Friday, 6 February 2015

Random variables and field sampling... by Murray Lark

Murray (left) getting some local help
It's not just geologists that travel to far away landscapes unearthing knowledge and understanding about the planet we live on! Each research trip needs it's own unique fieldwork team, a special blend of disciplines and expertise. In one trip to central Africa Murray Lark, an environmental statistician, accompanied two environmental geochemists, Michael Watts and Elliott Hamilton, on a search to understand the health implications of trace elements in the soil. Here Murray tells us why his role before-during-after hands-on fieldwork is so essential... 

It is always good when the environmental statistician gets to discuss field sampling with scientists before they head out to collect their material.  That way we can ensure, given all available prior information, that the data resulting from the sampling are most likely to be useful for the research. It is even better when the environmental statistician gets to participate in the field work.  That way we can see the practical issues of sampling on the ground, and sometimes make adjustments to the plan on the go.  And, in my case anyway, I enjoy field work as much as any of my geologist colleagues, particularly when it takes place in the landscape where I first developed an interest in environmental science: the miombo woodland of my native central Africa.

Sampling team, made up of BGS and colleagues from
Copperbelt University and the Zambian Agricultural
Research Institute, in front of a large termite mound.
A previous blog by Michael Watts 'Geochemistry brings societal benefits to sub-Saharan Africa... by Michael Watts' explained that BGS is beginning a five-year project in geochemistry with partners at the University of Nottingham and from universities and institutes in Zambia, Malawi and Zimbabwe.  The project will address a range of important questions related to human health through research to understand the behaviour of trace elements in the soil including essential micronutrients, often in deficiency, and potentially toxic elements, often associated with waste from mining activity past and present.  In November, under the aegis of the Centre for Environmental Geochemistry and BGS Global, Michael, myself and Elliott Hamilton visited Zambia to undertake some exploratory fieldwork with colleagues from the Copperbelt University and the Zambian Agricultural Research Institute.

Before travelling to Africa I met Michael and Elliott to discuss their objectives for the sampling.  First, they wanted to know what trace elements in the soil might be present in potentially harmful concentrations and whether these could be shown to arise from human activities such as smelters or the disposal of "tailings", waste products from the first stages of processing of ore.  Second, they needed to know something about the variability of elemental concentrations to aid planning of further more detailed surveys and the layout of experimental plots.   That means we need to know something about the importance of variation at different spatial scales.  For that reason I proposed a spatially nested sampling scheme around sampling "mainstations" or nodes set out on transects across potential environmental gradients of interest (e.g. downwind from a tailings heap, or a smelter).

Bagging first sample on cultivated land
near the margin of a dambo
In a nested sampling scheme we collect samples at clustered locations separated by fixed distances but at random directions from each other.  That allows us to analyse the variation of our observations into components associated with each distance, and so to identify the scale at which the important variation happens in a particular landscape.  That then allows us to answer questions such as "how densely do we need to sample to make a map of chromium concentration across a group of farms?" or "how can we design experiments to make sensitive tests of the effects of crop management on element uptake, given the variability of concentrations in the soil?"

We put these principles into practice in Zambia, sampling on transects to explore gradients expected due to the movement of material by wind from smelters or mine tailings and with a nested design at each node to explore variation at scales from 1m to 100m.  The soil is currently being processed in the laboratory, but soon we should be able to complete the spatial analyses, to learn something about the nature of the problem and how the next phase of sampling should be planned.

Murray giving a lecture on statistical methods used at the BGS
The miombo woodland (now largely cleared of trees for farming) is found on old land surfaces across central Africa.  These are drained by seasonal water courses or dambos, under grass with occasional scattered trees.  The soil in the dambos is fertile, with clay minerals reformed from weathering products that move laterally downslope.  Iron and aluminium is also mobilized by weathering, and forms hardened duricrusts or cuirasses where it approaches the surface.  The main "ecological engineers" in the soil of the miombo are the termites, and their enormous mounds, sometimes topped with trees, are a characteristic feature of the lanscape.  It was a great pleasure to reacquaint myself with this environment while designing and implementing sampling schemes in the bush!

While visiting the Copperbelt University, Zambian Agricultural Research Institute and, later, the University of Zimbabwe, I gave a lecture on the statistical methods that we use at BGS for analysis of spatial data and to plan robust and rational sampling schemes.  There was interest in how these methods could be used as the geochemistry project develops, and also in how they might be applied to elucidate other problems connected to soil carbon content under farming and the use of old soil survey data for current questions about how to monitor the soil.


To read Michael's previous blog follow this link.

No comments: