3.1. Land Use Changes
In this study, we analyzed land use and landscape pattern changes based on remote sensing data. The distributions of forestland, marshland, paddy field and dry farmland, and their proportions in 1982, 1995, 2000, 2005, 2010, and 2015 were demonstrated (see Figure 1
and Table 2
). The marshland area was 4336.4 km2
in 1982, accounting for 18% of the total study area (see Figure 1
It was indicated that the marshlands decreased remarkably by 63.29%, forestlands decreased by 12.88%, and dry farmlands decreased by 0.01% from 1982 to 2015. However, paddy fields increased 1.78 times during this period (see Table 2
). From 1982 to 2015, the increasing proportion of paddy fields was much higher than that of the decreasing proportion of marshlands.
In different periods, the characteristic of mutual conversion between paddy fields and dry farmlands was quite distinguishing. Paddy fields experienced large conversion into dry farmlands during 2005–2010 (1788.57 km2
), followed by a reverse conversion from 1995 to 2000 (2379.60 km2
) (see Table 3
). Therefore, the biggest amplitude of dry farmlands conversion to paddy fields was between 1995 and 2000 with a relative change of 103.05% (see Table 3
). In general, the exploitation scale of paddy fields was enlarged obviously. The total converted quantities of marshlands to dry farmlands were significantly higher than those of marshlands to paddy fields in various stages. The relative change of paddy fields was higher than that of dry farmlands, which indicated that the development speed of paddy field was faster than that of dry farmlands among the five periods. From 1995 to 2000 and 2010 to 2015, the relative change of paddy fields was always positive compared with that of marshlands and dry farmlands, so it was revealed that paddy fields expanded remarkably in these two periods.
There is no obvious difference in the proportions of forestlands from 1982 to 2015, with the maximum proportion of 29.95% in 1982 and the minimum proportion of 26.10% in 2015 (see Table 2
). The largest reduction of forestland area was 648.70 km2
from 1995 to 2000 and the relative change was −1.84% (see Figure 2
and Table 3
). There was 578.70 km2
of forestlands conversion to dry farmlands in this period (see Table 3
). Therefore, it was about 90% of the largest reduced forestland area from 1995 to 2000.
3.2. Landscape Pattern Changes
Delta (Δ) means difference in our study. As shown in Table 4
, from 1982 to 2015, ΔLPI, and ΔFRAC_AM of paddy fields were higher than those of marshlands and dry farmlands, which indicated stronger human intervention and severe landscape fragmentation. ΔNP of paddy fields and dry farmlands were obviously higher than that of marshlands, which showed that the intensity of cropland exploitation was increased especially after 1995.
ΔSPLIT of marshlands was 1807.34 from 1982 to 2015, obviously higher than that of paddy fields and dry farmlands, so they were much scattered. On the other hand, ΔAI of marshlands was smallest from 1982 to 2015, and meanwhile, ΔCOHESION of marshlands had the maximum amplitude with the trend of fluctuation, which indicated that patch connectivity was not compact.
Clear evidences of this fragmentation process can be observed on the remaining landscape-level metrics (see Table 5
). IJI significantly decreased with a more scattered pattern of landscape from 1982 to 2015. Meanwhile, NP rapidly increased, which led to a clear fragmentation process.
CONTAG had obvious difference with a range of 5.69%, and there were dominant patches with high connectivity. Because the range of CONTAG value was from 0 to 100%, the CONTAG in our study was about 70% in the direction of 100%. COHESION was about 99.9 with no particularly obvious change in different years, which showed that landscape connectivity was sustained.
The SHDI value reached the maximum in 2015, and meanwhile, the SHEI value in 2015 was much higher, which meant that the landscape area ratio tended to be further heterogeneous. Hence, there was a more even distribution of the patch types in landscape.
There are four subtypes of forestlands, including thick woodland, shrub land, sparse woodland, and others. As shown in Table 6
, NP, LPI and FRAC_AM values of thick woodland were much higher from 1982 to 2015, which indicated that there were more fragmented and severe human activities. ΔAI of thick woodland was smaller than that of sparse woodland and others from 1982 to 2015, and meanwhile, ΔCOHESION was smaller than sparse woodland and shrub land, which indicated that patch connectivity was not compact. SPLIT of others was highest, which showed that they were much scattered.
SHAPE_AM was 8.93 in 1982 (see Table 7
), which showed that shape of patches was more complicated and irregular than those in other years. CONTAG values were all over 89 and COHESION was about 99.9 with no particularly obvious change in different years, which indicated that landscape connectivity was sustained. IJI had obvious fluctuation with a decreased trend, which demonstrated the more scattered pattern of forestland landscape from 1982 to 2015. Meanwhile, NP substantially increased, which brought obvious landscape fragmentation. The SHDI value was the highest in 2010, so the same landscape was more diverse in different periods. The maximum value of SHEI was 0.20, and all of SHEI values were much lower than 1 in different periods, which indicated that the patch types in forestland landscape were unevenly distributed.