RESEARCH COMMUNICATIONS
*For correspondence. (e-mail: sarojkbarik@gmail.com)
disease for which apparently there is no vector, except a
report describing two blow fly species, namely Calliphora nigribarbis and Aldrichina grahami, which are
suspected7 to be mechanical transmitters for H5N1. More
recently, a mosquito species, namely Culex tritaeniorhynchus tested positive for H5N1, although its role as a
potential vector has not been evaluated8. The disease has
become a menace in recent times, causing death of
millions of poultry birds and leading to huge economic
losses9. Reports on sporadic human cases of H5N1 have
aroused concerns on triggering of a probable global influenza pandemic10,11.
The highly pathogenic strain H5N1 is a subtype of the
influenza A virus belonging to the family Orthomyxoviridae. The strain is endemic to some parts of Asia and
has established a niche in poultry12,13. It is reported to
have multiple sublineages in Vietnam14. The wild migratory
birds, considered to be the natural reservoir of all
influenza A viruses, are also dying due to H5N1 infection13,15. Although aggressive control measures are being
taken, including culling of more than 140 million poultry,
outbreaks have recurred in different parts of the
world13,16.
The wild migratory aquatic birds are suspected to be
the prime agents of geographic-scale spread of the pathogen at global level15,17, while the infected poultry and
ducks are important agents at the landscape level17,18.
However, explanations on actual transmission of the virus
are generally suggestive. For instance, water is argued to
be an effective medium for the transmission of the avian
influenza virus because of its high prevalence among the
aquatic wild birds that congregate in the wetlands15. The
transmission of the virus takes place through direct contact with the infected birds or by using the bird products
like faeces for fertilizer and aquaculture17. Other studies
have reported strong influences of environmental factors on
the transmission of avian influenza. For example, the
transmission of the HPAI in Nigeria and West Africa has
been linked to differences in plant phenology and landsurface reflectance19. Studies in Thailand and Vietnam
revealed that the transmission cycle of the pathogen is
strongly associated with duck farming, paddy cultivation
and human population density20,21. Thus, our understanding of the geographical spread and factors responsible for
the transmission of HPAI virus remains poor.
The number of HPAI outbreaks in the Indian subcontinent has increased over the years, causing huge economic
losses16. Its rapid transmission in the subcontinent has
been attributed to a number of causes, which are based on
wild speculations without having any credible scientific
basis. Thus, the transmission cycle of the virus largely
remains as an incomprehensible ‘black box’. In the present study, an attempt has been made to understand and
model the ecological niche of the HPAI (H5N1) considering the transmission cycle as a ‘black box’. The study
would help in predicting the areas having potential for its
CURRENT SCIENCE, VOL. 97, NO. 1, 10 JULY 2009
73
Modelling the ecology and distribution
of highly pathogenic avian influenza
(H5N1) in the Indian subcontinent
D. Adhikari, A. Chettri and S. K. Barik*
Centre for Advanced Studies in Botany, School of Life Sciences,
North-Eastern Hill University, Shillong 793 022, India
The highly pathogenic avian influenza (HPAI) type
H5N1 has become a menace in the recent years, causing huge economic losses to the poultry industry.
However, a proper understanding of the factors governing its transmission cycle and possible outbreaks in
space and time is lacking. The ecological niche modelling tool, Genetic Algorithm for Rule-Set Prediction
(GARP) has been used to model the ‘black box’ of
H5N1 transmission in the Indian subcontinent. Ninety
outbreak records for veterinary cases of H5N1 in
West Bengal and Bangladesh, and 18 environmental
parameters with a pixel resolution of ≈4 km were used
in the modelling. The final optimized rules generated
by GARP were projected on a broad geographical
scale covering India, Bangladesh, Nepal and many
parts of Pakistan to anticipate the potential areas for
the presence of H5N1. The niche model predicted vast
areas in the Indo-Bangladesh region for the presence
of H5N1. The areas having high level of model agreements are in the Gangetic Plains, Assam, Tripura,
West Bengal and some parts of Andhra Pradesh, Maharashtra and Gujarat. The model also predicted
large areas in northern Pakistan and some parts of
Nepal to be suitable for H5N1 transmission. The modelled ecological niche of H5N1 in the present study is
characterized by a mean annual temperature range of
21–26°C, lower slope angle, areas with a tendency to
pool water, a significant monthly variation in the
degree of greenness and human population density of
more than 100 persons/km2. The resultant niche
model validated 49 cases of HPAI outbreak out of the
total 52 past as well as the latest outbreaks in India,
Bangladesh, Pakistan and Nepal.
