Details of Award
NERC Reference : NE/M006131/1
Design-based and Model-based Statistics for Environmental Sampling
Training Grant Award
- Lead Supervisor:
- Professor RM Lark, British Geological Survey, Environmental Modelling
- Grant held at:
- British Geological Survey, Environmental Modelling
- Science Area:
- Atmospheric
- Earth
- Freshwater
- Marine
- Terrestrial
- Overall Classification:
- Atmospheric
- ENRIs:
- Biodiversity
- Environmental Risks and Hazards
- Global Change
- Natural Resource Management
- Pollution and Waste
- Science Topics:
- Environmental Statistics
- Optimal Design
- Sampling
- Spatial Statistics
- Statistical Estimation
- Statistics & Appl. Probability
- Abstract:
- There are many sophisticated methods to measure properties of the environment whether it is run-off from a catchment or the shape of pores in a clod of soil. However, the validity of the conclusions that we draw from these measurements depends on the way in which we select the sites or individuals that we measure. Without a sound sampling strategy it is all too easy to draw unreliable conclusions from data, and the considerable cost of field work and measurement can largely go to waste. Statistical sampling is a large and growing area of statistical science. It has its roots in the early days of applied statistics, but its concepts are still developing and the toolbox of sampling methods continues to evolve. Classical sampling is done by "design-based" methods, which mean that we use an element of randomization in our sampling. In this context we can be confident that our sampling is unbiased, and we can compute reliable statistics that describe the uncertainty in our resulting estimates. We can also ensure that we sample just enough to make acceptably precise estimates without wasting effort. Despite these good properties scientists surprisingly often avoid randomization in their sampling, questioning whether it is practically feasible. Often this is because they are unaware of some of the more advanced design-based sampling methods which are more complex than the simple sample designs taught as part of introductory courses. This course will equip early-career researchers with an understanding of design-based sampling and the range of techniques available. Another group of sampling methods are called "model-based". This is because they use a mathematical model of how environmental properties may vary to substitute for some of the properties of design-based sampling. Model-based sampling is usually more appropriate if we want to map a variable in space (or space and time) or if we need to sample a property at regular intervals in time and it can be particularly useful when we want to use "legacy" data sets, collected in the past, to answer new questions which were not envisaged when the original sampling was done. Model-based sampling can also be optimized in a variety of ways to make it more efficient and cost-effective. Most environmental scientists do not learn about these methods in their early training. This course will equip them to design efficient sampling schemes for tasks such as mapping and, often most importantly, to make statistically sound use of existing data sets collected on grids or transects (regular sampling along a line). This course will provide participants with an understanding of the theory behind design and model-based sampling, understanding of how sampling is implemented in practice, experience of analysing data from different sampling schemes using a free computer package (R) which they can then use in their own research, understanding of how to make sampling efficient, and practical tasks in sample planning and design based on real-world case studies. BGS staff involved are uniquely placed to offer this training. The course lead has substantial experience of research in statistical methods for environmental science, including sampling design. He is joint author of a recent textbook on sampling, and has experience of giving courses in applied statistics to researchers and students in the UK, Europe and Australia. Both staff involved are familiar with the R package and other computer tools. They are also involved in a daily basis in the design and implementation of sampling schemes. To take some recent examples, they have designed a geochemical sampling scheme for southern England (recently implemented), a national scale soil monitoring scheme for the United Kingdom (with collaborators), optimized sampling schemes for measuring weed density in agricultural fields, optimized networks for sensor arrays and a pilot scheme to monitor soil erosion across England and Wales.
- NERC Reference:
- NE/M006131/1
- Grant Stage:
- Completed
- Scheme:
- Doctoral Training
- Grant Status:
- Closed
- Programme:
- Advanced Training
This training grant award has a total value of £17,572
FDAB - Financial Details (Award breakdown by headings)
| Total - Other Costs |
|---|
| £17,572 |
If you need further help, please read the user guide.