Subsampling of geological data is a crucial step allowing us to visualize meaningful geological information as multiscale three-dimensional (3D) geological models. Multiscale 3D geological models allow us to better answer specific scale-dependent geological questions. To achieve this aim, we investigate the implications of different subsampling techniques in order to develop workflows best suited to subsampling geoscientific data (e.g. lithology, stratigraphy, structure) of different dimensionalities (0D: point observations, 1D: drilling data, 2D: lines and polygons on maps).  In this contribution, we present the findings of subsampling 1:500 000 knowledge-rich geological data and their implications to the Ninghan Syncline 3D geological model. The Ninghan Syncline is a mineralized, lithologically diverse and structurally complex area with regions of sparse and clustered sampling.  This dataset gives the opportunity to test the effectivity and limits of these subsampling workflows. The models are built through systematically subsampling the geological maps, ensuring 3D sample representativity and preserving heterogeneity. The geological maps are subsampled based on a combination of the stratigrapic hierarchy and structural complexity. The polygonal features are upscaled by simplifying the geometry and considering crucial vertices critical to preserving the shape and topology. The structural measurements are subsampled depending on the scale of interest, concentration of measurements and consistency of the orientation of these measurements. The subsampled data is used as inputs into a geomodelling engine, in this case, LoopStructural. The resulting 3D geological models are compared and assessed for the effects of subsampling on their geometries and the associated uncertainties from subsampling. The tools developed to subsample geological inputs for multiscale modelling are compiled as a part of the Python package map2loop (https://github.com/Loop3D/map2loop-2).

Measurements from the Gravity Recovery and Climate Experiment (GRACE) provide quantitative measurement of terrestrial water storage anomaly (TWSA) changes with unprecedented accuracy. Nevertheless, TWSA from the past and follow-on GRACE/GRACE-FO mission cover a relatively short period (< 18 years) with a twelve months temporal gap between the two missions and other several missing periods. This short-term time series remain major challenges for long-term studies (e.g., drought assessment). To improve the accuracy of reconstructing and bridging the gaps in the TWSA series, this study provides an examination of boosted regression trees (BRT) over the Yangtze River basin (YRB) of China. BRT incorporates the powers of: 1) regression trees algorithm, models which link the response with their predictors via binary recursive splitting, and 2) boosting algorithm, an adaptive merging process of several simple models in order to yield a robust predictive performance. Thus, the final BRT model could be defined as an additive regression model where the individual terms are simple trees implemented in a forward stagewise style. The network is trained to simulate TWSA based on its relationship with precipitation, surface temperature, soil temperature (level 2/3/4), surface air pressure, soil moisture, (2 meters) temperature, instantaneous moisture flux, and climate indices. Results showed that, the reconstructed TWSA series displayed a significantly high skill performance with a root mean-square error of 19.36 mm, which is akin with the accuracy of GRACE-TWSA datasets, and Nash-Sutcliffe efficiency coefficient of 0.86. Thereby, as this study provides a new perspective for reconstructing and filling the gaps of TWSA series, which is desired in drought characterization studies.

The number of in situ measurements remains lower or non-existent in many rivers of West Africa. However, remote sensing has been used to overcome this situation, and in this study, two methods to estimate the cross-section of rivers by remote sensing are proposed. The Lateral-method (LM) consists of following the river water lateral evolution on either side of a central line using daily satellite imagery, while the river depth is estimated using satellite altimetry. The Mixed-method (MM) consists of collecting the river bed data by spatial altimetry during its period of recession, and the heights of the banks are extracted from a digital elevation model (DEM). The resulting cross-section's area is estimated, and the equivalent section was calculated to estimate the wetted perimeter and the average river slope was obtained along a 30 km river reach. The extracted hydrological parameters were then used to estimate the discharge at the Middle Reach of Niger River, West Africa. Overall, results show an excellent linear relationship between the estimated cross-sections from LM and MM (R = 0.98), and a good Nash Sutcliffe Efficiency coefficient (>0.80) between the estimated and observed discharge. The obtained results show that the precision is acceptable for estimating the river discharge using LM and MM. However, some errors resulting in high values of RMSE are observed, which could be induced by the uncertainty of the trapezoidal method to estimate large areas with complex geometry during periods of high river flooding.

