Geostatistical Algorithms Research Articles

Spatio-temporal geostatistical modelling of sulphate concentration in the area of the Reocín Mine (Spain) as an indicator of water quality.

Water stored in open-pit lakes can be a water resource when the mine is closed. This study aimed to develop a reliable model to evaluate the water quality, based on the sulphate concentration, in the Reocín Mine area (Spain) by using geostatistical algorithms. To this end, water samples were taken from the beginning of the flooding period in November 2004 until August 2020. The model showed that the sulphate concentration was highest between February 2009 and February 2012 and decreased as the flooding process progressed. The area with the highest concentration (2000 mg L-1) was the central part of the study area, where the mine is located, while in the northeast and southwest, the values from the beginning of the flooding period were lower, below 500 mg L-1. In the last obtained model, the values decreased considerably to 1300 mg L-1 in the central area and below 250 mg L-1 in the northeast and southwest areas. The modelling conducted to assess the water quality in the area of influence of the mine determined that the flooding process has little influence on the water in the rivers and streams in the area, since the sulphate concentration measured in the adjacent rivers and streams was less than 250 mg L-1, indicating that anomalous concentrations were only found in the open-pit area. It was shown that geostatistical algorithms are useful tools that can be used to model the intensity and extension of water pollutants in space over time.

Application of a geostatistical algorithm for correcting well logging data when modelling complex hydrocarbon deposits at the stage of additional field exploration

Background. Additional exploration of oil and gas reserves requires the application of information technologies for modelling all technological processes and interpreting the results of geophysical well logging. Reliable models of the object under exploration and the use of best practices in the field of information systems reduce investment risks associated with planning exploration works, at the same time as optimizing the planning and design of the entire process of field development.Aim. Optimization of the selection process of multivolume, multiparameter logging data when modelling a complex oil reservoir at the stage of additional field exploration, grouping well data by quality, completeness and uniformity of their distribution over the study area, as well as identifying inaccuracies (errors) in the field data (registration of signals in probes, errors in measurements of directional survey, borehole picks, etc.).Materials and methods. Examples of well logging interpretation for a complex reservoir are given, along with its distribution options for a constructed lithology cube (2D, 3D) based on the data obtained during drilling from one pilot borehole – a horizontal well, a sidetrack and a directional well. It was established that the data obtained when distributing the reservoir and modelling the lithology cube of a complex development target can be significantly different for wells located close to each other. The results obtained when constructing 2D and 3D models by different specialists revealed discrepancies in the interpretation and cross-well correlation of well sections.Results. Local zones with a sharp change in the structure were established, in particular, the concentration of extrema around some wells. The presence of such anomalies points to the presence of wells, for which the values of stratigraphic depth marks for wells located close to each other are quite different. A geostatistical algorithm was applied to correct the well data using the “sliding window” method, which became a solution for the most accurate determination of the depths of the object under exploration.Conclusions. The possibility of using a geostatistical algorithm for correcting well data based on multiparameter geophysical data was assessed; the current state of modelling the processes of exploration and development of oil fields with a complex geological structure was analysed.

Geostatistical Algorithm Selection for Mineral Resources Assessment and its Impact on Open-pit Production Planning Considering Metal Grade Boundary Effect

Geostatistical Algorithm Selection for Mineral Resources Assessment and its Impact on Open-pit Production Planning Considering Metal Grade Boundary Effect

An Application of an Embedded Model Estimator to a Synthetic Nonstationary Reservoir Model With Multiple Secondary Variables.

A method (Ember) for nonstationary spatial modeling with multiple secondary variables by combining Geostatistics with Random Forests is applied to a three-dimensional Reservoir Model. It extends the Random Forest method to an interpolation algorithm retaining similar consistency properties to both Geostatistical algorithms and Random Forests. It allows embedding of simpler interpolation algorithms into the process, combining them through the Random Forest training process. The algorithm estimates a conditional distribution at each target location. The family of such distributions is called the model envelope. An algorithm to produce stochastic simulations from the envelope is demonstrated. This algorithm allows the influence of the secondary variables, as well as the variability of the result to vary by location in the simulation.

