9:05 AM - 9:30 AM
[ACG42-01] Nationwide estimation of the seagrass/algal bed distribution and their connection to terrestrial environments: Socioeconomic/Climate scenarios, Small Watersheds, and Land-use change within catchments.
★Invited Papers
Keywords:river basin, Coastal Environmental Management, GIS(Geographical Information System), remote sensing, Ecosystem Service
There have been many studies on the link between terrestrial and coastal ecosystems, but few have combined studies on a broad scale. Therefore, I would like to review the three studies we have conducted so far and discuss the direction of the research.
The first is a broad study of coastal changes, taking into account only large watersheds throughout the country, using four future scenarios (Kumagai et al. 2021).The second is the estimate of nationwide seagrass/algal bed beds distribution considering small watersheds by combining analysis of small watersheds and large watersheds (Yamakita et al. in prep).The third is the trial of the examination of the effect of terrestrial change in the basin unit using the Universal Soil Loss Equation(USLE) model (Yamaktia and Imaki 2019; Yamakita and Imaki Unpublished).
For the first part, we created a nationwide machine learning model by gradient boosting method using the area of eelgrass beds and seaweed beds as response variables. Seawater temperature, coast length, chlorophyll a, cropland ratio, wave height, tidal range, was used as static explanatory variables. Population and water quality were also used as explanatory variables based on scenarios. Based on the 4 national socio-economic scenarios, and national population and land-use change we predicted water quality at the river mouth. Based on these results, data were prepared by interpolating water quality. Data on climate change are also available.
The results suggest that, without climate change, the present seagrass beds would remain in a dispersed population society using artificial capital. On the other hand, the seagrass beds in Hokkaido would decline in a compact city population scenario using natural capital, probably due to a decrease in nutrients caused by a decrease in the local population. This decrease was about 2% of the seagrass area. When considering climate change the area could be reduced by an additional 3%. When these results were plotted on a GIS map, changes were large along the seagrass beds in eastern Hokkaido, where currently distributed with high density. Because the value of this region was the edge of the estimation range it might have a large variance in the estimate. Therefore, while it was appropriate to compare the degree of impact of each scenario nationwide, it was necessary to consider smaller scales and variance for further detailed analysis.
For the second part, the effects of watershed size on modeling were estimated by GLM and GAM. The number of grids containing eelgrass and seaweed beds in a certain distance buffer from the coastline or river mouth was considered as response variables. The explanatory variables were land drivers (Precipitation, forest/farmland ratio, and watershed area) and marine drivers (Wind waves, depth gradients, and water quality). The results showed that marine drivers accounted for more than 70% of the model variance in the model for eelgrass distribution by large watersheds (more than 10,000 km2), whereas drivers in land areas contributed more in small watersheds (less than 10,000 km2). In small watersheds, forest rates had a positive effect on the number of eelgrass grids. This suggests that land inflows directly affect aquatic vegetation, but it is important to consider the size of the watershed. Compared to such results on seagrass, seaweed was strongly affected by seawater temperature.
For the third part, a simple estimation of soil erosion and potential runoff from the land was conducted in individual watershed units. Here, we compare the temporal changes of estimation using the Universal Soil Loss Equation (USLE) model in the Pacific coastal basin of Tohoku as an example. At first, in order to generate river data from topography, we newly generated river data from a 10 m DEM using Whitebox GAT (Yamakita and Imaki 2019). Based on this GIS data, the coefficients of R: rainfall, K: soil, L・S: topography, C: crop, and P: soil management are calculated. Among these variables, a simple method using NDVI was used instead of using crop and soil management coefficients. The results reveal potential soil runoff and the impacts of land-use changes on coastal areas. In the presentation, I would like to show the estimated change of soil input due to the change in vegetation (NDVI) calculated from Google Earth Engine before and after the Great East Japan Earthquake in 2011.
References
Kumagai et al. (2021). https://doi.org/10.1007/s11625-020-00891-x
Yamakita and Imaki (2019). https://doi.org/10.5918/jamstecr.28.54
The first is a broad study of coastal changes, taking into account only large watersheds throughout the country, using four future scenarios (Kumagai et al. 2021).The second is the estimate of nationwide seagrass/algal bed beds distribution considering small watersheds by combining analysis of small watersheds and large watersheds (Yamakita et al. in prep).The third is the trial of the examination of the effect of terrestrial change in the basin unit using the Universal Soil Loss Equation(USLE) model (Yamaktia and Imaki 2019; Yamakita and Imaki Unpublished).
For the first part, we created a nationwide machine learning model by gradient boosting method using the area of eelgrass beds and seaweed beds as response variables. Seawater temperature, coast length, chlorophyll a, cropland ratio, wave height, tidal range, was used as static explanatory variables. Population and water quality were also used as explanatory variables based on scenarios. Based on the 4 national socio-economic scenarios, and national population and land-use change we predicted water quality at the river mouth. Based on these results, data were prepared by interpolating water quality. Data on climate change are also available.
The results suggest that, without climate change, the present seagrass beds would remain in a dispersed population society using artificial capital. On the other hand, the seagrass beds in Hokkaido would decline in a compact city population scenario using natural capital, probably due to a decrease in nutrients caused by a decrease in the local population. This decrease was about 2% of the seagrass area. When considering climate change the area could be reduced by an additional 3%. When these results were plotted on a GIS map, changes were large along the seagrass beds in eastern Hokkaido, where currently distributed with high density. Because the value of this region was the edge of the estimation range it might have a large variance in the estimate. Therefore, while it was appropriate to compare the degree of impact of each scenario nationwide, it was necessary to consider smaller scales and variance for further detailed analysis.
For the second part, the effects of watershed size on modeling were estimated by GLM and GAM. The number of grids containing eelgrass and seaweed beds in a certain distance buffer from the coastline or river mouth was considered as response variables. The explanatory variables were land drivers (Precipitation, forest/farmland ratio, and watershed area) and marine drivers (Wind waves, depth gradients, and water quality). The results showed that marine drivers accounted for more than 70% of the model variance in the model for eelgrass distribution by large watersheds (more than 10,000 km2), whereas drivers in land areas contributed more in small watersheds (less than 10,000 km2). In small watersheds, forest rates had a positive effect on the number of eelgrass grids. This suggests that land inflows directly affect aquatic vegetation, but it is important to consider the size of the watershed. Compared to such results on seagrass, seaweed was strongly affected by seawater temperature.
For the third part, a simple estimation of soil erosion and potential runoff from the land was conducted in individual watershed units. Here, we compare the temporal changes of estimation using the Universal Soil Loss Equation (USLE) model in the Pacific coastal basin of Tohoku as an example. At first, in order to generate river data from topography, we newly generated river data from a 10 m DEM using Whitebox GAT (Yamakita and Imaki 2019). Based on this GIS data, the coefficients of R: rainfall, K: soil, L・S: topography, C: crop, and P: soil management are calculated. Among these variables, a simple method using NDVI was used instead of using crop and soil management coefficients. The results reveal potential soil runoff and the impacts of land-use changes on coastal areas. In the presentation, I would like to show the estimated change of soil input due to the change in vegetation (NDVI) calculated from Google Earth Engine before and after the Great East Japan Earthquake in 2011.
References
Kumagai et al. (2021). https://doi.org/10.1007/s11625-020-00891-x
Yamakita and Imaki (2019). https://doi.org/10.5918/jamstecr.28.54