Multifractal analysis results exhibit the clear difference of
multifractality between the daily APIs, the pollution indices of SO2,
NO2 and PM10 in Figure 3. We can clearly recognize that they are
different in the shape of f(α)~ α curves different from the Marche
It is obvious that the shape of f(α) curves for APIs indices is like
hooks and PM10 is like a little bell to the right, however, SO2 and
NO2 are like hooks to the left. These different shapes reflect inner
dynamical characteristics of four pollution indices. Meanwhile, it
also shows that multifractal analysis as an effective and new tool
which completely reveals the differences from fractal structure of
complicated dynamic behavior among various pollutants.
Moreover, we calculated the multifractal parameters (Δα, Δf,
B) of four pollution indices. Δα values of APIs, PM10, NO2 and SO2
indices are 0.368, 0.393, 0.461 and 0.312, respectively; Δf are–
0.322,–0.264, 0.112 and 0.261, respectively; the B values are–
0.520,–0.397, 0.638, and 1.149, respectively. Compared with PM10
and NO2, the minimum Δα of SO2 shows that its multifractal
spectra f(α) are relatively dense and the degree of multifractality is
relatively weak, so as to further explain that the singularity of SO2
pollution index fluctuation is smaller, and there is insignificant
variation for highest and lowest point of the index. However, the
maximum Δα of NO2 then shows its strongest multifractal
characteristics. There exist great differences between multifractal
spectrum f(α) ~ α of PM10 and SO2 (NO2) indices, which are mainly
embodies in negative Δf and B values. This shows that spectra f(α)
~ α of PM10 indices (APIs) is to the right and the singular values for
range of α left are larger, which also shows the event o larger
pollution indices occupies the leading position with some local
pollution index felling.
The PSA results show that the daily APIs and SO2, NO2 and
PM10 indices exhibit the high persistence or long term memory in
about one year. As a result of the time evolution of air pollutant in
city and its air pollution control measures, every year time series
has a different dynamic characteristics, thus, they exhibit different
complicated characteristics. The difference may be quantitatively
demonstrated using multifractal spectrum. The research of
multifractal characteristics changes of every year time series (from
July 1998 to June 2012) can contribute to discussing time evolution
dynamic behavior of each air pollutant and developing advanced
techniques for air pollution forecasting. Therefore, in the present
study, the changes of multifractal characteristics for three
atmospheric pollutants were analyzed in the span of one year. So
1998 July to 2012 June is divided into 14 years, and their
multifractals were calculated, respectively. The results are shown in
Figure 4.
Figure 4a shows the multifractal parameters for SO2 indices in
each year. We can obviously see that Δα variation is stable at 0.23–
0.40, which shows that the singularity of fluctuation distribution for
SO2 indices in each year is similar and its difference between the
highest and lowest points is relatively stable. However, Δf values
change greatly, and this is because complex fractal structural
change to be resulted in by was different intrinsic dynamic
mechanism in each year. But the right endpoint lower or shorter
than the left end point for almost all multifractal spectrum (Δf>0),
which illustrates the probability for the index felling the lowest is
always shorter than that of the highest. As to B, their values in each
year are all positive which indicates the shape of f(α) curve inclines
to left and relatively low fractal indices are dominant.
Figure 4. Multifractal parameter for the daily APIs, the pollution indices of
SO2, NO2 and PM10 in each year. Lines with “●”, “★” and “▲” denote,
respectively, Δα, Δf and B.
Figure 4b shows the multifractal parameters for NO2 indices in
each year. For Δα, its values of the first six years (from 1998 July to
2004 June) and recent four years (from 2009 July to 2012 June) are
steadily rising from 0.382 (0.354) to 0.591 (0.572), indicating that
the singularity of fluctuation distribution and the difference of the
highest and lowest points are increasing year by year. These
fluctuating Δf values mainly results from rise or fall of f(αmin),
suggesting that the probability of event for largest index is also
fluctuating, but that of the smallest index is more stable. For B in
each year, they fluctuated between positive and negative values,
and overall it is negative in the first four years and B value of the
second year is much smaller than that of the first year, showing
that NO2 indices present a rising trend during this period, but rising
slightly significantly after 2002 June. And yet the index show a
downward trend after 2003 June, so repeatedly, the trend become
slight until 2006 June. This shows that although sometimes larger
index events is still dominant, but the index smaller event has
shown its dominant role, namely, due to various reasons resulting
in NO2 sometimes up, sometimes down, was extremely unstable,
but its last showing a small decline trend. These demonstrate that,
although sometimes the larger pollution events still dominate, the
smaller event has gradually shown its dominant role. In other
words, due to various reasons NO2 sometimes gets up and
sometimes down, it gets extremely unstable, but it last shows a
small drop.
Figure 4c shows the multifractal parameters for PM10 indices
in each year. It is obvious that Δα values are generally on a
downward trend indicating that the singularity of fluctuation
distribution and the difference of the highest and lowest points are
decreasing year by year, and the distribution of PM10 indices
presents more and more uniformity. For Δf, its changes are most
pronounced after July 2004, mainly resulting from f(αmin) increasing