Heinrich events, originally described in North Atlantic sediments, are expressed globally including tropical Asia, for example, as recorded in Asian speleothems. However, detailed connections of these events between the tropics and high latitude areas are still incomplete. Here, we present a composite stalagmite d¹⁸O record to show Asian summer monsoon variability from 31.5 to 28.4ka and tropical responses to Heinrich Event (HE) – 3. The record is based on two stalagmites collected from Thuong Thien cave located in Son La Province, northwestern Vietnam. The HE-3 signal was clearly identified in our record. At present, annual mean temperature and precipitation are 21C and 1328 mm respectively at our study region.  The cave site is located in the northern edge of present day local summer ITCZ with 80% of annual precipitation centered around May-September brought by the Indian Summer Monsoon. The ultrahigh temporal-resolution of d¹⁸O and d¹³C with excellent chronology allows us to investigate detail precipitation and vegetation changes in tropical monsoonal Asia during the HE-3. We also use numerical simulation to support the observed relationship between North Atlantic cooling and precipitation changes at the study site. Besides, our δ¹³C record variation shows consistent vegetation changes with d¹⁸O derived precipitation records.

In order to understand variability in sea surface temperature (SST) and sea surface salinity (SSS) in the South China Sea (SCS), we sampled two Porites coral cores from Dongsha Atoll and analyzed their trace element and stable isotopes (e.g., Sr/Ca, δ18O). We established monthly-resolution time series of 18 years and 25 years on corals cored from 2.5-meter and 8.6-meter depths, respectively. Both Sr/Ca and δ¹⁸O show strong seasonal variations over the past 20 years. Particularly, the Sr/Ca time series from the two cores are in good agreement with the instrumental SST observations from both in-situ and satellites. The linear regression between coral Sr/Ca and instrumental SST yields a slope of -0.07 mmol/℃ and -0.05 mmol/℃ (r=0.87), for two cores, respectively, within the published range of Porites corals. The regression remains nearly constant across different time intervals, and thus the slight difference in regression between the two cores could be attributed to their different oceanic settings and/or growth rate. Because coral δ¹⁸O records a combination of both sea surface temperature and salinity, the SSS addressed by seawater δ¹⁸O can be decoupled by removing the temperature component inferred from paired Sr/Ca ratio. Nevertheless, no significant correlations are observed between the SSS calculated using different methods and instrumental SSS regardless of coral cores. Our study highlights the sensitivity and utility of coral geochemistry from Dongsha Atoll in the South China Sea for reliable SST and SSS reconstructions.

The moisture transport pathway effect was recently proposed to explain the muted glacial-interglacial variability in Chinese speleothem δ¹⁸O records. However, the published cave δ¹⁸O records along the Indian Summer Monsoon trajectory are intermittent thus required further validation for this hypothesis. Here we reported supplemental speleothem U-Th dating and δ¹⁸O results from Lin Noe Twin Cave (LNT) located at ISM midstream in central Myanmar over the past 20,000 years. Our results show that LNT δ¹⁸O shifted from ~-5.5‰ during the LGM, -3‰ during H1, -6‰ in the Younger Dryas, and to ~-7.5‰ in present-day. The shift (~2‰) between the LGM and present-day is smaller than the ~5‰ recorded at coastal sites, but larger than the ~1‰ at sites further downwind, e.g., in Yunnan caves. This confirms the existence of spatial sensitivity of speleothem δ¹⁸O to glacial boundary forcings, thus reinforces the transport pathway effect. To further explore this hypothesis, we analyzed LNT samples for triple oxygen isotopes. Based on preliminary data of speleothem ∆¹⁷O (carbonate ¹⁷O anomaly) across well-defined climate events, we calculated the ¹⁷O-excess of speleothem parent waters with the consideration of temperature effect during calcite precipitation. We found that ¹⁷O-excess of monsoon precipitation during the LGM is ~30±6 per meg, distinctly larger than ~15±5 per meg in modern-times. If these values hold, the 15 per meg deceases in ¹⁷O-excess from the LGM to Holocene may indicate a ~15% drop in relative humidity (RH) in Indo-China Peninsula during the LGM than present-day, given the assumption that RH in the tropical Indian Ocean remains constant through time. Such a significant drop in RH can be attributed to a stronger continental re-evaporation but possibly suppressed plant transpiration during the glacial time. Therefore, both our speleothem δ¹⁸O and ∆¹⁷O results support the transport pathway effect. 