Keywords: Avian influenza virus, ecological niche,
H5N1 virus, Indian subcontinent.
TRANSMISSION of disease across a spatial and temporal
scale can be attributed to the overlapping ecological
niches of the vector, host and pathogens, and their associated ecological landscape1. It has been established that
the vector-borne diseases like chagas, malaria, dengue and
West Nile virus are transmitted due to such niche overlapping2–6. However, in the absence of a potential vector,
the widespread transmission of a disease in a geographic
domain becomes difficult to understand. The highly
pathogenic avian influenza (HPAI–H5N1) is one such
RESEARCH COMMUNICATIONS
presence in the Indian subcontinent, which would help in
taking necessary precautionary measures against any future
outbreak.
Considering the uncertainties associated with the transmission of the HPAI (H5N1), ecological niche modelling
(ENM) was used that follows an empirical approach for
niche reconstruction1. The technique was employed to
model the potential geographic distribution of many
infectious diseases earlier22–24. The ecological niche of
the ‘black box’ of HPAI (H5N1) transmission may be
defined as ‘a set of environmental conditions in which the
HPAI (H5N1) virus is capable of maintaining its number
by infecting its hosts without immigration’ (after ref. 25).
The modelling approach in the present study used genetic
algorithm as a tool to model the ecological niche of the
‘black box’ of HPAI (H5N1) transmission. Genetic algorithms are designed to solve problems following natural
selection to get an optimized solution for a particular
problem from a population of solutions. The algorithm
first starts with a set of solutions for a particular problem.
The best solutions then reproduce a new generation of
solutions. Crossovers and mutations occur in the process
of reproduction and the bad solutions are eliminated.
Genetic Algorithm for Rule-Set Prediction (GARP) is a
variant of the family of genetic algorithms, which has
gained considerable importance because of its ability to
solve problems related to species distribution using different solutions termed as rule types, i.e. atomic rule,
range rule, negated range rule and logistic regression rule
in an iterative and evolutionary fashion26. These rules are
subjected to the process of natural selection of crossovers
and mutations and the optimized rules are used in determining the species distribution. The ecological niche
modelling in the present study was carried out using a
desktop version of GARP, which was freely downloaded27
from www.nhm.ku.edu/desktopgarp.
Ninety records for the occurrence of H5N1 in West
Bengal and Bangladesh were collected through comprehensive search of news reports, field visits and from the
websites of the Ministry of Agriculture, Department of
Animal Husbandry, Dairying and Fisheries, Government
of India (http://dahd.nic.in/birdflue.htm); the Ministry of
Fisheries and Livestock, Government of the People’s
Republic of Bangladesh (http://www.mofl.gov.bd/), and
the World Organization for Animal Health (OIE)
(http://www.oie.int/downld/AVIAN%20INFLUENZA/A_
AI-Asia.htm; Figure 1). The geographic coordinates
of the locations of outbreaks were collected from
www.fallingrain.com, as well as from field visits. These
90 records were used for the development of the model.
For independent validation of the model, 52 additional
independent outbreak records in different parts of the
subcontinent, i.e. India, Pakistan, Bangladesh and Nepal
were downloaded from the website of the World Organization for Animal Health. These records were not used as
model training or testing records.
74
Eighteen environmental parameters pertaining to the
Indian subcontinent were used for modelling. These are
raster data on altitude and mean annual temperature with
a spatial resolution of 2.5 arc min (http://www.worldclim.
org)28; Normalized Difference Vegetation Index (NDVI)
layers for January to December derived from AVHRR;
slope, aspect, compound topographic index downloaded
from the USGS Hydro-1K dataset29, and raster data on average global human population per square kilometre with
a resolution of 2.5 arc min (http://sedac.ciesin.columbia.
edu/gpw). All the environmental layers were resampled
to a spatial resolution of 0.04° using the spatial analyst
module of ArcView 3.3. Coarsening of the resolution of
environmental layers was done to match with the approximate spatial resolution of the occurrence data.
Prior to predicting the areas having potential for the
presence of H5N1, the fitness of the ecological niche
model was evaluated following a series of calibration
tests, wherein different subsets of the entire area of
occurrence were subjected to reciprocal predictivity
tests19. To enable this, the occurrence records were divided into quadrants above and below the median latitude
and median longitude. The occurrence records were then
divided into subsets of north and south of the median latitude, west and east of the median longitude and on- and
off-diagonal.
