Earth Sciences Research

Highly specialised requests are first checked for feasibility from this information. Please do not send confidential data through the public form; describe it first and agree secure transfer afterwards.

1. 01Enquiry and requirement assessment
2. 02Technical scoping and proposal
3. 03Execution
4. 04Quality review and delivery, with an agreed round of revisions

Read the full project workflow.

In depth: how the work is done

1. Geostatistics and spatial interpolation

   Variograms describe how similarity decays with distance, and kriging uses them to interpolate with an estimate of uncertainty. Reliable variograms need enough sample pairs at each lag, and anisotropy and trends must be handled. We compare kriging with simpler methods, validate by cross-validation or hold-out data and show prediction variance maps beside the predicted surfaces, since interpolated maps otherwise look more certain than they are.

2. Multivariate analysis of geochemical and geophysical data

   Principal component analysis, clustering, discriminant analysis and related methods summarise many variables. Compositional data, such as element concentrations summing to a constant, require appropriate transformations, for example log-ratios, to avoid spurious correlations. We check assumptions, standardise suitably, test stability of clusters and interpret components in terms of geological processes only where supported by independent evidence.

3. Numerical modelling concepts

   Models of groundwater flow, heat transfer, deformation or reactive transport discretise space and time and need boundary and initial conditions that are often poorly known. We help formulate conceptual models, choose numerical approaches, run convergence tests and compare model output with observations. The conceptual model, the set of ideas about how the system works, is treated as a source of uncertainty in its own right.

4. Uncertainty quantification and non-uniqueness

   Inverse problems in earth science typically have many solutions that fit the data equally well. Bayesian and ensemble methods sample plausible solutions, while sensitivity analysis shows which parameters matter. Results are presented as ranges or probability distributions, together with the assumptions behind priors and error models. Deterministic best-fit models are described as one realisation among many.

5. Data integration across scales and methods

   Borehole logs, surface maps, geophysical surveys and remote-sensing products differ in resolution, support volume and error. Combining them needs consistent coordinate systems, careful upscaling and weighting by reliability. We document how each dataset was used and check that integration does not impose agreement that the data do not contain.

6. Reporting earth science results

   Reports should include sampling design, coordinates reference system, units, detection limits, treatment of below-detection values and quality control results. Figures show data locations so readers can judge coverage. Interpretations are separated from observations, and alternative explanations are discussed rather than omitted.

7. Choosing methods that match the sampling design

   Before choosing an interpolation or multivariate method, we ask how the samples were collected: regular grid, targeted at anomalies, along transects or opportunistic. Preferential sampling near high values biases variograms and means, and declustering or model-based approaches may be needed. Sample support, such as core length or composite size, affects variability and should be consistent. Documenting the design lets reviewers judge whether the statistical method is appropriate and whether conclusions can be generalised beyond the sampled locations.

8. Planning additional sampling

   Where uncertainty maps show poor constraint, further sampling is more valuable in some places than others. Locations with high prediction variance, near decision thresholds or in regions that control the conclusion are priorities. Simple designs such as stratified or spacing-filling samples can be compared by their expected reduction in kriging variance. This turns a limitation into a concrete recommendation for the next field or laboratory campaign.

Common pitfalls this service helps avoid

• Ignoring spatial autocorrelation. Treating clustered samples as independent overstates the precision of estimates.
• Applying standard statistics to compositional data. This produces spurious correlations among components.
• Presenting a single inverse model. Non-uniqueness means many models fit the data.
• Smoothing away real structure. Interpolation settings can hide or invent features.

What good reporting looks like

An earth-science report documents the study area, the sampling or survey design, the data and quality control, the spatial and multivariate methods and parameters, the models and their conditions, the uncertainty analysis, the interpretation with alternatives, and limitations. It supports research and does not replace professional geological assessments such as resource estimation or hazard certification, which follow recognised codes and require qualified persons.

An illustrative scenario

Hypothetical example written for explanation. It is not a client case, a result or a testimonial.

Imagine a researcher mapping soil contamination from sparse samples. The first kriged map shows a hotspot in a region with only two samples. A cross-validation reveals poor performance there, and the prediction-variance map confirms high uncertainty. The researcher adds a proposed sampling plan for the region and reports the original map with the variance map, noting that below-detection values were handled by a stated substitution method with sensitivity tested.

Questions to settle before you begin

• Is my sampling design adequate to estimate spatial structure?
• How do I treat below-detection values and compositional constraints?
• What validation will test my interpolation?
• Which conceptual models are plausible?
• How should I show uncertainty on my maps?

Key terms

Variogram

A function describing how spatial dissimilarity grows with distance.

Kriging

A geostatistical interpolation method that gives estimates and variances.

Log-ratio transform

A transformation for compositional data.

Non-uniqueness

The existence of multiple models that fit the same data.

Support

The volume or area over which a measurement is made.

Working together on Earth Sciences & Geosciences projects

Earth sciences engagements start with the dataset or field question: location, scale, sampling approach, measurement methods and the uncertainties in each. We ask for data tables, maps or logs in standard formats, the conceptual model you hold and any prior interpretations. Analyses make explicit how spatial and temporal sampling limits what can be concluded. Reports describe uncertainty and non-uniqueness, as in geophysical inversion, and separate observation from interpretation. Site-specific engineering or hazard decisions require qualified professionals and are outside the scope.

Few projects need a single service in isolation. These are the services most often combined with this one.

• Geology & Geophysics

  Analysis and interpretation support for geological and geophysical datasets, including seismic, gravity, magnetic and electrical methods, with attention to processing steps and non-uniqueness.

• GIS & Remote Sensing

  Spatial analysis and satellite or aerial image analysis for engineering and environmental research, including data preparation, classification and change detection.

• Hydrology & Hydrogeology

  Modelling and data analysis for surface-water and groundwater systems, including water balance, rainfall-runoff and groundwater-flow concepts.

• Seismology & Natural Hazard Modelling

  Analysis of seismic data and modelling of natural hazards such as earthquakes, landslides and floods, with probabilistic thinking and clear communication of uncertainty.