Application of newly developed ensemble machine learning models for daily suspended sediment load prediction and related uncertainty analysis

ABSTRACT

Ensemble machine learning models have been widely used in hydro-systems modeling as robust prediction tools that combine multiple decision trees. In this study, three newly developed ensemble machine learning models, namely gradient boost regression (GBR), AdaBoost regression (ABR) and random forest regression (RFR) are proposed for prediction of suspended sediment load (SSL), and their prediction performance and related uncertainty are assessed. The SSL of the Mississippi River, which is one of the major world rivers and is significantly affected by sedimentation, is predicted based on daily values of river discharge (Q) and suspended sediment concentration (SSC). Based on performance metrics and visualization, the RFR model shows a slight lead in prediction performance. The uncertainty analysis also indicates that the input variable combination has more impact on the obtained predictions than the model structure selection.     

Optimizing compaction calculation for improved probabilistic 2D mapping

Integration of all available data in reservoir characterization is critically important. 2D mapping is a reliable and robust technique that allows integration of multiple secondary data, including geological and geophysical surfaces and maps, to generate realistic summaries of reservoir quality at each location in an area of interest with an associated measure of uncertainty. This is achieved in 2D mapping with a more straightforward implementation, requiring significantly less time and fewer resources than three-dimensional modelling. In this paper, we propose an approach for the empirical calculation and optimization of differential compaction maps by leveraging existing well control for the use in 2D mapping. Success of the proposal is demonstrated through tests of accuracy, precision and fairness of the local uncertainty distributions for 100 new stratigraphical wells drilled in the Christina Lake and Kirby East area.     

Seismic risk assessment considering cumulative damage due to aftershocks

Calculating the limit state (LS) exceedance probability for a structure considering the main seismic event and the triggered aftershocks (AS) is complicated both by the time‐dependent rate of aftershock occurrence and also by the cumulative damage caused by the sequence of events. Taking advantage of a methodology developed previously by the authors for post‐mainshock (MS) risk assessment, the LS probability due to a sequence of mainshock and the triggered aftershocks is calculated for a given aftershock forecasting time window. The proposed formulation takes into account both the time‐dependent rate of aftershock occurrence and also the damage accumulation due to the triggered aftershocks. It is demonstrated that an existing reinforced concrete moment‐resisting frame with infills subjected to the main event and the triggered sequence exceeds the near‐collapse LS. On the other hand, the structure does not reach the onset of near‐collapse LS when the effect of triggered aftershocks is not considered. It is shown, based on simplifying assumptions, that the derived formulation yields asymptotically to the same Poisson‐type functional form used when the cumulative damage is not being considered. This leads to a range of approximate solutions by substituting the fragilities calculated for intact, MS‐damaged, and MS‐plus‐one‐AS‐damaged structures in the asymptotic simplified formulation. The latter two approximate solutions provide good agreement with the derived formulation. Even when the fragility of intact structure is employed, the approximate solution (considering only the time‐dependent rate of aftershock occurrence) leads to higher risk estimates compared with those obtained based on only the mainshock. Copyright © 2016 John Wiley & Sons, Ltd.     

Collocated Cokriging Based on Merged Secondary Attributes

There exist many secondary data that must be considered in in reservoir characterization for resource assessment and performance forecasting. These include multiple seismic attributes, geological trends and structural controls. It is essential that all secondary data be accounted for with the precision warranted by that data type. Cokriging is the standard technique in geostatistics to account for multiple data types. The most common variant of cokriging in petroleum geostatistics is collocated cokriging. Implementations of collocated cokriging are often limited to a single secondary variable. Practitioners often choose the most correlated or most relevant secondary variable. Improved models would be constructed if multiple variables were accounted for simultaneously. This paper presents a novel approach to (1) merge all secondary data into a single super secondary variable, then (2) implement collocated cokriging with the single variable. The preprocessing step is straightforward and no major changes are required in the standard implementation of collocated cokriging. The theoretical validity of this approach is proven, that is, the results are proven to be identical to a “full” approach using all multiple secondary variables simultaneously.     

Wave dispersion in high‐rise buildings due to soil–structure interaction

Nonparametric techniques for estimation of wave dispersion in buildings by seismic interferometry are applied to a simple model of a soil–structure interaction (SSI) system with coupled horizontal and rocking response. The system consists of a viscously damped shear beam, representing a building, on a rigid foundation embedded in a half‐space. The analysis shows that (i) wave propagation through the system is dispersive. The dispersion is characterized by lower phase velocity (softening) in the band containing the fundamental system mode of vibration, and little change in the higher frequency bands, relative to the building shear wave velocity. This mirrors its well‐known effect on the frequencies of vibration, i.e. reduction for the fundamental mode and no significant change for the higher modes of vibration, in agreement with the duality of the wave and vibrational nature of structural response. Nevertheless, the phase velocity identified from broader band impulse response functions is very close to the superstructure shear wave velocity, as found by an earlier study of the same model. The analysis reveals that (ii) the reason for this apparent paradox is that the latter estimates are biased towards the higher values, representative of the higher frequencies in the band, where the response is less affected by SSI. It is also discussed that (iii) bending flexibility and soil flexibility produce similar effects on the phase velocities and frequencies of vibration of a building. Copyright © 2014 John Wiley & Sons, Ltd.