The height of sample trees extracted from CHM largely differed from ground measurement results because the handheld GPS failed to perform differential processing, which led to major positioning errors of approximately 10 m in general. The position of sample trees could not be accurately matched with CHM, and there was a large displacement. The canopy height widely varied, showing significant discrepancy. Consequently, the error of CHM could not be accurately determined.
Dynamic extent
In 2015, the area of the mangrove patch was 8.16 ha (Fig. 2a). With an average annual growth rate of 3.7%, the patch area increased to 8.79 ha by October 2017 (Fig. 2a), accounting for an increase of 7.9%.
From 2015 to 2017, an area of 81,540 m2 in the patch remained stable. Nevertheless, the patch expanded by 6,356 m2 and shrunk by 19 m2 (Fig. 2c).
Due to the limitation imposed by the seawall, most of the patch expansions were sea-oriented (Fig. 2c, Fig. 3a1 and Fig. 3a2), and the maximum distance of expansion was 24 m (Fig. 3a3). The only exception is a landward extension of 89 m2 at the north-west end, filling the gap between the mangrove patch and the seawall. As no artificial afforestation was implemented in the study area, the process of patch expansion was natural. The patch expansion mainly occurred in the eastern part below the central part of the patch, and the expansion in northwestern part of the patch was very small. One possible explanation was that the water of the Naso River in the northern part of the study area and the ocean current during the ebb tide were affected by the branches and leaves of mangroves as they flow through the mangroves forest in the northwest of the patch. Blocked by stem and breathing roots, river and ocean currents were slowed down (Kathiresan 2003; Kathiresan and Rajendran 2005; Swales et al. 2007; Ong and Gong 2013). Sediment carried by rivers and ocean currents formed fine sand deposits in the southeastern part of the patch, resulting in rapid flattening of seafloor topography. This scenario created favorable conditions for the expansion of mangroves (de Boer 2002).
During the field investigation, local residents were found to collect Concha ostreae and raise ducks in the mangrove patch, but no evidence of artificial afforestation or other destructive human activities (such as clearing mangroves to build shrimp ponds) was found through image interpretation and field investigation. Therefore, the expansion and shrinkage of the patch occurred naturally. In other words, the death of the mangrove forest and destruction of the mangrove habitat were not attributable to human interference. The possible cause of shrinkage is the death of trees due to pests and diseases, as shown in Figs. 3b1, 3b2 and 3b3.
Spatial pattern of canopy height of patches and its evolution
CHM and d-CHM of the patch in 2015 and 2017 are shown in Fig. 4a1, Fig. 4b1, and Fig. 4c1, respectively. Their local 3D views are shown in Fig. 4a2, Fig. 4b2, and Fig. 4c2, respectively. UAV remote sensing exhibited a clear advantage over traditional field investigation. The conventional field investigation provided only a limited number of sample tree height, whereas UAV remote sensing provided patch CHM and d-CHM at a resolution of 1 m × 1 m. In addition to the stepless amplification and 3D view mentioned above, UAV remote sensing could also be used for canopy height classification and many other applications. It intuitively reflected the whole landscape of the patch and the distribution of the tree height in each section at the local scale.
In Fig. 4a1 and 4b1, the canopy height of each section of the patch appear to widely vary irrespective of the year (2017 or 2015). The canopy was higher in the northwestern section than in the southeastern section. Applying the spatial statistical method, the CHM of the two years was counted. The average height of the patch canopy was 2.8 m in 2015 and 3.4 m in 2017, as shown in Table 2.
Table 2
Statistics of crown height in 2015 and 2017
Year
|
Item
|
Total
|
0.1-1.0 m
|
1.1-2.0 m
|
2.1-3.0 m
|
3.1-4.0 m
|
4.1-5.0 m
|
5.1 m-
|
Mean (m)
|
2015
|
area (m2)
|
82588
|
461
|
16139
|
27683
|
30893
|
7367
|
45
|
2.8
|
%
|
100.0
|
0.6
|
19.5
|
33.5
|
37.4
|
8.9
|
0.1
|
|
2017
|
area (m2)
|
88856
|
126
|
23497
|
6584
|
23797
|
30058
|
4794
|
3.4
|
%
|
100.0
|
0.1
|
26.4
|
7.4
|
26.8
|
33.8
|
5.4
|
|
It can be seen from Table 2 that there is a great difference in canopy height in the two years, which is reflected in the large difference in area percentage of each height class, especially in the two height classs of 2.01–3.00 m and 4.01–5.00 m.
From Fig. 4c1, it could be concluded that the canopy height of patches increased generally from 2015 to 2017, with decrease in some areas. Canopy height increases of more than 1.0 m accounted for a large area, and those more than 1.5 m accounted for a small area, with a maximum increase of up to 3.2 m. Figure 4c1 shows that 88.2% of the patch area experienced canopy height increase, while 11.8% of the patch area experienced canopy height decrease from 2015 to 2017. Canopy height increased by more than 0.5 m in 70.9% of the patch area, and by more than 1.0 m in 38.9% of the patch area increased. In contrast, it decreased by more than 0.5 m in 5.0% of the patch area, as shown in Table 3.
