2:15 PM - 2:30 PM
[HCG20-03] Calibration and validation for modelling of coastal landscape evolution during a glacial-interglacial cycle - a case study on the Kamikita coastal plain, NE Japan
Keywords:Landscape evolution model, Calibration and validation, The last glacial-interglacial cycle, Kamikita plain
In Japan, the disposal depth for intermediate-level radioactive waste is relatively shallow, at 70 m or more during the safety period of 105 years, as compared with geological disposal, which exceeds 300 m. Therefore, assessing the long-term safety of landscape evolution by uplift, erosion, and sea level change is essential. To consider the uncertainty of future sea level change, numerical simulations of landscape evolution models (LEMs) are beneficial. Since most LEM parameters cannot be directly measured, a method for determining the unknown parameters is necessary for a proper safety assessment.
In this study, we developed a framework for calibrating and validating LEM parameters using simulations over the past 125,000 years. The process consists of the following two main steps. 1) Calibration: We identified the best parameter set that align with constraints of past landscape evolution, including terrace erosion rates, buried valley depth, longitudinal river profiles, and width of erosional valley floors. For efficient calibration, we initially selected parameters for calibration based on Morris sensitivity analysis. We then estimated the optimal parameter set by approximating the constraint conditions using second-order polynomial surrogate models of the selected parameters. The accuracy of this approach was verified by comparing the simulated topography. 2) Validation: Using the calibrated parameters, we conducted simulation over the past 125,000 years in an independent watershed, which shares similar geomorphic and geological conditions.
We tested this framework in the Kamikita coastal plain (calibration domain: the Tokusari River watershed, validation domain: the Futamata River watershed). The Japan Atomic Energy Agency developed the LEMs coupling of hillslope and fluvial transport, tectonics, marine sedimentation, sea level and climate change, and lithology. In this study, we enhanced the LEMs in terms of the fluvial transport (using a detachment- and transport-limited model) and marine processes (coastal erosion and deposition of drift sand). As the input data, the paleotopography and uplift rate were estimated using the marine terraces (MIS5e, 7, and 9) which are widely distributed in the study area. The distribution of alluvial deposits was inferred from previously published borehole and sonic data. Several LEM parameters were established based on measured data; fluvial incision parameters (exponents in DLM/TLM) were derived from slope-area analysis, and erodibilities were estimated from published uniaxial compressive strength data.
Through calibration, we identified five key parameters: the coefficients of hillslope transport (linear diffusion), fluvial incision, coastal erosion, and climate change. The developed surrogate models approximated the simulation results with a coefficient of determination of 0.7 or higher. Result from simulations indicated that the deviation between reproduced and recent topography is within ±15 m for 80% of the areas in both the calibration and validation domains. The validity of the calibrated parameter set was further confirmed by reproducing buried valley features, marine terrace cliffs, and terrace erosion rates. Our next step is to assess future landscape evolution while considering the uncertainty of sea-level changes by extrapolating the calibrated parameter set.
This study was funded by the Secretariat of Nuclear Regulation Authority, Nuclear Regulation Authority, Japan.
References
[1] Tanigawa, S. et al., The topographic change simulation methods, Japanese Patent No. 5422833, 2013.
[2] Tanigawa, S. et al., Transactions, Japanese Geomorphological Union, 37(2), 2016.
In this study, we developed a framework for calibrating and validating LEM parameters using simulations over the past 125,000 years. The process consists of the following two main steps. 1) Calibration: We identified the best parameter set that align with constraints of past landscape evolution, including terrace erosion rates, buried valley depth, longitudinal river profiles, and width of erosional valley floors. For efficient calibration, we initially selected parameters for calibration based on Morris sensitivity analysis. We then estimated the optimal parameter set by approximating the constraint conditions using second-order polynomial surrogate models of the selected parameters. The accuracy of this approach was verified by comparing the simulated topography. 2) Validation: Using the calibrated parameters, we conducted simulation over the past 125,000 years in an independent watershed, which shares similar geomorphic and geological conditions.
We tested this framework in the Kamikita coastal plain (calibration domain: the Tokusari River watershed, validation domain: the Futamata River watershed). The Japan Atomic Energy Agency developed the LEMs coupling of hillslope and fluvial transport, tectonics, marine sedimentation, sea level and climate change, and lithology. In this study, we enhanced the LEMs in terms of the fluvial transport (using a detachment- and transport-limited model) and marine processes (coastal erosion and deposition of drift sand). As the input data, the paleotopography and uplift rate were estimated using the marine terraces (MIS5e, 7, and 9) which are widely distributed in the study area. The distribution of alluvial deposits was inferred from previously published borehole and sonic data. Several LEM parameters were established based on measured data; fluvial incision parameters (exponents in DLM/TLM) were derived from slope-area analysis, and erodibilities were estimated from published uniaxial compressive strength data.
Through calibration, we identified five key parameters: the coefficients of hillslope transport (linear diffusion), fluvial incision, coastal erosion, and climate change. The developed surrogate models approximated the simulation results with a coefficient of determination of 0.7 or higher. Result from simulations indicated that the deviation between reproduced and recent topography is within ±15 m for 80% of the areas in both the calibration and validation domains. The validity of the calibrated parameter set was further confirmed by reproducing buried valley features, marine terrace cliffs, and terrace erosion rates. Our next step is to assess future landscape evolution while considering the uncertainty of sea-level changes by extrapolating the calibrated parameter set.
This study was funded by the Secretariat of Nuclear Regulation Authority, Nuclear Regulation Authority, Japan.
References
[1] Tanigawa, S. et al., The topographic change simulation methods, Japanese Patent No. 5422833, 2013.
[2] Tanigawa, S. et al., Transactions, Japanese Geomorphological Union, 37(2), 2016.