For building different subset models of the entire area
of occurrence, an algorithm threshold of 0.01 was selected with 1000 iterations as an upper limit for each run.
Then 50% of the total occurrence points were used to
generate the rule and the remaining 50% was used for
intrinsic testing for model optimization and refinement.
Figure 1. Locations of the 90 HPAI outbreaks in India and Bangladesh used for model calibration and generation of final rules.
CURRENT SCIENCE, VOL. 97, NO. 1, 10 JULY 2009
RESEARCH COMMUNICATIONS
Table 1.
Results of model calibration test performed at all threshold levels (1–10)
Sample size
Per cent of test area
predicted present
at all threshold levels (1–10)
Cumulative binomial
probability
Train
Test
No. of test points
correctly predicted
North-south
North predicts south
South predicts north
33
37
37
33
35
30
52.36
53.90
<0.001
<0.001
West-east
West predicts east
East predicts west
46
42
42
46
38
42
44.88
57.76
<0.001
<0.001
Diagonals
On predicts off
Off predicts on
25
23
23
25
10
25
25.91
80.06
0.66
<0.001
Subset test
Due to the stochastic nature of GARP in giving different
outputs at different runs, the model was set to perform
100 tasks. For selecting the 10 best subset models, the inbuilt best subset selection radio button in Desktop GARP
was checked with threshold level of 0% extrinsic hard
omission and 50% commission. All the 10 best subset
models were summed up using the calcgridsinview.ave
script in ArcView 3.3, to create a composite distributional representation of the ecological niche of the ‘black
box’ of HPAI (H5N1) transmission. Independent validation of the model was done by overlaying the occurrence
points with the model projections. To test the predictive
accuracy of the subset models, binomial probability test
was performed30.
After model calibration, the final predictions were
made by projecting the rules to the Indian subcontinent
covering India, Bangladesh, Nepal and some parts of
Pakistan. Overall, 90 occurrence records were used in the
final analysis, of which 45 were used for model building
and the remaining 45 were used for model validation. As
the final model projections were made in a much broader
unsampled area, GARP was set to perform 500 tasks with
a threshold level of 0% extrinsic hard omission and 50%
commission.
To assess the predictive ability of the ecological niche
model in a broader geographical range, the final countryscale projection was subjected to independent validation
test. This was done by overlaying 52 independent outbreak locations on the final model projection.
To visualize the ecological niche of HPAI (H5N1) in a
two-dimensional environmental space, the grid of the
composite ecological niche model was combined with
the original environmental data layers using the combine
grids option in ArcVew 3.3. The combined grids were
then exported in ASCII format. A subset of the combined
grids was then imported into MS EXCEL for visualization. The entire modelling exercise was performed on a
DELL Precision 690 WORKSTATION (X64 Edition).
The results of the model calibration tests performed at
threshold levels of 1–10 are summarized in Table 1. The
CURRENT SCIENCE, VOL. 97, NO. 1, 10 JULY 2009
northern occurrence points predicted 52.4% area of the
south to have potential for presence and correctly predicted 94.0% of the test points. The southern occurrences
predicted 53.9% area of the northern side to have potential for presence and correctly predicted 91.0% of the test
points.
Similarly, the HPAI (H5N1) occurrence points in the
west predicted 44.9% area of the eastern side to be suitable
for presence and correctly predicted 90.0% of the test
points. The eastern occurrence points predicted 57.8%
area of the west to be suitable and independently predicted 91.0% of the test points.
In the diagonal tests, the on-diagonal predicted only
25.9% area of the off-diagonal to be suitable and correctly predicted only 43.0% of the test points. However,
the off-diagonal predicted 80.0% of the on-diagonal area
to be suitable and correctly predicted all the test points.
The results of the cumulative binomial probability tests
for model calibration revealed that five subsets out of six,
correctly predicted the independent test points as evident
from significant level of binomial probability (P < 0.001).
However, the prediction through on-diagonal predicts offdiagonal subset did not yield a significant level of probability (Table 1).
The ecological niche model could independently predict 49 out of 52 outbreaks in the Indian subcontinent
between the years 2006 and 2009 (Table 2). All the outbreaks in Bangladesh, Nepal and the northwestern frontier
of Pakistan were successfully predicted. For India, the
model could predict 19 out of 22 outbreaks (Figure 2).
The three outbreaks which remained beyond the prediction of the model were located one each in Maharashtra,
Sikkim and Assam.
The country-scale prediction for the potential areas of
transmission of HPAI (H5N1) in India, Bangladesh,
Nepal and major portion of Pakistan revealed broad dispersion of the pathogen (Figure 2). The prediction shows
that the Gangetic region and West Bengal have the highest probability of HPAI transmission, as evident from the
higher model agreements. Besides, substantial areas in
75
RESEARCH COMMUNICATIONS
Table 2.
Validation of the ecological niche model using independent test points across spatial and temporal scale
Outbreak records during different years used for model validation
Country
Outbreaks correctly predicted
2006
2007
2008
2009
Total
2006
2007
2008
2009
Total
India
Bangladesh
Pakistan
Nepal
2
–
10
–
1
–
8
–
16
7
–
–
3
4
–
1
22
11
18
1
1
–
10
–
1
–
8
–
15
7
–
–
2
4
–
1
19
11
18
1
Total
12
9
23
8
52
11
9
22
7
49
Figure 2. Geographic projection of the final rules generated by the
GARP model in India, Bangladesh, Nepal and some parts of Pakistan.
Grey colour depicts the areas of agreement of (1–5) models and red
colour depicts agreement of (6–10) models. White colour depicts the
area where all 10 best subsets predict absence. Symbols in the map
denote 52 locations of HPAI outbreak during different years. These
locations were used as independent test points for final validation of
the model.
the northern and central India also have necessary elements for transmission of HPAI. The model also shows
some areas in Andhra Pradesh, Maharashtra and Gujarat
that may be suitable for HPAI transmission.
In northeastern India, some parts of Assam, Tripura
and Manipur possess suitable elements for HPAI transmission. The model predicted the waterlogged areas in
Assam, Tripura and Manipur to be suitable for transmission of HPAI virus. The mountainous areas of northeastern India indicate absence of these elements.
For Bangladesh, the model predicts a broad distributional range of the virus that spans across the country,
indicating the extensive presence of the elements necessary for its transmission. The confirmed reports on the
outbreak of avian influenza from 47 out of 64 districts in
76
Bangladesh (http://www.mofl.gov.bd/) corroborate the
model predictions.
The exploratory visualizations depict the ecological
characteristics of the modelled ‘black box’ of the H5N1
transmission in the training landscape of Indo-Bangladesh
region (Figures 3 and 4). The visualizations depict that
the niche of H5N1 is restricted to areas having lower
slope angle with a tendency to pool water, higher annual
variation in the degree of greenness and a mean annual
temperature range of 21–26°C. The combined observations
on slope angle, topographic index and monthly variation
in greenness index suggested specific land-use types, e.g.
stagnant water bodies, water courses, and wetlands,
including paddy fields, that were predicted as areas of
occurrence of the pathogen. Conforming to the model
prediction, the training landscapes in West Bengal and
Bangladesh have extensive areas under water courses,
stagnant water bodies, wetlands and agricultural lands
which constitute more than 50% of the total geographical
area31,32. The predicted areas of occurrence were also
characterized by seasonal fluctuations in the greenness
index due to variations in cultivation practice and cropping pattern (Figure 4). Gilbert et al.20,21 observed that
agricultural land use and rice crop intensity strongly drive
duck husbandry, and the abundance of free-grazing ducks
in intensive rice crop areas could be a potential factor for
transmission of H5N1. The model training landscape also
has abundant wetlands and water bodies such as shallow
ponds, which are seasonal and get dried up during winter
and summer months. These habitats favour aggregation of
wild fowl and domestic ducks for foraging on benthic
fauna, facilitating the transmission of HPAI virus.
Animal husbandry and livestock production contribute
substantially to the household income in West Bengal and
Bangladesh, where farm and backyard poultry are important components of the livestock production33. Unlike the
farm poultry, backyard poultry are free-ranging and are
generally reared under a scavenging system with feed
available from household wastes, homestead pickings,
crop residues in agricultural fields, wetlands and water
bodies. In the process, they come in close contact with
the wild migratory birds exposing them to the H5N1
transmission. Human population density could be another
surrogate element for the active local transmission of
CURRENT SCIENCE, VOL. 97, NO. 1, 10 JULY 2009
RESEARCH COMMUNICATIONS
Figure 3. Exploratory visualization of H5N1 niche in two-dimensional environmental space. Grey dots indicate the available environmental conditions and black dots indicate the modelled ecological niche of the virus.
