Articles | Volume 21, issue 8
Hydrol. Earth Syst. Sci., 21, 4195–4211, 2017
Hydrol. Earth Syst. Sci., 21, 4195–4211, 2017

Research article 23 Aug 2017

Research article | 23 Aug 2017

Estimating unconsolidated sediment cover thickness by using the horizontal distance to a bedrock outcrop as secondary information

Nils-Otto Kitterød1,2,3 Nils-Otto Kitterød
  • 1Norwegian University of Life Sciences, Environmental Sciences and Natural Resource Management, P.O. Box 5003, 1432 Ås, Norway
  • 2NIBIO – Norwegian Institute of Bioeconomy Research, Water Resources, P.O. Box 115, 1431 Ås, Norway
  • 3Irstea, UR HHLY, Centre de Lyon-Villeurbanne, 5 rue de la Doua BP 32108, 69626 Villeurbanne Cedex, France

Abstract. Unconsolidated sediment cover thickness (D) above bedrock was estimated by using a publicly available well database from Norway, GRANADA. General challenges associated with such databases typically involve clustering and bias. However, if information about the horizontal distance to the nearest bedrock outcrop (L) is included, does the spatial estimation of D improve? This idea was tested by comparing two cross-validation results: ordinary kriging (OK) where L was disregarded; and co-kriging (CK) where cross-covariance between D and L was included. The analysis showed only minor differences between OK and CK with respect to differences between estimation and true values. However, the CK results gave in general less estimation variance compared to the OK results. All observations were declustered and transformed to standard normal probability density functions before estimation and back-transformed for the cross-validation analysis. The semivariogram analysis gave correlation lengths for D and L of approx. 10 and 6 km. These correlations reduce the estimation variance in the cross-validation analysis because more than 50 % of the data material had two or more observations within a radius of 5 km. The small-scale variance of D, however, was about 50 % of the total variance, which gave an accuracy of less than 60 % for most of the cross-validation cases. Despite the noisy character of the observations, the analysis demonstrated that L can be used as secondary information to reduce the estimation variance of D.

Short summary
The GRANADA open-access database (NGU, 2016a) was used to derive point recordings of thickness of sediment above the bedrock D(u). For each D(u) the horizontal distance to nearest outcrop L(u) was derived from geological maps. The purpose was to utilize L(u) as a secondary function for estimation of D(u). Two estimation methods were employed: ordinary kriging (OK) and co-kriging (CK). A cross-validation analysis was performed to evaluate the additional information in the secondary function L(u).