Streamlining geospatial data processing for isotopic landscape modeling

AbstractStable isotopic landscape modeling has become a promising approach for answering research questions in multiple disciplines. However, its application has been hindered by the difficulty for individual researchers to collect, compile, and integrate environmental and isotopic data over large spatial and temporal scales and to develop and interpret geostatistical models. To address these challenges, we developed IsoMAP (http://isomap.org), a science gateway that enables researchers to access and integrate a number of disparate and diverse datasets, develop isoscape models over selected spatiotemporal domains using geostatistical algorithms, predict maps for the stable isotopic ratios, and associate a sample's isotope value with its most likely geographic origin. One main challenge in developing IsoMAP is to efficiently integrate large heterogeneous datasets into the modeling workflow to ensure real‐time query response and timely data update. In this paper, we described how the geospatial data processing workflow was implemented in the initial version of gateway and how it has been improved by leveraging the built‐in vector and raster data processing capabilities and the materialized view object of the PostgreSQL/PostGIS database. Our experience and lessons learned will be applicable to the development of other geospatial data workflows, a common task in the cyberinfrastructure of many science disciplines.

Modelling the Grade and Recovery of Sari Gunay Epithermal Gold with the Application of Geostatistical Algorithms

With the increase in complexity of mineral deposits, understanding the grade variability is of high significance in getting enough information about the deposit, preventing financial losses and identifying optimized plans during and after exploitation. Likewise, mineralogical variability, grade, alteration, … can highly effect metal recovery. Modeling these parameters is seriously substantial in order to have an optimized exploitation of the mine and to minimize the risks and costs. This study used the dataset of 91 drill holes from Sari Gunay epithermal gold deposit located in north-west of Iran to model gold grade and the recovery resulted from leaching tests. Initially, the oxide and sulphide zones were distinguished using Indicator Kriging and Simulation. Then the gold grade and recovery in each zone were modeled with the application of sequential Gaussian simulation and the developed block model. The grade-recovery curve was then drawn to estimate the total tonnage amount of the deposit which was related to the sulphide zone in more than 70%. Then with the help of recovery simulation data, the grade×recovery was considered a preferable parameter in depicting the curve and estimating the reserve.

Model Selection for Mineral Resource Assessment Considering Geological and Grade Uncertainties: Application of Multiple-Point Geostatistics and a Cluster Analysis to an Iron Deposit

The assessment of mineral resources requires the definition of a geological domain, and the quantitative results obtained should consider geological and grade uncertainties. However, when the available data are limited, it can be difficult to estimate the shape of an ore body owing to the spatial uncertainty. This study describes a methodology for defining geological domains using multiple-point geostatistics and selecting models for realizations using a cluster analysis. Then, the selected realizations were assessed for mineral resources using two-point geostatistics. Of the various multiple-point geostatistics algorithms, the single normal equation simulation was used to generate a realistic ore body, and various cluster algorithms, such as k-means and a density-based spatial clustering of applications with noise, were applied to select a model. Based on the outcomes, a mineral resource assessment that uses a sequential Gaussian simulation was performed, and a reasonable model that reflects the original data well was selected based on the grade statistics of the original data and mineral resource assessment per cluster. This study demonstrates that it is possible to define the ore domain and assess mineral resources while accounting for the uncertainty caused by a lack of spatial information.

Geostatistical algorithm for evaluation of primary and secondary roughness

Joint roughness is combination of primary and secondary roughness. Ordinarily primary roughness is a geostatistical part of a joint surface that has a periodic nature but secondary roughness or unevenness is a statistical part of that which have a random nature. Using roughness generating algorithms is a useful method for evaluation of joint roughness. In this paper after determining geostatistical parameters of the joint profile, were presented two roughness generating algorithms using Mount-Carlo method for evaluation of primary (GJRGAP) and secondary (GJRGAS) roughness. These based on geostatistical parameters (range and sill) and statistical parameters (standard deviation of asperities height, SDH, and standard deviation of asperities angle, SDA) for generation two-dimensional joint roughness profiles. In this study different geostatistical regions were defined depending on the range and SDH. As SDH increases, the height of the generated asperities increases and asperities become sharper and at a specific range (a specific curve) relation between SDH and SDA is linear. As the range in GJRGAP becomes larger (the base of the asperities) the shape of asperities becomes flatter. The results illustrate that joint profiles have larger SDA with increase of SDH and decrease of range. Consequencely increase of SDA leads to joint roughness parameters such Z2, Z3 and Rp increases. The results showed that secondary roughness or unevenness has a great influence on roughness values. In general, it can be concluded that the shape and size of asperities are appropriate parameters to approach the field scale from the laboratory scale.

Default versus Configured-Geostatistical Modeling of Suspended Particulate Matter in Potter Cove, West Antarctic Peninsula

The glacier retreat observed during the last decades at Potter Cove (PC) causes an increasing amount of suspended particulate matter (SPM) in the water column, which has a high impact on sessile filter feeder’ species at PC located at the West Antarctic Peninsula. SPM presents a highly-fluctuating dynamic pattern on a daily, monthly, seasonal, and interannual basis. Geostatistical interpolation techniques are widely used by default to generate reliable spatial information and thereby to improve the ecological understanding of environmental variables, which is often fundamental for guiding decision-makers and scientists. In this study, we compared the results of default and configured settings of three geostatistical algorithms (Simple Kriging, Ordinary Kriging, and Empirical Bayesian) and developed a performance index. In order to interpolate SPM data from the summer season 2010/2011 at PC, the best performance was obtained with Empirical Bayesian Kriging (standard mean = −0.001 and root mean square standardized = 0.995). It showed an excellent performance (performance index = 0.004), improving both evaluation parameters when radio and neighborhood were configured. About 69% of the models showed improved standard means when configured compared to the default settings following a here proposed guideline.

Geological process modeling and geostatistics for facies reconstruction of presalt carbonates

The presalt reservoirs of the Barra Velha Formation are complex and heterogeneous. Their deposition was controlled by a mixture of physical and chemical processes, which consequently affected the distributions of facies and variations in porosity and permeability. In this paper we present a presalt facies reconstruction of reservoirs within the Barra Velha Formation of the Búzios Field. This was achieved through the use of an innovative workflow that integrates geological process modeling with a truncated Gaussian simulation geostatistical algorithm with trends. Our results show that the structural highs are dominated by carbonate build-ups/mounds architecture seismic patterns which are dominated by spherulites and its intercalation with shrub sediments culminating in spherulitites and shrubby carbonate lithologies. Reworked sediments and facies are concentrated in regions dominated by debris seismic patterns and lake bottom seismic facies corresponding to laminites also occur within the study area. In chronostratigraphic terms the Barra Velha is muddier at the base transitioning to in-situ carbonates in the middle section, before becoming muddier again at the top. The western portion of the Búzios Field is mostly composed by shrubby carbonates and reworked facies whilst spherulitities are more abundant to the east intercalated with shrubby carbonates. Some uncertainties remain in relation to the parameters used in the proposed methodology. Nonetheless, the proposed workflow was effective for facies reconstruction of the presalt carbonates allowing not only a better understanding of sedimentological processes but also accurately distributing facies throughout the study area.

Accelerating geostatistical modeling using geostatistics-informed machine Learning

Ordinary Kriging (OK) is a popular geostatistical algorithm for spatial interpolation and estimation. The computational complexity of OK changes quadratically and cubically for memory and speed, respectively, given the number of data. Therefore, it is computationally intensive and also challenging to process a large set of data, especially in three-dimensional (3D) cases. This paper develops a geostatistics-informed machine learning (GIML) model to improve the efficiency of OK by reducing the number of points required to be estimated using OK. Specifically, only a very few of the unknown points are estimated by OK to get the weights and estimations, which are used as the training dataset. Moreover, the governing equations of OK are used to guide our proposed machine learning to better reproduce the spatial distributions. Our results show that the proposed GIML can reduce the computational time of OK by at least one order of magnitude. The effectiveness of the GIML is evaluated and compared using a 2D case. Furthermore, we demonstrate its efficiency and robustness by considering a different number of training samples on various 3D simulation grids.