Most precipitation records within the South American Summer Monsoon (SASM) domain bear a trend following the Southern Hemisphere Summer Insolation (SHSI). However, a growing coverage of precipitation records have revealed that rainfall histories also show a high degree of regional variability across the continent. Here we report a new speleothem δ¹⁸O record from central Brazil that underlies the South Atlantic Convergence Zone (SACZ). Our record reveals an invariant rainfall trend over the mid to late-Holocene, suggesting that the SACZ maintained its latitudinal position and intensity despite the rising SHSI and increasing monsoon strength in the core SASM zone. The record also shows no significant response to millennial-scale temperature anomalies in the North Atlantic. Instead, the long-term SACZ behaviour is likely dominantly influenced by subtropical South Atlantic sea-surface temperatures, which can affect the position and intensity of moisture convergence over the continent. Overall, our central Brazil speleothem record further suggests that orbital-scale rainfall trends in monsoon regions are not necessarily insolation-driven. Alternate drivers of hydroclimate change need to be explored to account for regional differences in rainfall patterns within the SASM system.

In recent years, the geophysical method has been widely used in groundwater investigation and exploration, especially non-invasive techniques such as Ground Penetrating Radar (GPR) and Electrical Resistivity Imaging (ERI). Here, we present how these techniques have emerged as valuable tools for investigating the groundwater parameters. We deployed multi-channel ERI with 100 electrodes (Wenner-Schlumberger configuration) and GPR (100 MHz) in the Guanyin beach (GB) and the Wu river (WR). The ERI results showed that the resistivity value in the GB (1-48 Ωm) is lower than in the WR (40-1400 Ωm) due to the presence of saltwater, whereas GPR results showed a high signal attenuation in the GB compare with the WR. In addition, we conducted two-dimensional forward modeling to understand and enhance the GPR results. Furthermore, we used several approaches to obtain the curve of soil water content (θ) versus depth, the highest θ in the GB is 0.35 and in the WR is 0.26. The estimated specific yields from the ERI measurements are 0.3 in the GB and 0.07 in the WR, where the results are consistent with those estimated from the GPR measurements. Both techniques suggest that the groundwater table of GB and WR were at the depth of 0.4 m and 16.56 m, respectively.

   In 2019, we conducted time-lapse Electrical Resistivity Imaging (ERI) surveys with 5 times measurements to evaluate the perched aquifer in one site at the Pingtung area in different seasons. From the 2-D inversion results, we distinguish the model into three parts of low-high-low resistivity distribution, which corresponds to surface soil, unsaturated gravel layer and saturated gravel layer respectively. However, some of the ERI inversion in wet season showed a low resistivity layer around 50-100 Ohm-m in the shallow surface, that might be due to the presence of a perched aquifer. Thus, this study focuses on identify the boundary of the perched aquifer and the effect of perched aquifer while estimating the ground water depth. In order to estimate the boundary of perched aquifer, first, we build two kind of model based on in-situ resistivity distribution and borehole data. And one of the models present perched aquifer present. Afterwards, Archie’s law is applied to calculate the resistivity change with depth and the saturation curve of Van Genuchten model is used to calculate the groundwater depth. The inversion results of both synthetic models show similar distribution to the in-situ results and that the model with a perched aquifer showed a low resistivity layer in the shallow part. The effect also reveal the groundwater difference of the model, where the groundwater depth of the one that without perched aquifer is 6.8 meters deep while the one with perched aquifer are 3 and 17.7 meters, shallower one is the depth of perched aquifer. The resultant of groundwater level is obviously overestimated and the error calibration is about 11 meters. Hence, we can adjust the groundwater level by adding error calibration to the result. Furthermore, the error due to the perched aquifer is related to the phenomena of current focusing in the aquifer.

