236
P. Maheras and C. Anagnostopoulou
With respect to the simulated monthly precipitation totals, loannina shows the
best results since the agreement between observed and simulated totals is acceptable
for all months except February. For some months, there is almost exact agreement
(January, March, August). The precipitation totals averaged over all stations show
very good agreement, except for January (when observed totals are higher than
simulated) and February (when the opposite occurs). Rhodes appears large
discrepancies between observed and simulated totals for December, January, March
and September, while agreement in other months is acceptable according to the X 2
test.
As for the annual cycle, the mean observed and simulated rainy days and the
corresponding precipitation totals were calculated for selected extreme daily
precipitation events. Daily amount thresholds of >2Smm, >30mm, >40mm and
>SOmm were selected. The results of these calculations for the validation periods for
annual, rainy and dry periods are presented in Fig. 2Sa and 2Sb.
The X 2 statistic was employed to test the agreement between observed and
simulated rain days and the corresponding precipitation totals, with the same
acceptance levels as for the annual cycle. According to the results of the X 2 test,
agreement between the simulated and observed values for all extreme events is
acceptable, not only for rain days but also for precipitation totals (except for
precipitation events >SOmm for Rhodes). Some more analytical details relating to rain
days are presented below.
Averaged over all stations, there is almost absolute agreement for rainy days for
events >2Smm, >30mm and >40mm for the annual, rainy and dry periods. For
Ioannina, there is almost absolute agreement during the dry period for events >30mm
and for annual, rainy and dry periods for events >40mm. However, for Rhodes, there
is absolute agreement for the annual period for events >2Smm and for the rainy
period for events >30mm only.
For precipitation amounts it results:
a) A veraged over all stations, there is almost absolute agreement between rainfall
totals for events >2Smm for the annual, rainy and dry periods, events >30mm and
>40mm only for the dry period. For the other cases, there is a slight
overestimation of the simulated totals with respect to the observed.
b) For Ioannina, almost absolute agreement between rainfall totals occurs for events
>2Smm, >30mm and >40mm for the dry period and for events >SOmm for the
annual, rainy and dry periods. On the contrary, for the other cases there is a slight
underestimation of the simulated totals with respect to the observed.
c) For Rhodes, a very good agreement between rainfall totals occurs only for
rainfall events >40mm for the annual, rainy and dry periods. For the other cases,
as for the results averaged over all stations, there is a slight overestimation of the
simulated totals with respect to the observed. As already mentioned, agreement
between the simulated and observed rainfall totals for events >SOmm is not
statistically significant at the O.OS level.
P. Maheras and C. Anagnostopoulou
With respect to the simulated monthly precipitation totals, loannina shows the
best results since the agreement between observed and simulated totals is acceptable
for all months except February. For some months, there is almost exact agreement
(January, March, August). The precipitation totals averaged over all stations show
very good agreement, except for January (when observed totals are higher than
simulated) and February (when the opposite occurs). Rhodes appears large
discrepancies between observed and simulated totals for December, January, March
and September, while agreement in other months is acceptable according to the X 2
test.
As for the annual cycle, the mean observed and simulated rainy days and the
corresponding precipitation totals were calculated for selected extreme daily
precipitation events. Daily amount thresholds of >2Smm, >30mm, >40mm and
>SOmm were selected. The results of these calculations for the validation periods for
annual, rainy and dry periods are presented in Fig. 2Sa and 2Sb.
The X 2 statistic was employed to test the agreement between observed and
simulated rain days and the corresponding precipitation totals, with the same
acceptance levels as for the annual cycle. According to the results of the X 2 test,
agreement between the simulated and observed values for all extreme events is
acceptable, not only for rain days but also for precipitation totals (except for
precipitation events >SOmm for Rhodes). Some more analytical details relating to rain
days are presented below.
Averaged over all stations, there is almost absolute agreement for rainy days for
events >2Smm, >30mm and >40mm for the annual, rainy and dry periods. For
Ioannina, there is almost absolute agreement during the dry period for events >30mm
and for annual, rainy and dry periods for events >40mm. However, for Rhodes, there
is absolute agreement for the annual period for events >2Smm and for the rainy
period for events >30mm only.
For precipitation amounts it results:
a) A veraged over all stations, there is almost absolute agreement between rainfall
totals for events >2Smm for the annual, rainy and dry periods, events >30mm and
>40mm only for the dry period. For the other cases, there is a slight
overestimation of the simulated totals with respect to the observed.
b) For Ioannina, almost absolute agreement between rainfall totals occurs for events
>2Smm, >30mm and >40mm for the dry period and for events >SOmm for the
annual, rainy and dry periods. On the contrary, for the other cases there is a slight
underestimation of the simulated totals with respect to the observed.
c) For Rhodes, a very good agreement between rainfall totals occurs only for
rainfall events >40mm for the annual, rainy and dry periods. For the other cases,
as for the results averaged over all stations, there is a slight overestimation of the
simulated totals with respect to the observed. As already mentioned, agreement
between the simulated and observed rainfall totals for events >SOmm is not
statistically significant at the O.OS level.