Earth Sciences Research Services

Earth science provides the environmental context for engineering: climate and weather, hydrology, oceans, soils, geomorphology and natural hazards. Engineering research increasingly depends on this context, for example in water resources, renewable energy siting, infrastructure resilience and land-use planning. Datasets are large, spatially and temporally structured, and often come from satellites, networks or models with their own biases. Our support concentrates on data analysis, statistical and numerical modelling, and the clear communication of uncertain, multi-scale results.

Rainfall, temperature, wind and radiation records form the basis for design and risk studies. Raw records contain gaps, station moves and instrument changes. Homogenisation, gap filling and bias correction require care, and each introduces assumptions that should be reported. Extremes are analysed with extreme-value methods whose uncertainty grows with return period.

We help clean and analyse time series, fit distributions, detect trends with appropriate statistical tests and present results with confidence intervals.

Global climate models operate at coarse resolution. Downscaling, statistical or dynamical, translates projections to local scale but cannot remove model structural uncertainty. Using ensembles, multiple emission scenarios and bias correction methods with clear documentation is standard. Projections are scenarios of possible futures, not forecasts, and should be described accordingly.

We support workflows that compare scenarios, quantify spread and communicate the limits of what the projections can say.

Rainfall-runoff models, groundwater models and reservoir operation studies need calibration and validation periods that are distinct. Performance metrics such as Nash-Sutcliffe efficiency can hide poor performance on extremes, so several metrics and visual checks are recommended. Parameter equifinality, where many parameter sets fit equally well, argues for uncertainty ranges.

We help set up calibration, sensitivity analysis and scenario comparison and write results transparently.

Satellite and airborne data supply land cover, vegetation, soil moisture, surface temperature and elevation. Products have resolution, revisit time and error characteristics that influence suitability. Classification and regression models need independent reference data and accuracy assessment, and spatial autocorrelation should be handled in validation.

We support image processing, time-series analysis, model validation and map preparation.

Hazard research combines historical records, mapping, physical modelling and probabilistic analysis. Hazard, exposure and vulnerability are distinct concepts and should not be conflated. Maps should state the return period or probability they depict and the data used.

We help structure analyses, quantify uncertainty and communicate with suitable cartography. Official hazard assessments and early-warning design are outside our scope.

Soil properties, erosion, land degradation and ecosystem processes interact with engineering works. Field sampling must be designed to capture spatial variability, and models such as erosion or carbon models need local calibration.

We help plan sampling, analyse data and compare management scenarios in research studies.

Combining climate, hydrological, ecological and economic models for assessment of adaptation options multiplies uncertainty. Transparent documentation of each component, assumptions and coupling variables is essential. Robust decision approaches evaluate options across many plausible futures rather than optimising for one.

We support the structuring of such studies and the write-up of results with suitable caution.

Good papers specify study area with maps, data sources and periods, methods with parameter values, validation statistics, and uncertainty. Limitations such as short records, sparse stations and model structural error should appear in the discussion.

Our manuscript services help improve structure, figures and reporting based on the authors' own analyses.

Tell us the region or system, the data available, the question and the intended output. We will propose an analysis plan and clarify what field verification would still be needed.

Autocorrelation, seasonality, non-stationarity and missing data complicate analysis. Standard tests assuming independence overstate significance, so methods such as prewhitening, block bootstrap or models with correlated errors are used. Trend estimates depend on the period chosen and on abrupt changes caused by instrumentation. Spatial fields add anisotropy and non-stationary variance.

We help choose suitable methods, test assumptions, run sensitivity checks on period and method and communicate trends with their uncertainty rather than as bare slopes.

Where several indicators or stations are combined, we help document weighting, standardisation and quality control so that composite indices are reproducible.

Sea level, storm surge, wave climate and coastal erosion connect earth science with coastal and marine engineering. Tide-gauge and satellite altimetry records have different coverage and corrections for land motion. Storm surge and wave statistics require long records, and projections of extremes are uncertain.

We support analysis of such records and the writing of coastal hazard and adaptation studies with scenario language.

Earth science data are valuable only if they can be found, understood and reused. Documenting units, coordinate systems, time zones, versions and processing steps, and depositing data in recognised repositories with persistent identifiers, make research reproducible and satisfy funder policies.

We help prepare data descriptions, README files and availability statements.

Design values for wind, snow, rainfall, temperature and seismic action come from standards based on regional data analysis. Research can refine these through improved statistics or climate-change adjustments, but must be clear about the status of the results relative to code values.

We help frame studies so that their relationship to standards is stated correctly and not overstated.

Gridded products, reanalyses and satellite retrievals are not measurements in the ordinary sense: they combine observations with models and algorithms, and their errors vary with region, season and surface type. Comparison with ground observations, where available, quantifies bias and error, and using more than one product shows the sensitivity of conclusions to the data source.

We help structure comparisons, report bias, correlation and error statistics by region, and write transparently about product limitations.

A regional assessment is stronger when it clearly defines the study area, the period, the indicators and the thresholds used to classify conditions. Comparing two or more regions or periods, with the same method, avoids drawing conclusions from a single case. Include maps of data coverage so readers can see where results rest on sparse information.

We help set up such designs and prepare the cartographic and statistical material.

Research on environment and hazards ultimately informs agencies and communities. Plain-language summaries, clear legends, uncertainty statements and links to datasets make results usable. Avoid presenting scenario output as prediction, and name the decisions the work can and cannot inform.

We help with summaries and visual communication, in addition to the technical paper.

Skill scores depend on the baseline: a model that beats climatology by a small margin may still be of limited use. Report skill relative to a simple reference, for different seasons and for extremes separately, and show reliability diagrams where probabilities are issued. Avoid generalising from one catchment or one period.

We help select baselines and metrics and write the evaluation honestly.

Field campaigns need a plan for sampling locations, frequency and quality control before the first measurement. Stratified designs cover heterogeneity, and repeat visits quantify temporal variability. Record coordinates, time, instrument serial numbers and conditions at the time of sampling.

It treats data gaps and biases explicitly, handles autocorrelation in statistics, compares products with independent observations, presents scenarios as scenarios and communicates uncertainty visually and in words.

Our support covers analysis, modelling and writing; official hazard or design values come from authorities and standards.