Poultry farm distribution models developed along a gradient of intensification

Highlights

• Intensive farms were found clustered in Nigeria, Thailand, Argentina and Belgium.

• The gradient of clustering did not reflect the gradient of intensification.

• We developed farm distribution models (FDM) to predict farm location and sizes.

• Improvements should consider covariates explaining better farm locations and sizes.

Abstract

Efficient planning of measures limiting epidemic spread requires information on farm locations and sizes (number of animals per farm). However, such data are rarely available. The intensification process which is operating in most low- and middle-income countries (LMICs), comes together with a spatial clustering of farms, a characteristic epidemiological models are sensitive to. We developed farm distribution models predicting both the location and the number of animals per farm, while accounting for the spatial clustering of farms in data-poor countries, using poultry production as an example. We selected four countries, Nigeria, Thailand, Argentina and Belgium, along a gradient of intensification expressed by the per capita Gross Domestic Product (GDP). First, we investigated the distribution of chicken farms along the spectrum of intensification. Second, we built farm distribution models (FDM) based on censuses of commercial farms of each of the four countries, using point pattern and random forest models. As an external validation, we predicted farm locations and sizes in Bangladesh. The number of chicken per farm increased gradually in line with the gradient of GDP per capita in the following order: Nigeria, Thailand, Argentina and Belgium. Interestingly, we did not find such a gradient for farm clustering. Our modelling procedure could only partly reproduce the observed datasets in each of the four sample countries in internal validation. However, in the external validation, the clustering of farms could not be reproduced and the spatial predictors poorly explained the number and location of farms and farm sizes in Bangladesh. Further improvements of the methodology should explore other covariates of the intensity of farms and farm sizes, as well as improvements of the methodology. Structural transformation, economic development and environmental conditions are essential characteristics to consider for an extrapolation of our FDM procedure, as generalisation appeared challenging. We believe the FDM procedure could ultimately be used as a predictive tool in data-poor countries.

Introduction

Population and income growth, urbanisation and technological advances have led to the intensification of livestock production systems over the past decades. Although this trend in intensification is observable globally, the level of intensification varies between countries and it mainly occurs today in low- and middle-income countries (LMICs). Intensification leads to large-scale and specialized production units, market-oriented and capital- and input-intensive. Such production units make heavy use of feed concentrates, antibiotics and vaccination. The increase in production meets the increasing demand for animal source food, but it has tremendous effects on livelihoods, animal and human health. Concentration of monogastric species have led to pollution of land and water due to the overload of improperly managed manure (Gerber et al., 2005). In terms of public health, intensive livestock production fosters the emergence of infectious diseases (Jones et al., 2013) and often leads to increased use of antibiotics, which is likely to increase antimicrobial resistance (Aarestrup, 2005; Vieira et al., 2011; Chantziaras et al., 2014; Van Boeckel et al., 2015). Livestock also traditionally represents an important source of livelihood for rural populations in LMICs but the intensification of livestock production seldom benefits the local population (FAO, 2018).

As production intensifies, the geographical distribution of monogastric animal farming changes. At a pre-industrial stage, chickens and pigs are mainly associated with human settlements for their role of waste converters (Perry et al., 1999; Steinfeld et al., 2006), with farms homogeneously distributed among rural populations. This situation is encountered in traditional production systems of LMICs (Steinfeld et al., 2006). Production intensification, in parallel to urbanisation, leads larger farms to locate within the peri-urban belt of major consumption centres (Steinfeld et al., 2006). Livestock products being perishable, the production profits from being located close to demand centres. However, as cities expand, land value in the periphery increases and, with improving transport infrastructures, farms move further from peri-urban belts, with the advantage of being closer to feed production or transportation areas (Steinfeld et al., 2006), and benefits from agglomeration economies (Roe et al., 2002a; Herath et al., 2005; Larue et al., 2011). This evolution is particularly pronounced for monogastric species, which are raised in landless production systems (Naylor, 2005; Steinfeld et al., 2006).

As the distribution of farms changes as it intensifies, so do the effects of livestock raising activities on their surroundings. Highly-detailed maps of livestock production allow the assessment of the local effects of intensive production (Robinson et al., 2011). In the case of disease outbreaks for instance, adequate control is planned based on mathematical modelling of infectious disease which require farm distribution and sizes (number of animals) (Hill et al., 2018; Tildesley and Ryan, 2012). Accurate data on the distribution and stock of farms rarely exist as agricultural censuses are time and resource consuming. In high-income countries, censuses are usually conducted but data access may be restricted due to confidentiality and privacy reasons. In the United States for instance, livestock data are only provided as a number of animal per county (Tildesley et al., 2010). In LMICs, censuses are not systematically carried out and data from occasional censuses are available as total numbers of animals per administrative units. However, the resolution at which these are provided vary greatly across countries (Robinson et al., 2011).

Downscaling procedures that increase the spatial resolution of existing livestock data, such as the Gridded Livestock of the World (GLW) are available (Wint et al., 2007; Robinson et al., 2014; Gilbert et al., 2018). However, these models predict livestock as a continuous, gradually varying, density of animals per pixel in a raster. Such data do not provide information at farm level, and densities of animal per pixel do not specify any information on the production systems. As production intensifies, livestock are fed on imported feed concentrates rather than on crops produced locally. Livestock production is hence less determined by environmental characteristics of the surrounding lands (Naylor, 2005; Steinfeld et al., 2006). Moreover, the spatial clustering of intensive farms is poorly explained by spatially continuous surfaces (Van Boeckel et al., 2012; Robinson et al., 2014). As suggested by Gilbert et al. (2015), a method accounting for the spatial clustering of intensive farms would be required to predict their distribution. In a previous study, we developed farm location models in Thailand, based on point pattern analysis methods (Chaiban et al., 2019). Other authors have predicted cattle population on individual farms in New Zealand with a zero-inflated Poisson regression, but only carried out an internal validation (van Andel et al., 2017). Burdett et al. (2015) modelled farm locations based on detailed geographical information as well as total animal and number of farms at county level. In this study, we present a methodology to predict both the location and the size of farms even where spatial information is scarce and which would account for the spatial clustering of farms.

We selected four countries to develop our farm distribution models, i.e. Nigeria, Thailand, Argentina and Belgium (hereafter referred to as the “sample countries”). These sample countries were assumed to have different levels of intensification as they have increasing levels of Gross Domestic Product (GDP) in the following order: Nigeria, Thailand, Argentina and Belgium. The GDP per capita (in purchasing power parity) is strongly associated with the level of intensification of livestock production of a country (Robinson et al., 2011; Gilbert et al., 2015). As farms tend to be more clustered as production intensifies (Abdalla et al., 1995; Roe et al., 2002a), we hypothesized that intensive farms would be spatially more clustered with increasing GDP. In addition to being at different stages of the intensification process, these countries present different ecological, demographic and socio-economic conditions. Developing farm distribution models (FDMs) along a wide range of conditions allows to assess how these model compare to each other in very different settings.

We first aimed to investigate the spatial distribution of commercial farms along the spectrum of intensification. Second, we built FDMs for intensive farms while accounting for the spatial clustering of farms. We then assessed the capacity of our modelling procedure to reproduce the farm distribution observed in Bangladesh, where farm locations and sizes were available.