The Value of Using Multiple Hydrometeorological Variables to Predict Temporal Debris Flow Susceptibility in an Alpine Environment

Prenner, David; Kaitna, Roland; Mostbauer, Karin; Hrachowitz, Markus

doi:10.1029/2018wr022985

Cited by 41 publications

(50 citation statements)

References 92 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, traditional statistical models are in general limited to certain requirements and assumptions on data, whereas machine learning methods are not. In the past three decades, common machine learning methods used for studying DFS mapping include back propagation neural network (BPNN) [20], decision tree (DT) [21], Bayesian network [22], and support vector machine [23]. With the advancement of researches, more and more models have been developed with better fitting performance.…”

Section: Introductionmentioning

confidence: 99%

Debris Flow Susceptibility Mapping Using Machine-Learning Techniques in Shigatse Area, China

et al. 2019

View full text Add to dashboard Cite

Debris flows have been always a serious problem in the mountain areas. Research on the assessment of debris flows susceptibility (DFS) is useful for preventing and mitigating debris flow risks. The main purpose of this work is to study the DFS in the Shigatse area of Tibet, by using machine learning methods, after assessing the main triggering factors of debris flows. Remote sensing and geographic information system (GIS) are used to obtain datasets of topography, vegetation, human activities and soil factors for local debris flows. The problem of debris flow susceptibility level imbalances in datasets is addressed by the Borderline-SMOTE method. Five machine learning methods, i.e., back propagation neural network (BPNN), one-dimensional convolutional neural network (1D-CNN), decision tree (DT), random forest (RF), and extreme gradient boosting (XGBoost) have been used to analyze and fit the relationship between debris flow triggering factors and occurrence, and to evaluate the weight of each triggering factor. The ANOVA and Tukey HSD tests have revealed that the XGBoost model exhibited the best mean accuracy (0.924) on ten-fold cross-validation and the performance was significantly better than that of the BPNN (0.871), DT (0.816), and RF (0.901). However, the performance of the XGBoost did not significantly differ from that of the 1D-CNN (0.914). This is also the first comparison experiment between XGBoost and 1D-CNN methods in the DFS study. The DFS maps have been verified by five evaluation methods: Precision, Recall, F1 score, Accuracy and area under the curve (AUC). Experiments show that the XGBoost has the best score, and the factors that have a greater impact on debris flows are aspect, annual average rainfall, profile curvature, and elevation.

show abstract

Section: Introductionmentioning

confidence: 99%

Debris Flow Susceptibility Mapping Using Machine-Learning Techniques in Shigatse Area, China

et al. 2019

View full text Add to dashboard Cite

show abstract

“…This concept was previously successfully applied in a wide range of environments (Gao et al, 2014;Gharari et 225 al., 2014;Fovet et al, 2015;Nijzink et al, 2016;Prenner et al, 2018).…”

Section: Hydrological Model Structurementioning

confidence: 99%

Using altimetry observations combined with GRACE to select parameter sets of a hydrological model in data scarce regions

