Of elephants, soil, crops, rocks and numbers. A random-walk of a career in soil(ish) science

“In our latest Your Career In Soil Or Plant Sciences segement, British Geological Survey’s (BGS) Murray Lark discusses his varied journey”

My career as a scientist started early, in central Africa where I spent my childhood.  The main questions that absorbed me were ecological ones.  In particular I was interested in the effect that the elephant population had on trees.  It occurred to me that this might be measured by comparing two airphotos of the same part of the Hwange Reserve, taken at different times.  Most 10-year olds would not have access to air photographs, but my father used them routinely in his work laying out building projects, and was able to help me acquire the necessary material,  two air photographs of the same area, about six years apart.

At this point I learned one of my first scientific lessons.  Even the most devoted ecologist could not count every tree on each photo.  Clearly I needed a sampling strategy.  The resulting data showed a marked decline in the number of trees, and I still use them today for teaching sampling with paired observations at fixed monitoring plots, and the selection of a sample size by power analysis.


A young Lark with friends, around 1976

So far so zoological, but I was not unaware of the soil.  I recall my long-suffering mother allowing me to dry soil samples in her oven so that I could compare the moisture content  of soils under two contrasting vegetation covers (Uapaca kirkiana and Brachystegia /Julbernadia).  I also determined their organic content by a (rather unreliable) loss on ignition method that involved an old spirit burner, a lot of meths, smoke and singed clothing and fingers.

Pedological awareness started to stir when I observed that my school, at a rather higher elevation than my house, had powdery talc-like red soil (which stuck tenaciously to school uniform), at home the duller-coloured soil was punctuated by lumps of massive lateritic ironstone, with pisolithic gravel which I could pick up with a magnet.  Further down slope, on the margins of a snake-infested vlei, the soils were black, heavy and sticky.  Some years later, when reading about  the soil catenas of central Africa I realized that I had grown up on one.

Time passed, and I found myself living in Britain.  As a sixth-former I started reading about statistics, and was intrigued.  I also got a voluntary position at the Natural History Museum in South Kensington where, along with routine jobs for the zoology department, I spent some time studying a collection of elephant jaws from Uganda, on which I based my first paper (Lark, 1984A comparison between techniques for estimating the age of African Elephants, Loxodonta africana. African Journal of Ecology. 22, 69–71).  I nearly melted a Sinclair Spectrum computer after programming it to calculate correlation matrices of measurements made on the original elephants. From statistics my interests turned to the mathematics of population dynamics, population genetics and evolution and game theory.  With those interests it seemed natural to apply to read Zoology at Oxford, and I started there in 1984, determined to specialize in mathematical biology.  Fate had other ideas.

In the first two terms at Oxford all biologists studied a common preliminary course.  Along with biological topics there were lectures on the physical environment given by geologists and Dr Philip Beckett from the soil science laboratory.  Philip’s first lecture was on the sequence of nick points, and their associated rejuvenated weathering surfaces along the Zambezi river, and the mosaic of soils and vegetation that these created.  I was transfixed, not only by the slides of the Zambezi landscapes which I had regularly traversed as a child, but by the combination of earth science and biology in an overall account of the landscape and its ecology.  After the lecture I talked with Philip, and he later showed me some maps and results from a 1960s University expedition to Zambia where he had met one Richard Webster, later to do his doctorate with Philip on the assesment of soil information derived from air photographs.  I started reading about soil science.  What clinched my future direction was reading a book by Peter Nye and Bernard Tinker “The Soil Under Shifting Cultivation”.  This considered shifting cultivation systems, drawing together data sets which they assessed, not by some tedious data mining but by a simple single-pool carbon model which allowed them to show how the frequency with which a particular piece of ground comes out of fallow determines the sustainability, or otherwise, of the soil’s fertility.  I summoned up my courage and told my college tutor, Richard Dawkins, that I was switching from Zoology to Pure and Applied Biology, a new course which allowed me to spend a lot of time on biometrics and soil science.  I had a glorious three years, culminating in a summer job after finals where I put together all the training from field classes and lectures on soil science and made a map of new farmland that the University had purchased to the south and west of Wytham wood.  I spent long hours with no company but sheep, and once emerged from a pit to find a small group of them standing in an intrigued semi-circle around me.

