Scholarlink

<p>The spatial characterisation of buried peat is of exceptional importance in various research disciplines yet remains understudied. The increasing availability of legacy boreholes forms an opportunity but also poses a challenge in geomodelling, especially in areas with buried peat which hold heterogeneous lithology with varying proportions. Improved three-dimensional modelling of buried peat, including quantification of uncertainty from legacy borehole descriptions, can support the development of decision support systems in land redevelopment. Within this context, a novel method for three-dimensional characterisation of buried peat based on legacy boreholes is compared to current practices. A 1,100-ha embanked floodplain of the river Scheldt in Belgium is presented with 190 legacy boreholes, from which descriptions have been encoded into lithology proportions. Peat lithology proportions of our input data showed an asymmetrical, bimodal distribution with many zeroes. We applied inverse distance weighting (IDW), widely used in practice, ordinary kriging (OK), popular in geospatial analysis, and indicator kriging (IK), with a novel propagation of uncertainty from legacy borehole descriptions through soft indicator coding. More specifically, an inclination-dependent power in IDW and variogram model parameters in kriging were optimised through cross-validation with a calibration subset of 164 boreholes. To ensure robust, independent and quantitative validation, we conditionally subsampled 26 validation boreholes, based on borehole attributes including sampling date, for assessment of the three-dimensional models and derived peat volume estimations. With a median absolute prediction error of 6.4%, 5.5% and 0.8%, and a correlation coefficient of 0.52, 0.53 and 0.55 for IDW, OK and IK, respectively, probabilistic methods prove to be more accurate in predicting peat proportions. The non-parametric IK approach with local conditional probability modelling captured data uncertainty and demonstrated versatility; it facilitates both buried peat reconstruction based on further geological expertise and uncertainty propagation in derived models, such as the peat volume estimations presented herein.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Three-Dimensional Spatial Reconstruction of Buried Peat Based on Legacy Borehole Data

  • 期刊论文

  • 发表日期：

    2025年7月18日

查看全文
  • Pablo De Weerdt,
  • Jeroen Verhegge,
  • Philippe De Smedt,
  • Tom Van Haren,
  • Roel De Koninck,
  • Ellen Van De Vijver

摘要

The spatial characterisation of buried peat is of exceptional importance in various research disciplines yet remains understudied. The increasing availability of legacy boreholes forms an opportunity but also poses a challenge in geomodelling, especially in areas with buried peat which hold heterogeneous lithology with varying proportions. Improved three-dimensional modelling of buried peat, including quantification of uncertainty from legacy borehole descriptions, can support the development of decision support systems in land redevelopment. Within this context, a novel method for three-dimensional characterisation of buried peat based on legacy boreholes is compared to current practices. A 1,100-ha embanked floodplain of the river Scheldt in Belgium is presented with 190 legacy boreholes, from which descriptions have been encoded into lithology proportions. Peat lithology proportions of our input data showed an asymmetrical, bimodal distribution with many zeroes. We applied inverse distance weighting (IDW), widely used in practice, ordinary kriging (OK), popular in geospatial analysis, and indicator kriging (IK), with a novel propagation of uncertainty from legacy borehole descriptions through soft indicator coding. More specifically, an inclination-dependent power in IDW and variogram model parameters in kriging were optimised through cross-validation with a calibration subset of 164 boreholes. To ensure robust, independent and quantitative validation, we conditionally subsampled 26 validation boreholes, based on borehole attributes including sampling date, for assessment of the three-dimensional models and derived peat volume estimations. With a median absolute prediction error of 6.4%, 5.5% and 0.8%, and a correlation coefficient of 0.52, 0.53 and 0.55 for IDW, OK and IK, respectively, probabilistic methods prove to be more accurate in predicting peat proportions. The non-parametric IK approach with local conditional probability modelling captured data uncertainty and demonstrated versatility; it facilitates both buried peat reconstruction based on further geological expertise and uncertainty propagation in derived models, such as the peat volume estimations presented herein.