For many years, several research groups and international organisations have developed high-resolution global maps of livestock based on reported statistics derived from census and survey data. Due to the dynamic nature of these populations (intensification of livestock production) this is an extremely challenging task that involves the collection of contemporary statistics, the development of complex algorithms to disaggregate these statistics at higher spatial resolutions and make predictions where they are absent.
The main objectives of our ongoing work are to evaluate quantitatively and qualitatively the different approaches that are used to predict and represent livestock distribution in high-resolution global databases. Our ultimate goal is to improve the existing methods by addressing some of their limitations, and to provide recommendations and revisions for future developments. This requires a detailed examination of both the conceptual and statistical models that are used to generate these databases, and the ways in which the outputs are presented and distributed to users. More specifically, the project is built around three specific objectives. More specifically, the current research activities involve: i) comparing the different methods and database available to map the distribution of livestock in terms of the goodness of fit of the results: how similar are the numbers in observed and predicted distributions, and in terms of spatial patterns; how similar are observed and predicted distributions in terms of heterogeneity and level of clustering; ii) examining how uncertainty in the predictions can best be estimated and, once estimated, how it can be best represented and communicated to users, and iii) identify innovative ways to improve the spatial realism of outputs.
Further developments involve breaking down livestock distribution maps according to different types of production systems. In addition, our current research primarily aims to understanding and predict the spatial distribution of livestock production as it is now, but future work will involve the development of methods to derive projections of future distribution according to different development scenarios.
Our main institutional collaborators are T. Robinson (International Livestock Research Institute, ILRI, Nairobi, Kenya), G. Cinardi and A. Mottet (Food and Agriculture Organization, FAO, Rome, Italy), W. Thanapongtharm (Department of Livestock Development, DLD, Bangkok, Thailand), H. Yu (Chinese Center for Disease Control, CDC, Beijing, China), and our main academic collaborators are W. Wint (ERGO, Oxford, UK), S. Hay (SEEG, Univ. Oxford, Oxford, UK), T. Van Boeckel (Dpt Ecol & Evol Biol, Princeton, USA) & S. Vanwambeke (UCL, Louvain-la-Neuve, Belgium).
Using Random Forest to Improve the Downscaling of Global Livestock Census Data
G. Nicolas, T.P. Robinson, G.R.W. Wint, G. Conchedda, G. Cinardi and M. Gilbert.
In "PLOS ONE", vol. 11, issue 3, 2016.
Income Disparities and the Global Distribution of Intensively Farmed Chicken and Pigs
M. Gilbert, G. Conchedda, T.P. Van Boeckel, G. Cinardi, C. Linard, G. Nicolas, W. Thanapongtharm, L. D'Aietti, W. Wint, S.H. Newman and T.P. Robinson.
In "PLOS ONE", vol. 10, issue 7, 2015.
Mapping the Global Distribution of Livestock
T.P. Robinson, G.R.W. Wint, G. Conchedda, T.P. Van Boeckel, V. Ercoli, E. Palamara, G. Cinardi, L. D'Aietti, S.I. Hay and M. Gilbert.
In "PLoS ONE", vol. 9, issue 5, 2014.