 |
| Titel |
Characterization of drought events in the Iberian Peninsula 1931-2009 |
| VerfasserIn |
Jose Luis Valencia, Maria Villeta, Ana María Tarquis, Antonio Saá-Requejo |
| Konferenz |
EGU General Assembly 2015
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 17 (2015) |
| Datensatznummer |
250107791
|
| Publikation (Nr.) |
 EGU/EGU2015-7505.pdf |
|
|
|
|
|
| Zusammenfassung |
| There are many definitions of drought due to an imbalance between water availability and its
use. In general, when there is an extended period of time in which the rainfall remains below
their mean values, drought incurred. It is important to assess the severity, duration and
frequency with which these phenomena occur, because its influence in the evolution of
ecosystems, agriculture and livestock. In this work we use the Standardized Precipitation
Index computed in a 3-month and 12-month time scale (SPI3 and SPI12) in two periods:
1931-1969 and 1970-2009, applied to synthetic series generated by Luna and Balairón
(AEMET). These daily rainfall series cover a squared region of the Iberian Peninsula
regularly.
We have fitted loglinear models to drought class transitions derived from Standardized
Precipitation Index. Each period is analyzed independently and we obtain separated squared
contingency tables. Each cell in these tables have two classification criterions (A,B) with
levels i,j, respectively, and two identifiers of the table (p and k) where level i is the
drought class at month t - 1, j is the drought class at month t, p is the period of time
and k identifies a short region inside the Iberian Peninsula. Then, the elements
of these tables can be represented by ni,j,p,k. The model selected to explain the
differences in the contingency tables for all the stations is the quasi association loglinear
model, where different parameters for different periods and regions (p and k) are
fitted.
log(Ekijp) = λ +λAi + λBj + αuivj + δ1i(i = j)
When adjusting these models, it is assumed that the values in the cells of the contingency
tables were taken by independent Poisson distributed variables. To validate the adjustment of
the model, we use the residual deviance G2 which is asymptotically distributed as a
Chi-Square for each contingency table.
( )
G2 = 2/ n log nl
l l Ënl
We have examined if there are differences in the expected number of transitions between all
drought classes by ANOVA linear model with fixed effects for the periods.
The main conclusion is that agricultural droughts become more frequent, the
duration of each event does not increase, but the droughts intensity is growing. The
probability of consecutive raining days has decreased, especially in autumn and in winter. |
|
|
|
|
|
|