Figure 4. Comparison of mean monthly NDVI values of 100 randomly selected pixels in the Indo-Bangladesh region. Solid line depicts the variation in mean monthly NDVI values of (a) predicted absence (annual variance = 11.80) and (b) predicted presence (annual variance = 50.07). Vertical lines in both the figures represent the range of NDVI values around the mean.
HPAI virus in the model training landscape. Confirming
this proposition, the exploratory visualization revealed
that human population with density >100 persons/km2 is
susceptible to H5N1 transmission (Figure 3). Higher human population density enhances the chance of greater
CURRENT SCIENCE, VOL. 97, NO. 1, 10 JULY 2009
trade interactions, increasing the probability of HPAI
transmission through trading of infected poultry34.
The present study shows that the ecological niche
model generated by GARP performed significantly better
than random expectations, and was successful in captur77
RESEARCH COMMUNICATIONS
ing the coarse-scale ecological conditions associated with
HPAI transmission in the Indo-Bangladesh region. The
land-use pattern, occurrence of seasonal wetlands, practices of backyard poultry and animal husbandry, and
density of human population were identified as risk factors for the widespread transmission of H5N1 virus. The
projection of the final rules generated from the training
landscape in the Indo-Bangladesh region could successfully predict the outbreaks in various parts of India, Nepal
and Pakistan from a spatial perspective. However, further
studies are required for predicting the time-specific outbreaks. Nevertheless, the present model could be used as
a guide map for possible future outbreaks in other parts
of India.
1. Peterson, A. T., Ecological niche modeling and understanding the
geography of disease transmission. Vet. Ital., 2007, 43, 393–400.
2. Peterson, A. T., Sánchez-Cordero, V., Ben Beard, C. and Ra,
J. M., Ecologic niche modeling and potential reservoirs for Chagas
disease, Mexico. Emerg. Infect. Dis., 2002, 8, 662–667.
3. Peterson, A. T., Martínez-Campos, C., Nakazawa, Y. and
Martínez-Meyer, E., Time-specific ecological niche modeling predicts spatial dynamics of vector insects and human dengue cases.
Trans. R. Soc. Trop. Med. Hyg., 2005, 99, 647–655.
4. Moffett, A., Shackelford, N. and Sarkar, S., Malaria in Africa:
Vector species’ niche models and relative risk maps. PLoS ONE,
2007, 9, 1–14.
5. Peterson, A. T., Robbins, A., Restifo, R., Howell, J. and Nasci, R.,
Predictable ecology and geography of West Nile virus transmission in the central United States. J. Vector Ecol., 2008, 33, 342–
352.
6. Wimberly, M. C., Hildreth, M. B., Boyte, S. P., Lindquist, E. and
Kightlinger, L., Ecological niche of the 2003 West Nile virus epidemic in the northern great plains of the United States. PLoS
ONE, 2008, 3, 1–7.
7. Sawabe, K. et al., Detection and isolation of highly pathogenic
H5N1 avian influenza A viruses from blow flies collected in the
vicinity of an infected poultry farm in Kyoto, Japan, 2004. Am. J.
Trop. Med. Hyg., 2006, 75, 327–332.
8. Barbazan, P. et al., Detection of H5N1 avian influenza virus from
mosquitoes collected in an infected poultry farm in Thailand.
Vector-Borne Zoonotic Dis., 2008, 8, 105–109.
9. Raynor, M. S. and Cliff, A. D., The geographical spread of avian
influenza A (H5N1): panzootic transmission (December 2003–
May 2006), pandemic potential, and implications. Ann. Assoc. Am.
Geogr., 2008, 98, 553–582.
10. Hien, T. T. et al., Avian influenza A (H5N1) in 10 patients in
Vietnam. New Engl. J. Med., 2004, 350, 1179–1188.
11. Ungchusak, K. et al., Probable person-to-person transmission of
avian influenza A (H5N1). New Engl. J. Med., 2005, 352, 333–
340.
12. Li, K. S. et al., Genesis of a highly pathogenic and potentially
pandemic H5N1 influenza virus in eastern Asia. Nature, 2004,
430, 209–213.
13. World Health Organization, Responding to the avian influenza
pandemic threat – Recommended strategic actions, WHO, Geneva,
2005.
