global mean temperature and this is supported by climate model simulations and short observational records. They go on to present the theory behind this expectation and explain how an increase in the occurrence of heavy precipitation may be associated with the increases in mean precipitation. The methodology used by Osborn & Hulme (2002) is therefore designed to detect increases in heavy rainfall events and it was successful: in winter, increases in total precipitation, mean rainfall on each rain day, heavy rainfall days (>15 mm) and multi-day sequences of heavy rainfall days right across the UK were found. These data for winter rainfall in the UK therefore appear to support the argument that higher global temperatures lead to increased regional rainfall. However, this study shows that these changes in winter EWR can be solely explained by changes in P & C indices (discussed  in the winter rainfall section of  "The 1989 to 2002 warm period in detail"). The main controlling factor in winter rainfall is the C index (Figure 22) and if  higher global temperatures are causing an increase in the mean amount of rain on each rain day or a greater incidence of days with rainfall above certain thresholds (e.g 15 mm), then logically the correlation between winter C index and winter EWR  should increase as the same level of cyclonicity would give increasing amounts of rainfall. The multiple regression models predicting seasonal EWR do not support this argument (differences (yellow line) in Figures 17, 18, 19 & 20) as there was no tendency for them to underestimate EWR in recent years in all four seasons which means that annual changes in weather types accounted for the few individual very wet winters of recent years. Table 26 compares the correlation coefficients between winter C index and winter EWR in each decade from the 1880s to 1990s (2000-2002 included in the 1990s data) and does not support this argument either. It is apparent that the 1990s did not have the highest correlation coefficient but the 1900s, 1910s and 1960s had correlations well over 0.9 indicating that for the same level of cyclonicity, greater winter EWR totals occurred in these decades than in the 1990s. The conclusive way to settle this point would be if the analyses of Osborn & Hulme (2002) were altered so the varying levels of  cyclonicity and indeed progressiveness (i.e days with westerly winds)  are accounted for. This study credibly explains EWR variations in all seasons  solely on the basis of weather types and  changes in weather type also probably explain the UK wide rainfall changes noted by Osborn & Hulme (2002).  The argument articulated by others that there is an association between global temperatures and these changes in winter rainfall in the UK is not supported by this study.
 

Table 26  Comparison of the correlation coefficients between winter P index and winter EWR each decade 1880s to 1990s

                       *     1990s includes data from 2000 to 2002.

In summer, a trend to lower rainfall since 1960 is certainly a characteristic of summer EWR (Figure 19) which supports UKCIP02 but a similarly dry run of summers prevailed from 1881 to 1909. The cause of both periods of drier summers was greater anticyclonicity, a very welcome characteristic of summer!  Table 23 demonstrates that there is an association between the winter P index and the summer C index which indicates that the drier, warmer summers of 1989 to 2002 are linked to the milder westerly type winters of this period. Both changes can therefore be linked to the positive phase of  the winter NAO with the consequence that when the NAO reverts to its negative phase, winters are likely to become drier and colder and summers cooler and wetter.

In autumns, in spite of the exceptional rainfall event of autumn 2000, autumn EWR does not show a trend to higher autumn rainfall. UKCIP02 do not predict wetter autumns but drier ones. 
 

Marks out of ten for UKCIP02

Iit will not be possible to assess the accuracy of UKCIP02 until later in the century but what can be done now is to compare recent UK climate observations with UKCIP02 to see if they are consistent.

1.  The UK will become warmer:  CET (Figure 1) shows the UK is becoming warmer and has been warming over 332 years so logically further warming can be expected (One mark).

2.  By the 2080s, the average annual temperature in the UK may rise by between 2 C for the Low Emissions Scenario and 3.5 C for the High Emissions Scenario:  average increases in 30 year running means from 1989 to 2002 (Table 22) indicate that rises of  2 C by 2100 are possible should this observed rate of increase continue (One mark).

3. There may be more warming in autumn than winter and spring:  autumn (Table 1) warmed around 1 C during the twentieth century, similar to winter but more than spring (One mark).

4. There may be more warming in summer than in winter and spring: summer (Table 1) only warmed by around 0.5 C during the twentieth century, about half the warming of winter and autumn (Zero mark).

5. There will be greater warming in the south and the east than the north and the west: increases in 30 year running means of  annual temperature (Table 21) show greater warming in England & Wales than in Scotland and Ireland (One mark).

6. Very cold winters will become increasingly rare: if a threshold of mean winter temperature 1 C below the 1961-90 average is applied to winter CET 1881 to 2002, there are 23 occasions (19%) when this threshold was exceeded (Figure 27). Therefore assuming no warming trend, the expected number of winters per decade with temperatures below this threshold would be 1.9. The 1980s had three such winters while the 1990s had two (Zero mark).

7. High summer temperatures will become more frequent: the mean summer temperature 1989 to 2002 was 0.7 C above the 1961-90 average but there was no trend to higher temperatures from 1881 to 1988 (Figure 27).  If a threshold of mean summer temperature 1.5 C above  the 1961-90 average is applied to summer CET 1881 to 2002, there are 7 occasions (6%) when this threshold was exceeded, three in the first sixty years and four in the last sixty years (Zero mark).

8. Winters will become wetter:  although there have been some very wet winters from 1989 to 2002, there were also some dry ones and no long term trend to wetter winters is apparent since 1881 (Figure 17). However, Osborn & Hulme (2002) showed increases in total winter precipitation UK wide but especially in northern and western regions (One mark).

