Landscape attributes driving avian influenza virus circulation in the Lake Alaotra region of Madagascar

Abstract

Abstract. While the spatial pattern of the highly pathogenic avian influenza H5N1 virus has been studied throughout Southeast Asia, little is known on the spatial risk factors for avian influenza in Africa. In the present paper, we combined serological data from poultry and remotely sensed environmental factors in the Lake Alaotra region of Madagascar to explore for any association between avian influenza and landscape variables. Serological data from cross-sectional surveys carried out on poultry in 2008 and 2009 were examined together with a Landsat 7 satellite image analysed using supervised classification. The dominant landscape features in a 1-km buffer around farmhouses and distance to the closest water body were extracted. A total of 1,038 individual bird blood samples emanating from 241 flocks were analysed, and the association between avian influenza seroprevalence and these landcape variables was quantified using logistic regression models. No evidence of the presence of H5 or H7 avian influenza subtypes was found, suggesting that only low pathogenic avian influenza (LPAI) circulated. Three predominant land cover classes were identified around the poultry farms: grassland savannah, rice paddy fields and wetlands. A significant negative relationship was found between LPAI seroprevalence and distance to the closest body of water. We also found that LPAI seroprevalence was higher in farms characterised by predominant wetlands or rice landscapes than in those surrounded by dry savannah. Results from this study suggest that if highly pathogenic avian influenza H5N1 virus were introduced in Madagascar, the environmental conditions that prevail in Lake Alaotra region may allow the virus to spread and persist.

Agro-Environmental Determinants of Avian Influenza Circulation: A Multisite Study in Thailand, Vietnam and Madagascar

“…A seasonal pattern has been observed for occurrence of HPAI H5N1, with higher risk of HPAI H5N1 outbreaks from November to January and April to June [45]. Temporal variations were also previously described for seroprevalence of low pathogenic avian influenza, including H9N2 [46] and cannot be excluded in Madagascar highlands [36]. Given this discrepancy in data collection across study sites, we may thus have compared the yearly situation captured in Thailand and Vietnam highlands, with a “high-risk” period in the Red River Delta (Vietnam) and with an AIV situation corresponding to the dry season only in Madagascar.…”

“…It is probable that water from the lake, rivers and rice paddy fields play an important role in the transmission dynamics of low pathogenic AIV in the poultry population of Lake Alaotra through an oral-fecal route. In Lake Alaotra, AIV circulation also may be favored on farms located close to water bodies by the presence in these areas of important wild waterfowls congregations [36], which are a natural reservoir of low pathogenic AIV [42]. Another interesting characteristic of Lake Alaotra is the presence of flocks of ducks and geese, which are brought during the day to graze on rice paddy fields like in Southeast Asia.…”

“…Given this discrepancy in data collection across study sites, we may thus have compared the yearly situation captured in Thailand and Vietnam highlands, with a “high-risk” period in the Red River Delta (Vietnam) and with an AIV situation corresponding to the dry season only in Madagascar. The temporal pattern of AIV circulation in poultry could be explained by different factors including varying contacts between poultry and wild birds, virus survival under climatic conditions, proportion of land covered by flooding [39], but also seasonality of poultry production and trade [36]. Despite the fact that data on the distribution in space and time of both AIV and all these factors are very difficult to obtain at high resolution, further studies should seek to include simultaneously agro-environmental, climatic and poultry trade related variables to eliminate any possible confounding.…”

“…In Vietnam, a series of MODIS images collected in 2005 at 500-m resolution (https://lpdaac.usgs.gov/products/modis_products_table/mod09a1) were processed to map water bodies and rice paddy fields, using methods described elsewhere [34], [35]. In Madagascar, one Landsat 7 enhanced thematic mapper plus (ETM+) satellite image centered on Lake Alaotra, of 30-m resolution and dated from March 2007 was analyzed using supervised classification [36]. Landcover classifications in Vietnam and Madagascar were validated by field visits [16], [36].…”

“…In Madagascar, one Landsat 7 enhanced thematic mapper plus (ETM+) satellite image centered on Lake Alaotra, of 30-m resolution and dated from March 2007 was analyzed using supervised classification [36]. Landcover classifications in Vietnam and Madagascar were validated by field visits [16], [36]. The Euclidian distances from each study location (village or farm) to the nearest permanent body of water were calculated.…”