Anthropogenic activities and climate change have badly affected the balance between natural recharge and discharge of groundwater across the globe, especially in the low-to-middle income countries where groundwater is the primary source of irrigation, and domestic consumption. Implementation and operation of artificial recharge of groundwater or Managed Aquifer Recharge (MAR) for storage and reutilization of surface runoff is indispensable to achieve sustainable groundwater practices. However, delineation of a prospective MAR site and its success depends on the detailed knowledge of the subsurface geology and associated aquifer system. This can be effectively accomplished by electrical resistivity studies. In this contribution, time-lapse electrical resistivity tomography (TL-ERT) in conjunction with borewell investigation was performed at two locations inside the IIT Kanpur campus to examine: (i) reduction in resistivity values before and after monsoon and (ii) variation in sand strata with depth. The percent change in the TL-ERT resistivity data with respect to pre-monsoon resistivity values was calculated for two locations and were correlated with the bore well lithologs. In the first location, a ~15 m thick channel comprising of sand strata at a depth of ~10-15 m was observed whose resistivity values were reduced by 45%. Similarly, in the second location, two channels of sand strata were observed whose resistivity values were reduced by 20-80% with respect to pre-monsoon resistivity values. Also, it was evident that one of the channels was extending at least up to 50 m, However, due the limited depth of investigation (i.e., 50 m) it was difficult to ascertain if the channel extended beyond 50 m. Thus, we could infer that changes in the resistivity values observed at two sites is primarily due to change in the saturation level of the permeable strata after the monsoon.

About 93% of the excess energy stored in the climate system due to anthropogenic greenhouse gas emissions has been absorbed by the oceans, leading to thermal expansion and sea level rise. For more than 600 million people living in low-elevation coastal areas future sea level rise is one of the main damaging aspects of climate change. Crucial decisions about adaptation to sea level rise for populous coastal megacities and communities in low lying small islands, where a considerable fraction of global economic activity and critical infrastructure exists, will be made based on future sea level projections. Confidence in these projections depends on the ability of climate models to simulate the various components of future sea level rise. In this study we analyse the ability of climate models to reproduce the observed changes in sea level. Analysis of thermosteric sea level component, defined by the heat uptake by the ocean, provides evidence that CMIP6 model simulations are underestimating observed sea level rise over the 20^th century.  The rate of CMIP6 ensemble mean of global mean thermosteric sea level rise is 0.2 ± 0.1 mm yr^-1, which is less than half of the observed rate (0.5 ± 0.02 mm yr^-1). To place these results in a wider context we use paleo data and introduce a global mean thermosteric sea level sensitivity as a function of the changes in the global mean thermosteric sea level (as rate in cm per century) to changes in global surface temperature (°C per century). We further discuss the difference in sea level sensitivity to the changes in global surface temperature over the historical and future periods, suggesting that current models are conservative in simulating heat uptake by the ocean and therefore sea level will likely rise faster  than previously estimated.

Relative sea-level (RSL) records from far-field regions distal from ice sheets remain poorly understood, particularly in the early Holocene. Here, we extended the Holocene RSL data from Singapore by producing early Holocene sea-level index points (SLIPs) and limiting dates from a new ~40 m sediment core. We merged new and published RSL data to construct a standardized Singapore RSL database consisting of 88 SLIPs and limiting data. In the early Holocene, RSL rose rapidly from -21.0 m to -0.7 m from ~9,500 to 7,000 cal. yrs. BP. Thereafter, the rate of RSL rise decelerated, reaching a mid-Holocene highstand of 4.0 ± 4.5 m at 5,100 cal. yrs. BP, before falling to its present level. There is no evidence of any inflections in RSL when the full uncertainty of SLIPs is considered. When combined with other standardized data from the Malay-Thai Peninsula, our results also show substantial misfits between regional RSL reconstructions and glacial isostatic adjustment (GIA) model predictions in the rate of early Holocene RSL rise, the timing of the mid-Holocene highstand and the nature of late Holocene RSL fall towards the present. It is presently unknown whether these misfits are caused by regional processes, such as subsidence of the continental shelf, or inaccurate parameters used in the GIA model.