Geostatistical and deterministic methods for rainfall interpolation in the Zayandeh Rud basin, Iran

ABSTRACT A new approach to select a model for fitting to a seasonal daily average rainfall semi-variogram was developed in this study by interpolating rainfall using geostatistical methods. Using two deterministic algorithms (Thiessen polygon, THI and inverse distance weighting, IDW) and two geostatistical algorithms (ordinary kriging, ORK and universal kriging, UNK) based on a rain gauge network, a 20-year daily and annual rainfall time series was generated for the Zayandeh Rud Dam Basin (Isfahan, Iran). Seven models were fitted to the experimental semi-variogram. Evaluation criteria – root mean square error (RMSE) and correlation coefficient (R) – identified that the Gaussian model is the best fit to the experimental semi-variogram. The ORK and UNK algorithms performed better in generating data when network sizes of 27, 15 or 10 gauges were used, whereas THI produced better results at a network size of five gauges. The results show that increasing the number of gauges does not necessarily produce better estimates, and stochastic methods are more sensitive than deterministic algorithms to network size.

Numerical simulation of polydisperse dense particles transport in a random-orientated fracture with spatially variable apertures

Numerical experiments are conducted to quantify the transport behaviour of polydisperse dense particles through a randomly orientated fracture with spatially variable apertures. To be specific, fracture apertures are created by implementing the geostatistical algorithm of random field generation, while, based on the classical Local Cubic Law (LCL), pressure distribution and velocity field across the fracture are determined under given boundary conditions. Subsequently, the existing constant-spatial-step particle-tracking equations are modified by incorporating both particle settling gravity and fracture orientation to simulate particle transport behaviour. Using literature data, the simulation model in this study has been well validated. By simulating a set of well-designed scenarios, effects of fracture orientations, particle size-distribution, and aperture field heterogeneity and anisotropy on particle transport are systematically examined, while sensitivity analysis of the transport behaviour has been evaluated and compared. Compared to particle number breakthrough, mass breakthrough efficiency of particle plume is found to be usually much lower only except for the vertical orientation of a fracture. An increased inclination either parallel or perpendicular to the overall flow direction leads to an earlier particle breakthrough with a shorter tail on the breakthrough curves. A more heterogeneous fracture results in an enhanced longitudinal spreading of particle plume with distinct non-uniform fore- and back-end distribution. For a larger deviated particle size-distribution, particle plume during transport is evolved with more particles at its two-ends. In addition, it is observed from a vertical fracture that aperture anisotropy orientated parallel to the flow direction significantly enhances particle transport and spreading than that perpendicular to the flow direction. Fracture inclinations parallel and perpendicular to the flow direction yield a similar sensitivity on particle transport, while, compared to the heterogeneity of particle size-distribution, the aperture field heterogeneity affects particle transport more significantly. Comparing the aperture field consisting of a larger transverse correlation length with that of a larger longitudinal correlation length, the former is found with a more sensitive dependence of particle transport behaviour on aperture field anisotropy.

Prediction of CO₂ Saturation Spatial Distribution Using Geostatistical Inversion of Time-Lapse Geophysical Data

Carbon dioxide sequestration in deep saline aquifers and depleted reservoirs relies on numerical models for the prediction of the spatial distribution of CO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> saturation during injection and migration. Due to the limited knowledge of the rock and fluid properties before injection, model predictions are often uncertain and must be updated when new measurements are available. The spatial distribution of CO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> saturation and the plume location can be monitored using time-lapse geophysical data, such as seismic and controlled source electromagnetic surveys. We propose a geostatistical inversion approach for the prediction of the time-dependent spatial distribution of CO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> saturation from geophysical data. The methodology is based on the application of a stochastic optimization method, the Ensemble Smoother, for the solution of the inverse problem, using rock physics and geophysical models. The inversion is applied to the difference in the geophysical data acquired before and during injection. The predicted models of CO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> saturation are obtained by updating an ensemble of geostatistically generated prior realizations, based on the misfit between geophysical model predictions and measured data. The novelty of the approach is the integration of geostatistical algorithms and stochastic optimization methods for the joint inversion of geophysical data. The proposed approach allows including hydrological constraints in the prior model and quantifying the prediction uncertainty due to the noise and resolution of the data and approximations in the physical relations. The method is applied to the Johansen formation model, offshore Norway, using synthetic seismic and electromagnetic data.