9. Heavy winter precipitation will become more frequent: this study only considered winter totals of  EWR so cannot reliably assess this point. However, Osborn & Hulme (2002)  show conclusively that heavy rainfall events in winter have increased since 1960 (One mark).

10. Summers may become drier:  there is a clear trend to lower summer rainfall since 1960 (Figure 19) with two summers since 1970 with less than 100mm (1976 & 1995)  (One mark).

With a score of seven out of ten, UKCIP02  are mostly supported by recent UK climate data.
 

Final comments

Iin addition  to higher than average  temperatures, highly anomalous  synoptic weather patterns during winter and summer were characteristic from 1989 to 2002. However, strong similarities between the winter synoptic weather patterns of 1989 to 2002  and those of 1903 to 1916 and between the summer synoptic weather patterns of 1989 to 2002 and those of 1881-1909 were found proving that in the historical context of 1881 to 2002, 1989 to 2002 is not unprecedented in terms of  winter and summer synoptic weather patterns.  The positive phase of the NAO which commenced in winter 1988/89 was found to be associated with the occurrence of  these anomalous weather patterns in  both  seasons  and the NAO is an entirely natural phenomenon which has been known to science for some time. When the next negative phase of the winter NAO commences, and there is some evidence that it has already done so, winters are likely to become colder and drier and summers cooler and wetter, reversing the trends  seen since 1960.

From the perspective of anthropogenic climate change, the clear long term warming of spring and autumn weather types found in this study, which is supported by the analyses of Brown (2002) is  far more relevant than the unusual synoptic weather patterns of 1989 to 2002. The fact that the warming in both these seasons occurred throughout the record from 1881 onwards evidently accounts for the long term warming shown by the CET.  However, the fact that neither winters nor summers showed  any warming of weather types or  indeed any long term  warming trend from 1881 to 1988 does not support the hypothesis that higher temperatures in the atmosphere caused by higher concentrations of CO₂ are driving this temperature trend in CET.  This should logically cause warming in all seasons. Quite what the factor is that is causing a warming in spring and autumn but not in winter and summer is a very interesting and relevant question but the author has no suggestions.

Finally, the argument proposed by some (e.g Osborn & Hulme, 2002) that recent increases in winter rainfall in the UK, in particular the incidence of heavy rainfall days, may be linked to increases in global temperature enhancing the hydrological cycle, thereby leading to heavier rainfall in some regions is not supported by this study. The multiple regression analyses of this study credibly explain rainfall changes in all seasons from 1881 to 2002 on the basis of  weather types and the recent positive phase of the winter NAO.
 

Update:  Spring temperature trends examines the warming of spring weather types in further detail   14 March 2003

REFERENCES

Brown, P. R. (2002).  Relationships between the Lamb Daily Weather Types and Central England Temperatures from 1881 to 2000. Journal of Meteorology, Vol 27, No 373, 326-334.

Jenkinson, A.F. &  Collison, F.P. (1977). An initial climatology of gales over the North Sea. Synoptic Climatology Branch Memorandum No. 62, Meteorological Office, Bracknell.

Jones, P.D., Hulme, M. &  Briffa, K.R. (1993). A comparison of Lamb circulation types with an objective classification scheme. Int. J. Climatol. 13, 655-663.

Jones, P.D., Osborn, T.J., Wigley, T.M.L, Kelly, P.M. and Santer, B. D. 1997: Comparison between the microwave sounding unit temperature record and the surface temperature record for 1979 to 1996: real differences or potential discontinuities. Journal of Geophysical Research, 102, 30135-30145.

Jones, P. D. & Briffa, K. R. (2001). The Little Ice Age: local and global perspective. Climatic Change, 48, 5-8.

Lamb, H. H. (1972). British Isles Weather types and a register of daily sequence of circulation patterns, 1861-1971. Geophysical Memoirs, 16, HMSO 85pp.

Manley, G. (1974). Central England temperatures: monthly means 1659 to 1972. Q. J. of Royal Meteorological Society, 100, 389-405.

Murray, R. & Lewis, R. P. W. (1966). Some aspects of the synoptic climatology of the British Isles as measured by simple indices. Meteorological Magazine, 95, 193-203.

Murray, R. & Benwell, P. R. (1970). PSCM indices in synoptic climatology and long range forecasting. Meteorological Magazine, 99, 232-245.

Murray, R. (1972). Monthly mean temperature related to synoptic types over Britain specified by PSCM indices. The Meteorological Magazine, 101, 305-311.

Osborn, T. J. & Hulme, M. (2002) Evidence for trends in heavy rainfall events over the UK. Phil. Trans. R. Soc. London., 360, 1313-1325.

Perry,  A. (1968). The regional variation of climatological characteristics with synoptic indices. Weather, 23, 325-330.

Perry, A. (1969). The PSCM index and regional anomalies of temperature, rainfall and sunshine. Weather, 24, 225-228.

Rodwell, M.J., Rowell, D.P. and Folland, C. K. (1999) Oceanic Forcing of the wintertime North Atlantic Oscillation and European climate. Nature, 398, 320-323.

Wigley, T. M. L., Lough, J. M. & Jones, P. D. (1984) England and Wales precipitation: a discussion of recent changes in variability and an update to 1985   J. of Climatology,  4,  1-25.