Hulsman¹,

Winsemius²,

Michailovsky³

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

Abstract. To ensure reliable model understanding of water movement and distribution in terrestrial systems, sufficient and good quality hydro-meteorological data are required. Limited availability of ground measurements in the vast majority of river basins world-wide increase the value of alternative data sources such as satellite observations in modelling. In the absence of directly observed river discharge data, other variables such as remotely sensed river water level may provide valuable information for the calibration and evaluation of hydrological models. This study investigates the potential of the use of remotely sensed river water level, i.e. altimetry observations, from multiple satellite missions to identify parameter sets for a hydrological model in the semi-arid Luangwa River Basin in Zambia. A distributed process-based rainfall runoff model with sub-grid process heterogeneity was developed and run on a daily timescale for the time period 2002 to 2016. Following a step-wise approach, various parameter identification strategies were tested to evaluate the potential of satellite altimetry data for model calibration. As a benchmark, feasible model parameter sets were identified using traditional model calibration with observed river discharge data. For the parameter identification using remote sensing, data from the Gravity Recovery and Climate Experiment (GRACE) were used in a first step to restrict the feasible parameter sets based on the seasonal fluctuations in total water storage. In a next step, three alternative ways of further restricting feasible model parameter sets based on satellite altimetry time-series from 18 different locations, i.e. virtual stations, along the Luangwa River and its tributaries were compared. In the calibrated benchmark case, daily river flows were reproduced relatively well with an optimum Nash-Sutcliffe efficiency of ENS,Q = 0.78 (5/95th percentiles of all feasible solutions ENS,Q,5/95 = 0.61 – 0.75). When using only GRACE observations to restrict the parameter space, assuming no discharge observations are available, an optimum of ENS,Q = −1.4 (ENS,Q,5/95 = −2.3 – 0.38) with respect to discharge was obtained. Depending on the parameter selection strategy, it could be shown that altimetry data can contain sufficient information to efficiently further constrain the feasible parameter space. The direct use of altimetry based river levels frequently over-estimated the flows and poorly identified feasible parameter sets due to the non-linear relationship between river water level and river discharge (ENS,Q,5/95 = −2.9 – 0.10); therefore, this strategy was of limited use to identify feasible model parameter sets. Similarly, converting modelled discharge into water levels using rating curves in the form of power relationships with two additional free calibration parameters per virtual station resulted in an over-estimation of the discharge and poorly identified feasible parameter sets (ENS,Q,5/95 = −2.6 – 0.25). However, accounting for river geometry proved to be highly effective; this included using river cross-section and gradient information extracted from global high-resolution terrain data available on Google Earth, and applying the Strickler-Manning equation with effective roughness as free calibration parameter to convert modelled discharge into water levels. Many parameter sets identified with this method reproduced the hydrograph and multiple other signatures of discharge reasonably well with an optimum of ENS,Q = 0.60 (ENS,Q,5/95 = −0.31 – 0.50). It was further shown that more accurate river cross-section data improved the water level simulations, modelled rating curve and discharge simulations during intermediate and low flows at the basin outlet at which detailed on-site cross-section information was available. For this case, the Nash-Sutcliffe efficiency with respect to river water levels increased from ENS,SM,GE = −1.8 (ENS,SM,GE,5/95 = −6.8 – −3.1) using river geometry information extracted from Google Earth to ENS,SM,ADCP = 0.79 (ENS,SM,ADCP,5/95 = 0.6 – 0.74) using river geometry information obtained from a detailed survey in the field. It could also be shown that increasing the number of virtual stations used for parameter selection in the calibration period can considerably improve the model performance in spatial split sample validation. The results provide robust evidence that in the absence of directly observed discharge data for larger rivers in data scarce regions, altimetry data from multiple virtual stations combined with GRACE observations have the potential to fill this gap when combined with readily available estimates of river geometry, thereby allowing a step towards more reliable hydrological modelling in poorly or ungauged basins.

show abstract

“…The steep slopes provide loose material for debris flow [ 45 ]. The slope angle in the northwest of the study area is mainly 0–5°, while the slope in the southeast is generally near 10°, and, in a few areas, it is more than 20°.…”

Section: Data Preparationmentioning

confidence: 99%

A Method for Improving Controlling Factors Based on Information Fusion for Debris Flow Susceptibility Mapping: A Case Study in Jilin Province, China

Dou

Qin

Zhang

et al. 2019

Entropy

View full text Add to dashboard Cite

Debris flow is one of the most frequently occurring geological disasters in Jilin province, China, and such disasters often result in the loss of human life and property. The objective of this study is to propose and verify an information fusion (IF) method in order to improve the factors controlling debris flow as well as the accuracy of the debris flow susceptibility map. Nine layers of factors controlling debris flow (i.e., topography, elevation, annual precipitation, distance to water system, slope angle, slope aspect, population density, lithology and vegetation coverage) were taken as the predictors. The controlling factors were improved by using the IF method. Based on the original controlling factors and the improved controlling factors, debris flow susceptibility maps were developed while using the statistical index (SI) model, the analytic hierarchy process (AHP) model, the random forest (RF) model, and their four integrated models. The results were compared using receiver operating characteristic (ROC) curve, and the spatial consistency of the debris flow susceptibility maps was analyzed while using Spearman’s rank correlation coefficients. The results show that the IF method that was used to improve the controlling factors can effectively enhance the performance of the debris flow susceptibility maps, with the IF-SI-RF model exhibiting the best performance in terms of debris flow susceptibility mapping.

show abstract

The Value of Using Multiple Hydrometeorological Variables to Predict Temporal Debris Flow Susceptibility in an Alpine Environment

Cited by 41 publications

References 92 publications

Debris Flow Susceptibility Mapping Using Machine-Learning Techniques in Shigatse Area, China

Debris Flow Susceptibility Mapping Using Machine-Learning Techniques in Shigatse Area, China

Using altimetry observations combined with GRACE to select parameter sets of a hydrological model in data scarce regions

A Method for Improving Controlling Factors Based on Information Fusion for Debris Flow Susceptibility Mapping: A Case Study in Jilin Province, China

Contact Info

Product

Resources

About