Despite the high capabilities of data-based smart models for classification and clustering purposes, to the best of our knowledge, no previous research has explored their application in sediment tracking. In this respect, here, we explore the abilities of three classic machine learning (ML) models (e.g., k-means, k-means++ and k-medoids) and eight deep learning (DL) models (e.g., auto-encoder based data clustering (AEC), deep clustering network (DCN), deep embedded clustering (DEC), improved deep embedded clustering (IDEC), dip-based deep embedded clustering with k-estimation (DipDECK), DipEncoder, deep k-means (DKM) and N2D) for the clustering and classifying of 132 potential source samples for loess deposits in the Ili Basin, Central Asia. A stepwise discriminant function analysis was applied to calculate the source samples classification accuracy, based on different classification schemes presented by ML and DL models. The main goal is comparing the performance of the ML and DL models with the conventional source samples classification technique. The reclassification rate of the source samples classification for the different models was as follows: DipDECK and DKM > DEC, k-means and k-means++ > AEC, IDEC, N2D, and k-medoids > DCN > DipEncoder. Overall, all classic ML and DL models exhibited an accuracy above 90% (apart from DipEncoder with an accuracy of 89.4%), while all these models outperformed the conventional tests (DFA-based reclassification rate: 84%) used in the sediment source tracking studies. Two DL models (DipDECK and DKM) were selected as the most accurate for source samples classification (reclassification rate: 99.2%), opening a new era in sedimentation studies.
We present a robust three-dimensional shear-wave velocity model of the central-southern Tanlu Fault Zone (TLFZ) and adjacent regions in eastern China, derived from joint inversion of Common Conversion Point (CCP) stacking of teleseismic receiver functions and ambient-noise surface-wave dispersion data recorded by 145 broadband stations (2019–2023). The model reveals pronounced Moho depth variations, strong crustal and upper-mantle velocity heterogeneities, and marked along-strike segmentation of the TLFZ. North of Jiashan, persistent across-fault velocity contrasts extend to ~50 km depth, indicating a lithosphere-scale fault that facilitates stress accumulation and elevated seismicity, including the 1668 Tancheng M8.5 earthquake. Southward, high-velocity bodies near the Moho extending across the fault, eliminating the clear velocity contrast observed across the fault farther north. These high-velocity bodies are interpreted as more rigid structures that are less prone to deformation, which correlates well with the observed seismic quiescence. Mid-crustal low-velocity zones and across-fault anomalies suggest that the area has been affected by multi-phase tectonic reworking driven by Triassic collision, Mesozoic Pacific subduction, magmatic ascent, and Cenozoic compression. These findings illuminate the complex geodynamic evolution of the TLFZ and provide critical geophysical constraints for regional seismic hazard assessment.
The overall coarseness of marine sand and gravel (MSG) is a critical parameter for quality assessment, ore-body delineation and suitability evaluation. Current MSG exploration adopts the Udden-Wentworth scale for grain-size classification, whereas there is no dedicated coarseness index matched to this system. The fineness modulus (Mx) derived from the American Society for Testing Materials sieve series is widely applied for coarseness evaluation. This study developed a new index, the coarseness exponent (Ce), based on the Udden-Wentworth scale. The Ce value is 0.25 greater than the corresponding Mx value, and MSG is classified into five grades according to Ce: ultra coarse, coarse, medium, fine and ultra fine. Measured data from 111 sandy samples validated the capability of Ce to characterize the overall coarseness of MSG (R2 = 0.99, RMSE = 0.03). The practical performance of Ce and Mx was compared in MSG grading and ore-body delineation. The results show generally consistent classification and delineation outcomes for the two indices. Furthermore, Ce allows valid grading of ultra-coarse MSG that cannot be defined using Mx. Adopting Ce in MSG exploration unifies the sand-mud boundary at 0.063 mm and avoids the need for two independent grain-size analyses. Consequently, Ce exhibits promising prospects for MSG quality grading, ore-body delineation and exploitation-suitability assessment.
The Khourmo Cu deposit is located in the northeastern part of the Ravar Copper Belt (RCB), near the southwestern edge of the Tabas Block, situated within East-Central Iran. Copper exploration activities in this region have focused on three mining blocks: A, B, and C. The Khourmo deposit is primarily hosted by the Garedu Red Bed Formation (Fm.), which spans from the Late Jurassic to the Early Cretaceous and is composed of alternating layers of sandstone, micro-conglomerate, and siltstone. The sandstone units are subdivided into oxidized, mineralized, reduced, and bleached zones. The copper mineralization is mainly confined to the reduced gray sandstone layers, especially within ancient fluvial paleochannels, and is represented by a range of sulfide and secondary minerals, including chalcopyrite, bornite, digenite, chalcocite, covellite, pyrite, as well as supergene minerals such as malachite, chrysocolla, and various iron oxides. Geochemical analysis indicates that the sandstone Formations hosting the Khourmo Cu deposit derived from igneous parent rocks of intermediate to felsic composition on an active continental margin. Evidence such as the presence of permeable host rocks, the formation of bleached, oxidized, and reduced mineralized zones, the Cu mineralization within paleochannels, the association of copper with plant fossils, and the occurrence of salt domes beneath the Garedu Red Bed Fm. suggest that the Khourmo deposit represents a stratabound, sediment-hosted Cu mineralization. This type of mineralization likely developed immediately after primary diagenesis of the host rock and before deep burial.
There are a variety of multicriteria decision-making approaches to pinpoint the areas with significant mineral potential. Since exploration data need to be weighted or transformed into the same range before integration, and because there are multiple approaches to weighting, the chosen weighting strategy has a strong influence on the resulting prospectivity model. Consequently, a major challenge in mineral prospectivity mapping is the sensitivity of the integration methods to input data and the methods applied to generate weighted exploration evidence layers, which results in inconsistency in the recognized targets. In this paper, we perform a noise analysis to assess the sensitivity of integration and modeling results to the variation of weighted layers of the same input exploration data. For this, we use a dataset of porphyry copper mineralization in the Chahargonbad district within the Urumieh-Dokhtar Volcanic Belt of Iran, and we employ the root mean square (RMS) function as an alternative to the existing geometric average (GA) integration approach for mineral exploration targeting. Furthermore, the isolation forest (IForest) algorithm, an unsupervised anomaly detection method, is employed for the identification of anomalous patterns for comparison purposes. All three methods were evaluated using the prediction–area plot and the normalized density index, demonstrating the superiority of the RMS method over other approaches in identifying Cu mineralization. The proposed integration framework exhibits reduced sensitivity to noise and weighting variability in exploration datasets. The outcomes demonstrate that the RMS method outperforms both the GA and IForest approaches in recognizing exploration targets.