I stayed on after graduation to do a doctorate with Philip Beckett on multivariate and spatial analysis of remote sensor data.  After a University of Wales fellowship I started a Post-Doc at the then Silsoe Research Institute (BBSRC) working on statistical problems in precision agriculture.  I was given a permanent job in the Maths Group, and built up a team working on environmental statistics with particular emphasis on soil monitoring and management.  In due course this group moved to Rothamsted where Dick Webster joined us as a Lawes Trust fellow.  We had a good deal of fun developing geostatistical methodology for soil science, paying a lot of attention to sampling problems, and using an emerging mathematical method, the wavelet transform,  to learn about the spatial variability of nitrous oxide emissions in the landscape. This gave us insight into how one of the key assumptions that underlies geostatistical analysis can break down in complex landscapes.  This is the assumption of stationarity in the covariance, whereby our model assumes that the variance and scale dependence of a property is uniform in space.  By couching the underlying statistical problem in terms of a strategically important problem in land management (greenhouse gas emissions, again), we persuaded BBSRC to fund a project in which we developed methodology to handle such non-stationary variation in the geostatistical model.

Almost every project involved the collection of new data, whether through laboratory analysis of soil or field measurements of soil properties, remote sensing, hand harvesting of plots or plot combining and yield monitors on commercial combines.  We got a lot dirtier and wetter than is usual for statisticians.


The writer making holes in Bedfordshire

In late 2010 I was offered a new role at the British Geological Survey at Keyworth near Nottingham.  I had collaborated with BGS, particularly Barry Rawlins, for some years.  BGS was keen to make more use of statistical methodology across its programme.  I decided it was time for a change.  I now work with colleagues on topics from volcanic hazards to surface water geochemistry, marine habitat mapping and the uncertainty in 3-D geological models of the subsurface.  However, there is still plenty of soil science.  Most recently I have become involved with a network project, funded by the Department for International Development through the Royal Society.  This is based in Zambia, Zimbabwe and Malawi, and is focussed on soil geochemistry and agriculture with particular interest in micronutrient status of crops and also the uptake of potentially hazardous elements near mines and mine spoil.  This has given me the opportunity to return to my native Africa and the soil landscape which first captured my imagination, but this time with statistical tools to plan sampling and analyse data, and rather better lab back-up than my old spirit burner.


A Subsistence farmer’s house and fields, in the shadow of a smelter, the Copper Belt, Zambia

So what advice would I offer to someone starting a career in soil science?  Here are a few stratified random thoughts.

Although you, gentle reader, are probably a lot smarter and more organized than I am, I would suggest that excessively detailed career-planning is largely a waste of time.  Follow what excites you, and have the flexibility to take up opportunities that the contingencies of history throw across your path.  Most scientists will develop a personal agenda of questions that they want to answer.  Make sure that you develop such an agenda.  However, in the real world of scarce resources, one of the most important skills is an ability to show others (with more money than you have) why their strategic problems will be best solved if you can tackle your scientific ones.  With food security still a matter of global concern, and climate change a continuous background threat, there are many opportunities to do this in soil science.

image description

The writer in a hole in Malawi looking at a Vertisol soil

Second, while it is important to develop an area of expertise in which you are known to be useful, don’t be put in a box.  Philip Beckett advised his students completely to change their area of focus every decade or so, and while that probably isn’t manageable in the modern world  make sure that you never stop learning about new subjects and take the opportunity to mix things up a bit from time to time.

Third, never underestimate the value of working scientific relationships.  It is important to cultivate the skills of a team-worker, someone who makes a positive virtue of working with others and who knows how to make connections between practitioners in different disciplines.  I have learned an enormous amount from previous and current colleagues and students.

Finally, never dig a hole without having a clear idea how you intend to analyse the resulting data, and never plan a data analysis unless you can say what hypothesis you are setting out to test.  Very little progress has ever been made by the strategy of measuring everything in sight and then leaving it to a GIS/ machine learning algorithm/random forest to extract the information

Oh, and one more thing.  Let your kids mess around with photogrammetry, set fire to stuff and fraternize with dangerous wildlife.  And encourage any interest in numbers, you never know what it might lead to.

Author: Murray Lark

Twitter: @GeostatLark