204
Murat Tiirke~
relationships for the Markov persistence are also seen to be satisfied approximately
in summer series of four stations, in one autumn series, and in the annual series of
four stations. On the other hand, the significant negative L-1SCs in spring series are
associated closely with a marked short-period oscillation rather than a persistence,
which comes out in the spectral analysis.
4.2.2 Aridity Index Series
Fluctuations and trends. As an expected result of the decreasing trends in annual
precipitation of most stations, annual AI series have generally tended to decrease at
many stations of Turkey. However, significant downward trends are found at some
stations of the MED region (Table 2). At the stations of the Aegean part of the MED
region with a significant decreasing trend, there is a marked change from the humid
or near humid conditions of the 1960s to the dry sub-humid or semi-arid climatic
conditions of the mid and late 1980s and early 1990s (Fig. 12). In contrast, annual AI
values tend to increase significantly towards humid or semi-humid climatic conditions
at 5 stations, which are located over the northern part of the CCAN region of Turkey
(Table 2).
In addition to the secular trends from the Mann-Kendall u(t) statistics, the
beginning of secular or partial trends in AI series, the duration of significance and the
change points were investigated by means of time-series plots of sequential values of
the statistics u( t) and u' (t) from the Mann-Kendall test. Plots of the selected stations
are shown in Fig. 13. When the curve of the u(t) values exceed the absolute value of
the critical value, which is 1.96 at the 5 percent significant level, a significant period
of trend is indicated. On the plots of Balikesir, Akhisar, Dikili and Bodrum, few
values of u(t) passed the 5 per cent significant level, although the u(t) statistics of
these stations indicated a significant decreasing tendency in their AI series. Thus, as
pointed out by Sneyers (1990), it may be assumed that, if the trend is real, its effect
is very recent, and it is recommended to wait for later observations to confirm the
trend. For an abrupt change, or for the beginning point of a secular trend, u(tJ and
u' (tJ curves would exhibit the same behaviour at the change point over the series. The
two lines should also intersect at that point. This clearly shows up on the plots of
Afyon, Ankara and Merzifon stations. Upward trends in AI series of Ankara and
Merzifon began in 1957 and 1970, and become significant during the periods of
1971-1993 and 1979-1993, respectively.
Persistence and periodicity. Tiirke~ (1999) applied the L-I SC test to the annual AI
values, and found followings. The AI series of eight stations indicate a significant
positive L-lSC (Table 2), although most AI series are dominated by a positive serial
correlation. By considering the exponential relationships r 2 == r/ and r3 == r/, it was
found that only the AI series of Konya station in the CCAN region is characterized
Murat Tiirke~
relationships for the Markov persistence are also seen to be satisfied approximately
in summer series of four stations, in one autumn series, and in the annual series of
four stations. On the other hand, the significant negative L-1SCs in spring series are
associated closely with a marked short-period oscillation rather than a persistence,
which comes out in the spectral analysis.
4.2.2 Aridity Index Series
Fluctuations and trends. As an expected result of the decreasing trends in annual
precipitation of most stations, annual AI series have generally tended to decrease at
many stations of Turkey. However, significant downward trends are found at some
stations of the MED region (Table 2). At the stations of the Aegean part of the MED
region with a significant decreasing trend, there is a marked change from the humid
or near humid conditions of the 1960s to the dry sub-humid or semi-arid climatic
conditions of the mid and late 1980s and early 1990s (Fig. 12). In contrast, annual AI
values tend to increase significantly towards humid or semi-humid climatic conditions
at 5 stations, which are located over the northern part of the CCAN region of Turkey
(Table 2).
In addition to the secular trends from the Mann-Kendall u(t) statistics, the
beginning of secular or partial trends in AI series, the duration of significance and the
change points were investigated by means of time-series plots of sequential values of
the statistics u( t) and u' (t) from the Mann-Kendall test. Plots of the selected stations
are shown in Fig. 13. When the curve of the u(t) values exceed the absolute value of
the critical value, which is 1.96 at the 5 percent significant level, a significant period
of trend is indicated. On the plots of Balikesir, Akhisar, Dikili and Bodrum, few
values of u(t) passed the 5 per cent significant level, although the u(t) statistics of
these stations indicated a significant decreasing tendency in their AI series. Thus, as
pointed out by Sneyers (1990), it may be assumed that, if the trend is real, its effect
is very recent, and it is recommended to wait for later observations to confirm the
trend. For an abrupt change, or for the beginning point of a secular trend, u(tJ and
u' (tJ curves would exhibit the same behaviour at the change point over the series. The
two lines should also intersect at that point. This clearly shows up on the plots of
Afyon, Ankara and Merzifon stations. Upward trends in AI series of Ankara and
Merzifon began in 1957 and 1970, and become significant during the periods of
1971-1993 and 1979-1993, respectively.
Persistence and periodicity. Tiirke~ (1999) applied the L-I SC test to the annual AI
values, and found followings. The AI series of eight stations indicate a significant
positive L-lSC (Table 2), although most AI series are dominated by a positive serial
correlation. By considering the exponential relationships r 2 == r/ and r3 == r/, it was
found that only the AI series of Konya station in the CCAN region is characterized