Study area
The survey was conducted in the riparian area of the Akigawa River (Fig. 4), covering a regulated section spanning 33.57 km with a catchment area of 101.9 km2, managed by the Tokyo Metropolitan Construction Bureau (Tokyo Metropolitan Construction Bureau, https://www.kensetsu.metro.tokyo.lg.jp/jimusho/nishiken/kanri-ka/4-kasen/kasen.html, accessed on August 30, 2023). This study focused on the area near the confluence with the Tama River, considered a midstream region with relatively stable environmental factors, including riverbed gradient, that can influence plant growth.
Upon reviewing monthly precipitation data from Hachioji City, the closest meteorological observation point to our survey site, it was evident that the monthly rainfall for October 2019 substantially exceeded 600 mm, indicating heavy rainfall associated with a typhoon (Appendix S1 Fig. S1). Additionally, between June and August in 2019, 2020, and 2021, there were smaller peaks in rainfall, suggesting the presence of seasonal rain (Appendix S1 Fig. S1).
Study area location: Akigawa River.
Evaluation of disturbance impact
We assessed disturbance impact by categorizing land cover types within river channels, specifically bare ground, grassland, and forest, before and after the large-scale flooding event in October 2019, as well as one year (2020) and two years (2021) postflooding. Although directly quantifying flood disturbance in riverine areas can be challenging, the presence of bare ground within river channels, a result of flooding effects, serves as a reliable indicator of such disturbance3,16,33.
Aerial photographs, captured between June and November 2019, were used to measure areas of land cover, with such photographs of the river preflooding sourced from GSI Maps (https://maps.gsi.go.jp, accessed on August 30, 2023). Although the typhoon passed in October 2019, the aerial photographs showed walking paths and bridges in their pretyphoon state, indicating that the images were captured prior to the typhoon occurring. Additionally, post-typhoon aerial photographs were obtained from GSI Maps, featuring images from October 13, 2019, i.e., shortly after the typhoon (https://maps.gsi.go.jp/development/ichiran.html#t20191012typhoon19_tamagawa_1013do, accessed on August 30, 2023). Aerial photographs taken one year post-typhoon (captured in August 2020) were acquired from NTT InfraNet Corporation’s GEOSPACE aerial photographs (https://www.ntt-geospace.co.jp/geospace/koukuu.html, accessed on August 30, 2023). Finally, aerial photographs taken 2 years after the flood (May 2021) were obtained using a multicopter DJI Phantom 4 Pro (https://www.dji.com/jp/phantom-4-pro, accessed on August 30, 2023), and an orthophoto was created using Pix4Dmapper ver.4.6.4 (https://www.pix4d.com/product/pix4dmapper-photogrammetry-software/, accessed on August 30, 2023). Although the shooting months varied each year, no significant differences in land cover were expected due to the timing of photographs, as they were all taken during the plant growing seasons. The aerial photographs were interpreted, and terrestrial areas within the river channel were classified into three land cover categories: bare ground, grassland, and forest (dominated by woody plants). Polygon data were generated using QGIS 3.10 (https://qgis.org/en/site/, accessed on August 30, 2023).
To quantify the changes in land cover as an indicator of disturbance impact, survey units were established. Gregory et al. (1991) recommended that a section length of 10–100 times the channel width is an appropriate spatial scale to assess the impact of natural disturbances on river structure19. In this study, the survey area predominantly featured a channel width of 20–30 m. Therefore, buffer polygons measuring 300 m in width were established at approximately 200 m intervals from stream centerline data, sourced from the Ministry of Land, Infrastructure, Transport, and Tourism’s National Land Numerical Information (https://nlftp.mlit.go.jp/ksj/gml/datalist/KsjTmplt-W05.html, accessed on August 30, 2023). Each buffer polygon was subdivided into north and south sections (Fig. 5), resulting in a total of 52 units for analysis. These divided buffers were used as units for land cover and subsequent field surveys (as described later).
Survey unit description. Each unit spans 300 m in width at approximately 200 m intervals from the stream centerline.
The bare ground, grassland, and forest areas in each unit were derived from the created polygons. To quantify changes in land cover over time, we calculated the areas of each category in each unit, both before and after the typhoon in 2019, 2020, and 2021. Areas of land cover that were water or other types, such as agricultural land, artificial land, and constructions, were excluded from the analysis.
