Discussion

2.4.1. Overall Representativeness

The greater vegetation cover as indicated by both VIs, the overall gentler slopes, and lower altitude of the FFDP in relation to the FEF, and the significant differences in slope aspect between the FFDP and the FEF (Figure 2.3) all suggest that the plot does not adequately represent the greater landscape, both in terms of vegetation and topographic heterogeneity. Given the much smaller area of the FFDP (25 ha) compared to the FEF (1097 ha) and the rough mountain topography of the reserve, it is not surprising that the FFDP cannot adequately represent the topography of FEF.

However, although the FFDP covers only 10% of the elevation range of the FEF, it covers more than 50% of the landform variation (TPI) and slope steepness of the FEF.

32

The high coverage of the landform variability and slope steepness, however, does not lead to a high degree representativeness in vegetation as the range of VIs of the FFDP covers only 9% of the NDVI and 23% of the NDII ranges of the FEF (Figure 2.3). The lack of significant correlation between the topographical variables and vegetation indices within both the FFDP and the FEF (Table B1) explains that the moderately high topographical representativeness of FFDP for the FEF does not lead to high representativeness of vegetation indices.

Elevation is the only topographical factor examined that had any significant correlation with VIs. The negative relationship between elevation and VIs can partially explain the greater VIs in FFDP than the FEF because the FFDP has a considerably lower mean elevation than the FEF. Many environmental factors important to plant growth, such as temperature and precipitation vary with elevation such that vegetation cover is likely to vary with elevation. Notably, the correlation between elevation and VIs are stronger for the FEF than the FFDP. Elevation is likely a more important factor for structuring vegetation cover at the broader landscape than within the FFDP, which has smaller elevational range. Many studies illustrate that tropical forest tree species diversity decreases with elevation (Lieberman et al., 1996; Vázquez G. & Givnish, 1998; Chi et al., 2015), including those from northern Taiwan (Hsieh et al., 1998), the narrow elevation range within the FFDP probably limits its representativeness of the landscape tree species richness by overestimating its diversity and missing species that only grow at higher elevations (Hsieh et al., 1998). It has also been reported that forest NDVI is closely related to tree diversity (see review by Rocchini et al., 2015), thus the difference in NDVI between the FFDP and the FEF suggests that the FFDP could have different species composition from the greater FEF.

Higher values and narrower ranges of NDVI were positively related to LAI (White et al., 1997) and NDII, which were, in turn, positively related to vegetation water content, suggesting that the FFDP overestimates LAI (NDVI, White et al., 1997) and forest water status (NDII, Cheng et al., 2006) at the landscape scale, but underestimating their variability. Thus, the FFDP is unlikely a completely

33

representative sample for primary productivity and other important environmental processes at the landscape (i.e., FEF) scale.

2.4.2. Typhoons Damages Intensity in the Plot and the Reserve

The significant differences of typhoon-induced ΔVIs between the FFDP and the FEF (Table B4) suggest that the FFDP cannot adequately represent the disturbance effects of the greater FEF. The low representativeness of typhoon disturbance within the FFDP is also evident from the differences in the proportion of the cells affected with respect to typhoon frequency between the FFDP and the FEF (Figure 2.6). For both VIs, less than 9% of the cells of the FEF were hit by all of the five typhoons examined (Table B4), and the small size of the FFDP relative to the FEF certainly limit its ability to capture cells with high typhoon frequencies (Figure 2.6). Effects of topography on the cyclone damage distribution within forested landscapes are widespread across regions affected by tropical cyclones (Turton, 2008; Marra et al., 2014; Abbas et al., 2020). In this study, elevation was the only topographical variable that correlated with typhoon damages for three of the five studied typhoons, although relationships were weak (ρ < 0.2). This suggests that slope steepness and TPI, which are relatively well represented by the FFDP (Figure 2.3), have limited influence on forest sensitivity to wind disturbances at the 30 m resolution in the reserve and in the permanent plot. This contrasts with the report of lesser forest damage by disturbance in flatter surfaces of the FFDP in a ground survey of aboveground biomass (McEwan et al., 2011). More studies are required to examine if such differences are related to differences in the survey methods (i.e., remote sensing vs. ground-based approaches), typhoon variability, or scale. Notably, FFDP representativeness for typhoon disturbance effect is considerably better than its representativeness of overall vegetation cover, as it represents between 30% and 75.9% of the FEF IQRs associated with ΔVIs across the five typhoons, but only less than 7% of the FEF IQRs for VIs. Thus, despite covering a small fraction of the vegetation variability and elevational range of the greater landscape, the FFDP provides a relatively good representation of

