Five years ago I gave at talk at the International Mycological Congress, IMC10, in Bangkok, on the distributions of crop pests and pathogens around the world. One aspect which hadn’t received much attention was the problem of pseudo-absences in the pest distribution data we use, obtained from the CABI Knowledge Bank. CABI’s pest distribution data is very commonly used on global analyses of pest distributions, and may be the best dataset out there. CABI have been mining publications for observations of forestry and agriculture pests since the 1940s, and there are now more than 2000 such organisms in their databases.
I used a statistical technique, in a paper published in New Phytologist, to estimate the numbers of unobserved pests and pathogens in different countries around the world. Briefly, the statistical model included per capita GDP and investment in R&D per country as predictors. Unsurprisingly, both these variables are predictors of numbers of pests and pathogens reported per country, because they indicate the likely scientific and technical capacity of that country to detect, identify and report pest presence.
To estimate the true numbers of pests per country, I made predictions from the model, but changing all countries’ per capita GDP and R&D investment to that of the USA, because the USA is likely to be very good at detecting and reporting pests. The results show that many developing countries are likely under-reporting hundreds of agricultural pests, with serious implications for crop protection and food security – if we don’t even know what’s there, how can we deal with it? China was listed as the country with the most ‘missing’ pests…more than 1000, in fact.
This prompted one of the audience at my talk, Peter Mortimer of Kunming Institute of Botany, to ask me whether the CABI database included observations published in the Chinese literature. The Chinese literature is largely independent and isolated from Western academia, and only in recent years have Chinese scientists begun publishing in English-language journals (to great success – China is now the world’s second-largest country by papers published).
We decided to test this hypothesis by making predictions of pest presences using the existing CABI database, and checking these predictions by searching for pest observations in the Chinese literature. Elsa Field, a student at Oxford University, obtained a BSPP vacation bursary to visit China and work with Peter’s team to search through the Chinese literature for one hundred ‘missing’ pests. Of those hundred, Elsa and researcher Gui Heng found 27 reported in China, and these were the ones that our predictive model said would be most likely to be present.
Our predictive model itself is simple, using estimates of the global distributions of host crops, distance from the coast, and presence of pests in neighbouring regions to estimate pest presence. This makes the model a potentially useful method to home in on the most risky pests and pathogens when making decisions about which of the thousands of threats to global agriculture we should focus our attention upon. Pest Risk Analysis, or PRA, is mostly done using expert opinion to prioritize effort. Our model provides a way to statistically quantify this exercise.