Classification of large-scale flooding impact
We categorized survey units into three classes based on the disturbance intensity caused by large-scale flooding: high, medium, and low intensity. This classification was based on a comparison of land cover data from before and after the 2019 typhoon, serving as a direct reflection of the large-scale flooding event’s impact. Classification criteria were based on the loss of natural terrestrial area, encompassing the total area of the three land categories we established. If > 50% of the land area was lost due to the flooding, the unit was designated as high intensity. For land area loss between 25% and < 50%, units were assigned medium intensity. If land area loss was < 25%, the unit was classified as low intensity.
Classification of periodic flooding impact
Further classification of survey units into three classes, within each large-scale flooding impact category, was based on the disturbance intensity caused by periodic flooding. This classification involved the use of land cover data from 2020 and 2021, comparing land cover changes between years. As no large-scale flooding occurred in these years, land cover changes were indicative of the effects of periodic disturbances. Classification criteria were chosen according to changes in natural terrestrial area, considering the total area of the three land categories for this period. Differences in the natural terrestrial area of each unit between 2020 and 2021 were calculated and ranked within each large-scale flooding intensity category. Units with the largest third of differences were designated as high intensity, the next third as medium intensity, and the smallest third as low intensity. Consequently, each of the three large-scale flooding intensity categories had three periodic flooding intensity categories within them.
Vegetation survey
Vegetation surveys were conducted in each survey unit from early August to early September in 2020 and from late July to late August in 2021. The survey involved establishing a 20 m line transect from the boundary between the river flow and the terrestrial area in each unit, and recording the plant species along the transect. In each survey transect, we recorded the occurrence of annual plant species, perennial plant species, and other plant species, including woody species. The surveys focused on the presence of species. For each unit in each year, we counted the total species number, annual species number, and perennial species number as an α-diversity index. We also calculated the Jaccard index based on the composition of plant species among each unit as a β-diversity index. The Jaccard index, J, was calculated as follows:
$$J = \frac{{S_{AB} }}{{S_{A} + S_{B} – S_{AB} }},$$
(1)
where SA and SB represent the species number in site A and B, respectively, and SAB indicates the common species number between site A and B. The Jaccard index is bounded between 0 and 1, with 0 indicating no shared species between the two sites, and 1 indicating identical species composition. We calculated the Jaccard index for transect pairs in the same class of large-scale disturbance and in the same class of periodic disturbance within the large-scale disturbance class. The former reflected β-diversity between large-scale disturbances, whereas the latter reflected β-diversity between the periodic disturbance classes within the same class of large-scale disturbance.
Statistical analysis
To evaluate the impact on natural terrestrial area by both large-scale and periodic disturbances across the entire study area, we used a generalized linear model (GLM) with Gaussian distributions (identity link) and a Tukey’s multiple comparison test to compare the areas of the three types of land cover among study periods. The explanatory variables were years; these were 2019 before the typhoon, 2019 after the typhoon, 2020, and 2021. If significant differences among these years were detected, we compared land cover areas between before and after the 2019 typhoon, between after the 2019 typhoon and 2020, and between 2020 and 2021 for each combination in each land cover type. To evaluate the influence of periodic disturbances on terrestrial area in each land cover type in each class of large-scale flooding impact, we compared the areas of the three land cover types between 2020 and 2021 using the same approach.
To assess the influence of large-scale disturbances on plant communities in the riparian area, we examined the total species number, annual species number, and perennial species number in each unit using generalized linear model (GLM) with Poisson distributions (log link) and a Wald test. The explanatory variable was the large-scale disturbance category (low, medium, and high) in the survey units. Given the categorical nature of the explanatory variable, the medium level served as the reference standard. Similarly, we assessed the Jaccard index in each unit using a GLM with Gaussian distributions (identity link) and a Wald test. The explanatory variable was the large-scale disturbance category in the survey units. Additionally, to evaluate the influence of plant communities resulting from the interaction between large-scale and periodic disturbances, we conducted a similar analysis, using periodic flooding classes in each large-scale disturbance class. For example, in the high-intensity class of large-scale disturbance, three classes of periodic disturbance were present. Thus, each periodic disturbances class was defined as follows: (1) high–high, (2) high–medium, and (3) high–low intensities (referring to the large-scale disturbance–periodic disturbances combination). We analyzed both species numbers and Jaccard index using GLMs based on periodic flooding classes as the explanatory variable in the survey units.
All statistical analysis were performed using the R statistical package (version 4.4.2; R development core Team, https://www.r-project.org/, accessed at 30, November, 2024).