Australia is located in the far-field of past and present polar ice sheets and therefore Holocene relative sea-level (RSL) data from this region is driven by meltwater influx to ocean basins following the last deglaciation and processes such as equatorial syphoning and hydro-isostatic loading of the continental crust. Together, these processes resulted in a mid-Holocene RSL highstand across its coastlines. However, the timing and magnitude of the highstand and the nature of mid- to late-Holocene RSL fall to present level is disputed, which may in part be due to comparisons of RSL records generated using different methodologies. Further, the RSL record from Australia is incomplete, with notable spatial gaps in western Australia. To address these knowledge gaps, we aim to establish reliable records of RSL change in Western Australia through the collection of new field data and the standardization of existing RSL reconstructions. Here, we present a preliminary, quality-controlled compilation from the last 12 ka to present that spans from 12.2°S (Bonaparte Gulf) to 33.9°S (Esperance) and 113.8°E (Houtman Abrolhos Islands to 127.8°E (Cape St. Lambert). The database consists of 93 index points and 47 limiting points from sedimentary (upper intertidal sediment and estuarine mud), geomorphic (coastal dunes, coral reefs), and fixed biological indicators (oysters and encrusting bivalves). We divide the database into two sub-regions to investigate the influence of GIA and other local- to regional-scale processes on RSL. Finally, we apply an error-in-variables integrated Gaussian process model to the data to evaluate the timing and magnitude of the mid-Holocene highstand across the two regions and examine temporal patterns of change from the highstand to present.

Holocene relative sea-level (RSL) data from the China coast, a location far-field from past ice sheets, has been used to infer sea-level equivalent changes. However, there are notable misfits between geological RSL data and current glacial-isostatic adjustment (GIA) model predictions in the timing and amplitude of the RSL highstand and temporal variation over the Holocene. These misfits bring into questions: 1) the interpretation and accuracy of proxies; 2) whether models using 3D viscosity structure improve the fit to data compared to 1D models and 3) whether other local- to regional-scale processes may have significantly influenced RSL. Here we present an updated sea-level database using standardised protocol and improved quality control, which consists of 105 index points and 134 limiting points extending geographically from 119.19°N (Fuzhou) to 111.82°N (Xisha Islands) and 21.82 °E (Qinzhou) to 25.22 °E (Putian). These data are generated from sedimentary (salt marsh, mangrove, tidal flat, lagoon), geomorphic (chenier ridge, beach rock), fixed biological (oyster), and coral indicators. We divide the database into 4 subregions due to the relative influence of GIA, deltaic compaction, and tectonics (Fujian and Taiwan Strait, Han River Delta, East Guangdong, and West Guangdong and Hainan). We employ a spatio-temporal empirical hierarchical model to estimate the rate and timing of RSL change and compare the results to 1D and 3D GIA models. We observe spatial variability over a shorter spatial scale than the long-wavelength GIA signal, which suggests RSL is influenced by local- to regional-scale processes. Finally, we evaluate the effects of processes such as sediment compaction, tidal range change, and tectonic vertical motion on RSL change on the south China coast. 

The mid-Holocene highstand, when relative sea level (RSL) reached higher than present levels, is a common character that revealed in Holocene RSL records from regions distal from ice sheets (far-field). It’s magnitude and timing vary spatially due to hydro-isostatic processes including ocean syphoning and continental levering. While the timing, magnitude and source of ice-equivalent sea level in the middle to late Holocene are not well determined. Here, we compare Glacial Isostatic Adjustment (GIA) model predictions to a standardized database of sea-level index points (SLIPs) from Southeast Asia where we have near-complete Holocene records. The database has more than 130 SLIPs that span the time period from ~9.5 ka BP to present. We investigate the sensitivity of mid-Holocene RSL predictions to GIA parameters, including the lateral lithospheric thickness variation, mantle viscosity (both 1D and 3D), and deglaciation history from different ice sheets (e.g., Laurentide, Fennoscandia, Antarctica). We compute gravitationally self-consistent RSL histories for the GIA model with time dependent coastlines and rotational feedback using the Coupled Laplace-Finite Element Method. The preliminary results show that the timing of the highstand is mainly controlled by the deglaciation history (ice-equivalent sea level), while the magnitude is dominated by Earth parameters (e.g., lithospheric thickness, mantle viscosity). We further investigate whether there is meltwater input during middle to late Holocene and whether the RSL records from Southeast Asia can reveal the meltwater source, like Antarctica.