Large wood remobilization in Asakura (North Kyushu, Japan): Adapting strategies to climate change and rural population depletion

In the aftermath of the heavy rainfall events of 2017 in North Kyushu, Asakura (Japan), driftwood-related disaster risk has proven to be as real as it was in 1933, when wood-laden debris-flow invaded the city centre of Kobe City near Sannomiya station. Despite a temporary truce obtained thanks to extensive Sabo dam constructions, climate change, the return of forest on agricultural land due to rural exodus and an ageing population are once again increasing driftwood hazard and disaster risk. In the face of these novel challenges, new management strategies may need to be considered, and the present contribution aims to propose a new approach to driftwood hazards and disaster risk. For this purpose, the present contribution is analysing the 2D hydrodynamic of scenario-based floods on driftwood deposited after the 2017 heavy-rainfall event in Asakura. The boundary conditions for the simulation were generated using UAV photogrammetry calibrated against existing DEM data to generate a DEM at 25 cm horizontal resolution and an orthophotographs at 5 cm resolution. The results show that with the new river configuration (post-2017), an instantaneous peak flow of 90 m3/s is necessary to flood the areas where the driftwood has stopped with a water depth of >40 cm. This is 9 times the mode discharge calculated from the inlet geometry. The majority of the driftwood is thus not an immediate hazard, and leaving the wood instead of removing it should be considered as a management strategy whenever it is feasible. Recent research in Western Europe and Northern America has already shown the importance of driftwood for biodiversity and wildlife, and combining both environmental and hazards and disaster risk objectives may be a solution. For this purpose, the authors propose that medium-size ditch and artificial “abandoned channels” could be created in order to trap the wood in the floodplain, so that the cost from removing the wood can be alleviated, horizontal trapping could complement vertical trapping at slit dams, and the local wildlife could benefit from the wood decomposition in the floodplain.


Introduction
Large wood is the portion of the driftwood transported by overland flow and coastal waves, which is characterized by single trees and tree-stems, by opposition to accumulations of small branches, foliage (including single elements and lumps).Thus, it is often defined as large pieces of wood > 0.1 m in diameter.In Northern America and in Europe, research on driftwood benefit for the ecological system and the environment has progressed in the recent decades [1,2], but this approach has seen less traction in East Asian nations with short floodplains and dense population, as the disaster-risk still trumps ecological priorities.Those risks are often structural (bridges, weirs… [3,4,5], but the absence of any reliable warning system (except the radar-based in-flow detection system proposed by Gomez et al., [6]), has led to catastrophic impacts [7].In the present contribution, the authors are only interested in the large wood travelling in river floods.Large wood can be set in motion during a flood event due to bank erosion, slope failures and debris-flows linked to the flooding channel, and contribution from tributaries.In tectonically active countries, earthquakes can also be a source of driftwood [8].Besides purely natural processes, anthropogenic modification of the landscape [9] and landcover change due to population depletion in rural areas are also increasing sources of large wood [7].
In mid-and high-mountain of Japan, local settlements in the shape of hamlets are now disappearing from the landscape [10].Before ageing became a problem for Japan, this trend had already been recorded from 1960 to 1980 with a recorded 1712 lost hamlets [11].As population migrated to urban areas for work, vacant dwellings and farmland management issues have thus emerged [12].
In Island-arc countries of East-Asia (Indonesia, Japan, the Philippines), short-plain and highmountain ranges and volcanoes combined with high-population densities have led to channel-based strategies, where the large wood can be stopped by letting the water flow through by the use of slit-dam.However, as population in rural areas is decreasing in Japan, horizontal strategies can now be implemented as well, and instead of blocking large wood, it can also be driven to retention zone on the floodplain, as first introduced by Schmocker and Weitbrecht in USA [13].The objective of the present research is thus (1) to define the discharge level that brought the large wood to stop on the floodplain and (2) to investigate the potential for a large wood trapping solution within the study area.

