No area of work is more closely associated with data analysis than agriculture. Statistics as a concept grew out of the need to quantify agricultural production for taxation purposes. Modern statistics as a coherent, mathematically-based field was shaped by a need to quantify agricultural investigation. Most generic methods can be traced back to an agricultural point of origin. The great landmark adventure of recent years, the sequencing of the human genome, concludes an intellectual story line that started with a statistical view of inheritance in peas. Astronomy and mathematics owe most of their history to the need to better predict seasons and the flooding of farmlands.
And it is inevitably so. No other strategic concern of the human race is so much at the mercy of a statistical universe. Within the controlled environments of manufacturing, ever smaller numbers of components escape the ever-tightening loop of quality control. In the fields, we remain concerned with the percentage of crop that can be brought to usable yield.
Agriculture has, to be sure, done serious quality control loop tightening of its own; but fewer of the affecting factors are amenable to human control than is the case in manufacturing – and the degree of control is less complete by many orders of magnitude. If production quality for left-handed silicon widgets is temperature dependent then I can easily control the temperature of the manufacturing environment within a few degrees. If I am a wheat farmer, however, I am limited to modest and partial amelioration of what the climate throws at me. The individual ear of wheat cannot be protected; my aim has to be maximising the number of ears per hectare that survive to be harvested and eaten.
Some day, perhaps, all food will be synthetic: biomass grown in orbital tanks, on nutrient broth mined from the asteroids and Saturn’s rings, then knitted into palatable forms, all under quality control comparable to current silicon chip production. Perhaps. That day is a long way off, though. More modest industrialisations, such as greenhousing, polytunnel cloching and hydroponic production are with us already, but are generally focused on market value position or marginal capacity. In the foreseeable future, a rising population will for the most part depend upon nutrition grown in dirt, under sky, battered by weather. In improving the efficiency with which we discern and exploit the patterns within that environment, computerised methods are as important as in any other area of science.
In this battle to maximise yield, data analysis is applied across a range of scales, but they can be split crudely into tactical and strategic approaches. Strategic application seeks to refine global understanding of the factors that affect success; tactical use is aimed at specific application of that understanding to a particular area, location or context. Linking the two, embracing both, is a mesh of research and modelling cycles.
All of this makes for a huge subject, and trying to address agricultural data analysis in a few pages has more than a hint of hubris. All that anyone can hope to achieve is a representative lucky dip into diverse topics. [read more...]
No comments:
Post a Comment