Extreme Discharges and Thresholds of Boulder Mobility in Steep Mountainous Streams on Maui, Hawai'i

Abstract:

Quantifying the conditions required for boulder mobility in steep mountainous streams remains a relevant problem across spatial and temporal scales from land-management to landscape evolution. One challenge is that the floods capable of moving large boulders are infrequent and their impacts are rarely observed. In 2018 a series of large storms generated extreme floods on the island of Maui. Discharges from these storms estimated in the valleys of Honokohau and Waihe'e suggest flood sizes approximately 1.5 (345 m³s^-1 and 525 m³s^{- 1} respectively) times larger than the 100-year flood. Field observations in these valleys and the neighboring valley of Honolua provided an opportunity for assessing boulder mobility and the impacts of extreme floods. Evidence of boulder mobility included fresh woody debris pinned beneath boulders, precariously balanced and imbricated boulders, and lateral and mid-channel bars consisting almost entirely of fresh deposits of imbricated boulders. Across multiple reaches in each valley, we measured the intermediate axes of the largest mobile boulders. The mean largest mobile boulder observed for Honolua, Honokohau, and Waihe'e are 0.8, 1, and 1.2 meters in diameter, respectively. We used channel cross-sections marked by flood debris to determine the hydraulic radius and estimate τ_b (the boundary shear stress ranging from 700 to 1000 Pa). Assuming τ_b = τ_c, where τ_c is the threshold shear stress for the mean largest mobile boulder, preliminary analysis indicates the non-dimensional shields parameter τ_c^* ranges from 0.049 to 0.033. These results are interesting since they provide insight into the conditions of boulder mobility in steep mountainous streams (4 to 6%). We speculate that extreme floods may generate conditions, consistent with recent experimental work, where the dependence of τ_c^* on hydraulic resistance and thus the relative roughness may reduce τ_c^* to similar values observed in lowland streams and rivers. To evaluate if this possible and to allow comparisons to other settings we plan on using deep neural network algorithms and Structure from Motion surveys from lateral and mid-channel bars to analyze the spatial patterns and distributions of grain sizes to make informed estimates of τ_c^* for the median observed grain size.