Table 3
Range of
change (m)
|
Total
|
-3.2–1.5
|
-1.5–1.0
|
-1.1–0.5
|
-0.5–0.0
|
0.0-0.5
|
0.5-1.0
|
1.1–1.5
|
1.5–3.2
|
Mean (m)
|
Area (m2)
|
88855
|
932
|
1031
|
2490
|
6069
|
15369
|
28432
|
19743
|
14789
|
0.78
|
%
|
100.0
|
1.0
|
1.2
|
2.8
|
6.8
|
17.3
|
32.0
|
22.2
|
16.6
|
|
In the GIS software environment, six vertical profiles (a1 to a6, distance between each profile line was 90 m to 130 m) were set up from northwest to southeast. These profiles were basically perpendicular to the seawall; three horizontal profiles (b1 to b3, at distances of 19 m, 43 m, and 72 m, respectively, from the seawall) were also generated and they were basically parallel to the seawall. In each vertical and horizontal profile line, 100 evenly located sample points were set. The heights of all 100 sample points in the CHM of 2015 and 2017 were extracted and the canopy height curve of each section is shown in Fig.5.
Figure 5a1–5a6 show that the canopy height exhibited a decreasing trend from inside (near the seawalls) to outside (to the seaward) longitudinally (perpendicular to seawall) in both 2015 and 2017. The height of trees was higher near the seawall than in offshore areas. This trend was not as clear in profile line a6 as it was close to a culture pond in the southeast (Fig. 2a and 2b). This trend is different from that of a mangrove patch in the West Alligator River in northern Australia, where the mangroves in the middle section were higher than those seaward and landward (Lucas et al. 2002). Horizontally (parallel to the seawall), the canopy height on the landward side (Fig. 5b1, about 19 m from the seawall) and the seaward side (Fig. 5b3, about 72 m from the seawall) showed a decreasing trend from the northwest to the southeast in both years. In other words, the trees were higher in the northwest and lower in the southeast. However, this regularity was not obvious in the middle of the patch (Fig. 5b2).
Figures 4c1 and 5a1–5a6 show that, in general, the canopy height increase from 2015 to 2017 was slightly higher in mangroves on the seaward side than those on the landward side. Nevertheless, this trend was not obvious. The increase of canopy height was similar in some areas (Fig. 5a4), but it widely varied in most areas (Fig. 5a2, 5a3 and 5a5). In some areas, the canopy height increase of mangroves on the seawall side was greater than those on the seaward side (Fig. 5a3 and Fig. 5a6). In other sections, the canopy height increase on the seaward side was larger than that on the seawall side (Fig. 5A1 and Fig. 5a5). Longitudinally, the canopy height increase exhibited a strong regularity. The crown height increase during 2015–2017 was more than 1 m in most areas from 210–360 m and in some areas from 415–475 m to the northwest of the patch and in some areas from 10–70 m to the southeast of the patch, and also the newly formed mangrove, especially along the middle profile line (Fig. 5b2). The decrease in canopy height did not exhibit any spatial regularity. Canopy height decreased in many areas of the patch, with a maximum decrease of 3.1 m. In addition, there was no linear relationship between canopy height increase during 2015–2017 and that in 2015.
The regular distribution of canopy height increase in the patch might be related to the time of the formation of the mangrove patch and the nutrient content of the bottom soil. The following possible explanations are proposed: 1) The patch was originally on an open beach. After the seawall was built in 1969, the Naso River and ocean currents were slowed down by the seawall, during which sediment deposition began in front of the seawall. This led to the gradual rise in the seafloor. After the seafloor was elevated above the average sea level, a large number of drifting propagule (Hypocotyl of Avicennia marina mainly) from the Naso River and ocean currents began to take root, settled, and grew in front of the seawall. As a result, the mangroves first formed in front of the seawall, and then gradually spread seaward (Li and Zhou 2017). The height of the canopy in the coastal area is higher than that of those on the seaward side. 2) Likewise, the mangrove formation time in the northwest section near the Naso River estuary was earlier than that in the southeast section. Consequently, the height of the tree canopy was higher in the northwest section and slightly shorter in the southeast.
The main reason for the decrease in canopy height might be the death of forest due to pest infestation. Regional pest disasters frequently occur in the study site and its surrounding areas due to the marine pollution and disappearance of large area primary vegetation in the adjacent coastal zone, which were replaced with Eucalyptus plantation, and other factors. The mangroves of this path were damaged by Hyblaea puera in September and October 2015, and was seriously damaged by Oligochroa cantonella in May 2016 (Fan and Wang 2017), and about 80% of the leaves were found to be eaten during the field work in August 2016. From afar, the mangrove forest appeared yellow in color (Li and Zhou 2017). During field investigation for sample trees in October 2017, about one-third of the trees were found to be dead in some areas, leaving large gaps in the mangrove forest. By comparing the two temporal DOM, it can be found that the forest in the patch was very prosperous in 2015 (Fig. 6a), but by 2017, due to the death of a considerable number of trees in the middle of the patch, the dead trunks were clearly visible on the image, and many gaps with a diameter of more than 5 m appeared (Fig. 6b), resulting in a significant decrease in the canopy height (Fig. 6c).
The increase of canopy height was the result of mangrove forest growth. As shown in Fig. 7, the tree density increased in 2017 and the tidal channels became smaller, compared to those in 2015.
According to the remote sensing images, the trees were thriving in most area in 2015 and 2017, but the increase of canopy height was so high that it is difficult to establish a specific and reasonable explanation. Possible reasons are as follows: 1) During the monitoring period, some area had soil rich in nutrients, promoting the growth rate of trees above the normal level. 2) In October 2017, the wind speed was high during UAV photo acquisition. The wind might have blown the branches upwards, increasing the apparent height of the tress. It is noteworthy that the mangrove habitat is in great danger, and the growth of mangroves is slow. Therefore, further study and analysis are required to fully understand the actual cause of the increase.