Research Location
On July 5 th -6 th 2017, heavy rainfalls fell in North Kyushu, in the Asakura area.Daily rainfall reached 516 mm, when monthly rainfalls only reached 600 mm to 800 mm for the wettest months of the last 40 years [7].In turn, this event generated more than 1,500 mass movements.An estimated 3.9 to 13.4 million m 3 of sediments [14] were trasnported from the mountainous areas, eventually reaching the Chikugo River (Fig. 1).At the exit of the Myoukengawa river catchment, the authors selected a river bend where large wood stopped over the true right pointbar (Fig. 2) to conduct the different simulations.
During the 2017 heavy-rainfall event, large wood recruitment did not occur predominantly from terrace and river bank erosion, because those are mostly directed to agriculture production, and the wood entered the river from debris-flows and other mass movements (Figure 3-1).The trees mobilized on the slopes travelled as stems with sometimes the roots still attached (Figure 3-2 & 3).In combination with the water flow, the sediments and large wood, the event destroyed bridges (Figure 3

Methodology
To quantify the discharge that deposited the large wood in the study area and then test the inclusion of a retention zone on the true right side of the river, first a finite volume estimation of the equation of St-Venant (also known as the shallow-water equation) were calculated over a triangular grid using the mass conservation equation (eq. 1) and the momentum conservation equations (eq. 2 and 3) in the two horizontal orientations [15] , where h is the depth, u and v are the horizontal velocity, m is the mass conservation term,   is the shear in the channel direction and   in the cross-channel direction, g is the gravitational acceleration and  is the density (for more details see [15]). ( (2) (3) This first simulation was tested to obtain values of velocities and depth over the pointbar where the large wood was deposited.This simulation was performed by using inlet discharge between 10 and 80 m 3 /s, in order to identify the necessary discharge to flood the area of interest.Then a second set of simulation using SPH (Smooth Particle Physics) was performed to test the feasibility of creating a secondary channel in the outer bend of the channel, in order to trap wood debris.
The SPH solver used in the present study is the one of DualSPHysics code [16], where the shallowwater equation are rewritten in a meshless system using smoothing kernels in space for SPH purposes.This simulation was ran over an .stlfile generated in 3DSmax to generate a simplified boundary condition.The secondary channel design was developed using a secondary crest system to temporarily trap large wood (Fig. 5).

Results
The results present in a first step, the analysis of discharge (velocity and flow depth) to deposit the large wood found in the aftermath of the 2017 heavy rainfall event, and in a second step, it investigates the potential of a secondary trapping channel as a test study.

Deposition and remobilization discharge
At the study areas, clump of large woods have been deposited over the floodplain pointbar, and using the upstream dam inlet to simulate a flow entry in the channel, different levels of flooding have shown that a minimum of 50 m 3 /s peak discharge (Fig. 6) was necessary to flood the pointbars, generating depths of 60 cm above the pointbar, which can be considered as sufficient depths to deposit large wood stems (Fig. 6 and 7).

Conceptual testing of a secondary trapping channel using SPH simulation
Using the space of the pointbar, converted to a lower large wood trapping area (as in figure 5), and a peakflow of 50 m 3 /s hydrograph, the flow behaviour in the channel shows that the flow can either flow out in the trapping area at the front of the flow (Fig. 8) or when the flow reach peak discharge at 200 seconds (Fig. 8).

Figure 8
Simulation result for a single-peak flow using the SPH in a channel sized upon the sample area (50 m length in the inner bed.).

Conclusions
The preliminary results shown in this study proposes that scientists and engineer can back-calculate the flow discharge based on the position of the deposited large wood, to then calibrate an engineering solution.Such approach allows for a reduction of the number and depth of expensive laboratory trials, by using field data into a computer simulation.
-4 & 5)  as well as clogged check dams (Figure3-6 &7).Consequently, developing new and complementary appraoches to large wood management during flood has appeared as an essential task.

Figure 1 .
Figure 1.Research location in North Kyushu at the exit of the Myougengawa river catchment.

Figure 4 .
Figure 4. Series of Zooms on the study location showing the large wood trapped on a pointbar on the true right of the Myougengawa River, just downstream of a retention pond.

Figure 5 .
Figure 5. Floodplain fitted with a secondary trapping channel and metallic combs across the trapping channel, to keep the wood floated in the trapping channel.

Figure 6 .
Figure 6.Flow depth in the channel and the floodplain for hydrographs with a peak discharge comprised between 15 and 75 m 3 /s.

Figure 7 .
Figure 7. Instantaneous flow velocity in the channel and over the floodplain for simulations with inlet discharge comprised between 15 and 75 m 3 /s. :