Performance of clustering for the decision of stationarity; A case study with a nickel laterite deposit

The decision of stationarity is a fundamental prerequisite to geostatistical estimation and uncertainty characterization of natural resources. Great effort is given to delineate relevant spatial-statistical populations for modeling. Recently developed spatial clustering methodologies claim improvements over traditional methodologies for stationary decision making. However, these novel methods are generally unproven with respect to the geostatistical algorithms that generate the models upon which important and costly decisions are based. In this study, a new method to generate the prior parameter uncertainty for the decision of stationarity is established, and the effects of considering different (and uncertain) decisions of stationarity are explored through a K-Fold cross validation study applied to a nickel laterite dataset with complex multivariate relationships.Results show that, in terms of univariate regression metrics, geostatistical models generated with clustering-based modeling domains are competitive or better than those generated with the traditional merged-lithological domains. However, models generated from the clustering-based stationary domains are superior in terms of multivariate feature reproduction when compared to the merged-lithology domains. We also demonstrate a geostatistical modeling workflow that incorporates uncertainty associated with stationary domaining, with only a minimal extension to established techniques for geostatistical modeling with parameter uncertainty.

Geostatistical estimation of Ground Class prior to and during excavation for the Caldecott Tunnel Fourth Bore project

One of the most significant challenges when tunneling in rock is predicting the true tunneling conditions along the alignment. Prior to excavation, engineers rely on sparse borehole information and a knowledge of the regional geology and tectonics to estimate anticipated tunneling conditions and associated spatial uncertainty. As more data become available during excavation, in some projects, geologists and engineers construct maps of the tunnel face that record the tunneling conditions at closely-spaced intervals along the tunnel, which can be used to update predictions of tunneling conditions just ahead of the face. This work used a variogram-based geostatistical algorithm (indicator kriging) with both borehole data and face maps from the completed Caldecott Tunnel Fourth Bore in California, USA to predict anticipated tunneling conditions and associated uncertainty both prior to and during excavation to integrate information available at different times during the life of the project. First, borehole data available prior to excavation were used to estimate anticipated tunneling conditions using two methods which attempt to represent a relationship between measured borehole data and anticipated tunneling conditions; these methods were compared using the face maps collected during excavation as verification. When comparing the two methods in the context of two geologic units from the Caldecott Tunnel Fourth Bore, one of the methods produced reasonable uncertainty predictions for both geologic units, while the second produced excellent uncertainty predictions in one geologic unit and poor uncertainty predictions in the other geologic unit due to the geometrical configuration of the boreholes. The former method was used to assess the impact of face maps on predicted tunneling conditions in both geologic units, and it was found that a large reduction in uncertainty can be achieved 1–5 m ahead of the face when incorporating face maps into updated geostatistical calculations.

Application of unconditional simulation methods for quantifying the uncertainties in mud window design of gas reservoirs based on 3-dimensional mechanical earth modeling

Mud window is an inseparable part of well design. It is so crucial for having a successful drilling operation, especially in gas fields. Reliable and accurate design of mud weight is the key to having a safe wellbore. Many wellbore stability problems occur because of the inaccurate mud weight in both gas and oil reservoirs. Inaccurate and even sometimes wrong mud weight is caused by lack of adequate knowledge about the field geomechanics. Quantifying the geomechanical uncertainties allows taking the knowledge incompleteness into account. It can reduce the risks of unreliable mud weight design. The geomechanical parameters are only known at some sparse points of measurements. Therefore, it is necessary to estimate the parameters in the whole reservoir as if there is a decision to drill an infill well or conduct a well work-over or side-track an old well. Geostatistical algorithms are utilized to create 3D mechanical earth modeling (MEM) of the system. Several MEM realizations are created to span the range of geomechanical uncertainty adequately. The mud window is designed under uncertainty based on Mohr-Coulomb failure criteria. In this article, at first, 1D MEMs at well locations are built by the available data, and then ordinary Kriging is used to propagate 1D MEMs onto 3D MEM. Smooth plus rough – nonlinear problem (LMAP) and Rough plus smooth – nonlinear problem (RML approximate) are the methodologies applied to quantify the geomechanical uncertainties. In the end, the optimum mud window and the corresponding uncertainty ranges are identified based on the posterior probability distribution of geomechanical parameters in one of the gas reservoirs in the south of Iran. The resulting mud window is narrower compared to the mud window designed with prior information. Consequently, mud window design would be more reliable with less possibility of kick or any wellbore instability.