34

disturbance effects occurring at the landscape scale. In addition, our results did not suggest any over- or under-exposure to cyclone damage within the FFDP relative to the FEF over the long term, because differences in ΔVIs are not consistent across typhoons. Moreover, even though statistically significant, the differences between FFDP and FEF may have limited ecological meaning as the ΔVIs values are often very close in absolute value (e.g., a difference of 0.01 of mean NDVI for Typhoon Herb, Table B2).

2.4.3. Vegetation Cover and Topographical Representativeness with Alternative Strategies Across all strategies, most of the alternative plot designs improved the representativeness of vegetation cover in comparison with the FFDP (lower minED values), although the representativeness of topographical parameters, slope steepness, and TPI decreased (Figure 2.7). However, the elevation range of the plot increased from less than 10% to over 30% of the FEF range due to the alternative plot designs.

The lack of consistency between vegetation and topographical representativeness could be explained by the lack of correlation between TPI, slope steepness, and the VIs. Moreover, the one-square strategy led to fewer replicates with improved vegetation representativeness than the other strategies with spatially separate subplots (Table C1), likely because its elevational range is narrower than the other strategies as elevation was significantly correlated with VIs. Nevertheless, the elevation range may not be the only factor explaining increased VI representation as dispersed subplots across the landscape may also provide a more representative sample of the vegetation diversity than a single large plot in landscapes with high spatial heterogeneity such as Fushan. Indeed, it has been reported that spreading multiple plots across a forest landscape led to increased representativeness of plot biomass and biodiversity patchiness in comparison with a single large plot (Salk et al., 2013).

2.4.4. Disturbances Representativeness with Alternative Strategies

The evaluation of the different plot design strategies with respect to disturbance effect led to an unexpected trend. Despite its smaller elevational range, the FFDP can

35

be more representative of the FEF (i.e., smaller minED) in terms of typhoon effects than alternative plot designs with greater elevational ranges. This was observed for Typhoons Dujuan, Herb, and Soudelor, which is 75% of cases (Typhoon Aere was not included in the analysis of alternative plot design strategies). On the other hand, several alternative plot designs showed improved ΔVI representativeness for Typhoon Nari (Figure 2.8). Since the same alternative plot placements were used to study all disturbances, the difference between Nari and other events may be the product of less cloud cover in the Landsat images used to study the effects of this typhoon. Clouds obstructed almost 50% of the FEF for Typhoons Dujuan and Soudelor, but less than 20% for Typhoon Nari. Clouds were mostly located at higher elevations and elevation has weak but significant correlations with ΔVIs; therefore, the representativeness of disturbance effects across the FEF within the alternative plot design strategies (e.g., minED values) may vary among typhoons because of varying cloud cover and elevational ranges of the observable area. Nevertheless, for Typhoon Nari, alternative plots had ΔVIs that were significantly different from the reserve and most remained underexposed just like the FFDP (Table C2). Thus, a complete representation of typhoon disturbances may not be attainable by using the current plot nor by alternative plot design strategies based on 25 ha surface but wider elevational ranges.

2.4.5. Comparison of Strategies

The comparison of the alternative plot design strategies based on minED for both the overall vegetation cover (four strategies, including one-square) and disturbance effects (three strategies) did not show an overall improvement in plot representativeness with respect to the FFDP (Table C3). However, the strategies that used multiple subplots ranked better than one large square and two rectangular plot strategies, which is in agreement with the observations of a previously published study by Salk et al. (2013). Furthermore, in contrast to the result from a simulative study by Reese et al. (2005), we do not find greater representativeness of rectangular

36

shaped plots (i.e., transects) relative to squares suggesting that the possibility of rectangular quadrats to cover a wider range of vegetation heterogeneity is not universal.