Stochastic Modeling of Chemical Compounds in a Limestone Deposit by Unlocking the Complexity in Bivariate Relationships

Modeling multivariate variables with complexity in a cross-correlation structure is always applicable to mineral resource evaluation and exploration in multi-element deposits. However, the geostatistical algorithm for such modeling is usually challenging. In this respect, projection pursuit multivariate transform (PPMT), which can successfully handle the complexity of interest in bivariate relationships, may be particularly useful. This work presents an algorithm for combining projection pursuit multivariate transform (PPMT) with a conventional (co)-simulation technique where spatial dependency among variables can be defined by a linear model of co-regionalization (LMC). This algorithm is examined by one real case study in a limestone deposit in the south of Kazakhstan, in which four chemical compounds (CaO, Al2O3, Fe2O3, and SiO2) with complexity in bivariate relationships are analyzed and 100 realizations are produced for each variable. To show the effectiveness of the proposed algorithm, the outputs (realizations) are statistically examined and the results show that this methodology is legitimate for reproduction of original mean, variance, and complex cross-correlation among the variables and can be employed for further processes. Then, the applicability of the concept is demonstrated on a workflow to classify this limestone deposit as measured, indicated, or inferred based on Joint Ore Reserves Committee (JORC) code. The categorization is carried out based on two zone definitions, geological, and mining units.

PIN133 A GEOSTATISTICAL ALGORITHM TO BETTER IDENTIFY CONTEXTUAL AND CLINICAL FACTORS ASSOCIATED TO HPV VACCINE COVERAGE IN FRANCE

Large disparities of the Human Papillomavirus (HPV) vaccine coverage rates (VCR) are observed over the French territory. The study aims at identifying clinical and contextual factors explaining such variations using geostatistical algorithms which will be adapted to HPV vaccination in France. Understanding which community-level or individual-level factors are associated with HPV vaccination is a key aspect for designing targeted programs toward VCR improvement. At the community level, the HPV VCR were crossed over the French territory with a large number of data gathered in sociological, economic, clinical, political and behavioral categories for the year 2016. It includes agglomerated and geolocalized web and social network data. The analysis was driven in a spatially varying way in order to determine geographical areas which are homogeneous in regard to the observed correlations. In each geographical area, a factor analysis was performed to reduce redundant information. The study shows that HPV VCR spatial variations in France cannot be fully explained by a unique model. Rural areas were more influenced by political and sociological factors (especially socio-professional categories and education level), while urban areas were preferentially associated with economic and migration related factors. Two secondary geographical areas were determined: Ile-de-France and 10 North Eastern departments. Discriminating them was contributing to a better characterization of the VCR variations. The Ile-de-France was showing even more economic factors impact than the urban area as a whole, and the North Eastern departments located in the urban area were showing very high level of VCR despite a poor economic situation. Our geostatistical modeling approach leads to a better comprehension of the HPV VCR disparities. That is the first time that such innovative method is applied to identify factors influencing HPV VCR. Using this model, future localized vaccination programs should be targeted in specific sub-populations of interest.

Uncertainty, sensitivity analysis and optimization of a reservoir geological model

Geological models are mainly obtained via interpolation of simulations using geostatistical algorithms, and the simulation results are highly uncertain. In this paper, several factors that have great influences on model results are analyzed. Through an analysis of the sensitive uncertain factors, it is found that the uncertainty of the lithofacies model has the greatest influence on the geological model, and the variogram has the second greatest influence. Therefore, ensuring the rationality of lithologic model can effectively reduce the uncertainty of the model. It is found that using seismic impedance as constraints can effectively improve the accuracy of the lithofacies model, and one can use the seismic impedance to obtain a reasonable variogram (effectively overcoming the shortage of well data), thus reducing the uncertainty of the model. Although the stochastic simulation technology used in geological modeling can yield many simulation results, only a limited number of models can be applied in practical work. In this paper, a probabilistic analysis method is proposed to optimize the model, which can provide a reference for business decision-making and risk assessment of oil and gas companies. Reserve uncertainty analysis in the exploration stage can determine the most probable reserve of the reservoir through the probabilistic characteristics of the reserve so as to determine the next exploration target. In the development stage, probabilistic analyses of the porosity, remaining oil saturation and other parameters related to oil field production can maximize the development of remaining oil. The idea of model optimization based on sensitivity analysis of model uncertainties proposed in this paper can effectively reduce the uncertainty of the model, improve the accuracy of the geological model and meet the needs of different research stages and levels.