14. Nguyen, T. D., Nguyen, T. V., Vijaykrishna, D., Webster, R. G.,
Guan, Y., Peiris, J. S. M. and Smith, J. D., Multiple sublineages of
influenza A (H5N1), Vietnam, 2005–2007. Emerg. Infect. Dis.,
2008, 14, 632–636.
15. Causey, D. and Edwards, S. V., Ecology of avian influenza in
birds. J. Infect. Dis., 2008, 197, 29–33.
78
16. http://www.oie.int/downld/AVIAN%20INFLUENZA/A_AI-Asia.htm
17. Kilpatrick, A. M., Chmura, A. A., Gibbons, D. W., Fleischer,
R. C., Marra, P. P. and Daszak, P., Predicting the global spread of
H5N1 avian influenza. Proc. Natl. Acad. Sci. USA, 2006, 103,
19368–19373.
18. Gilbert, M. et al., Free-grazing ducks and highly pathogenic avian
influenza, Thailand. Emerg. Infect. Dis., 2006, 12, 227–234.
19. Williams, R. A. J., Fasina, F. O. and Peterson, A. T., Predictable
ecology and geography of avian influenza (H5N1) transmission in
Nigeria and West Africa. Trans. R. Soc. Trop. Med. Hyg., 2008,
102, 471–479.
20. Gilbert, M., Xiao, X., Chaitaweesub, P., Kalpravidh, W., Premashthira, S., Boles, S. and Slingenbergh, J., Avian influenza,
domestic ducks and rice agriculture in Thailand. Agric. Ecosyst.
Environ., 2007, 119, 409–415.
21. Gilbert, M. et al., Mapping H5N1 highly pathogenic avian influenza risk in Southeast Asia. Proc. Natl. Acad. Sci. USA, 2008,
105, 4769–4774.
22. Peterson, A. T., Lash, R. R., Carroll, D. S. and Johnson, K. M.,
Geographic potential for outbreaks of Marburg hemorrhagic fever.
Am. J. Trop. Med. Hyg., 2006, 75, 9–15.
23. Peterson, A. T., Bauer, J. T. and Mills, J. N., Ecological and geographic distribution of filovirus disease. Emerg. Infect. Dis., 2004,
10, 40–47.
24. Levine, R. S., Peterson, A. T., Yorita, K. L., Carroll, D., Damon,
I. K. and Reynolds, M. G., Ecological niche and geographic distribution of human monkeypox in Africa. PLoS ONE, 2007, 1, 1–7.
25. Grinnell, J., Field tests of theories concerning distributional control. Am. Nat., 1917, 51, 115–128.
26. Stockwell, D. R. B. and Peters, D., The GARP modelling system:
problems and solutions to automated spatial prediction. Int. J.
Geogr. Inf. Syst., 1999, 13, 143–158.
27. Scachetti-Pareira, R., Desktop GARP v.1.1.4. The University of
Kansas Biodiversity Research Group, Kansas, USA, 2001.
28. Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G. and Jarvis, A., The WorldClim interpolated global terrestrial surfaces.
Version 1.3, 2004.
29. United States Geological Survey, HYDRO1k elevation derivative
database. United States Geological Survey, Washington DC, 2001.
30. Anderson, R. P., Lew, D. and Peterson, A. T., Evaluating predictive models of species’ distributions: criteria for selecting optimal
models. Ecol. Model., 2003, 162, 211–232.
31. www.wbgov.com/e-gov/admin/newgovtpublications/upload/Pubstatus_on_land.pdf
32. ASB Land use. In Banglapedia: National Encyclopedia of Bangladesh, Asiatic Society of Bangladesh, Dhaka, 2003, vol. 6, pp.
235–239.
33. FAO, Poultry sector country review. FAO Animal Production and
Health Division, Food and Agriculture Organization of the United
Nations, Rome, 2008.
34. Clerc, M. G., Lebarbenchon, C. and Thomas, F., Recent expansion
of highly pathogenic avian influenza H5N1: a critical review. Ibis,
2007, 149, 202–214.
ACKNOWLEDGEMENTS. We are grateful to the organizers of the
training programme on Ecological Niche Modelling at Bangalore during 13–17 December 2006, where the first author was a participant. We
thank Dr Mohammed Irfan-Ullah for valuable suggestions while preparing exploratory visualizations of the ecological niche models. We
also thank the anonymous referee for useful comments. Financial assistance to the first author in the form of Research Associateship by the
UPE (Biosciences) Programme under the School of Life Sciences,
North-Eastern Hill University, Shillong, is acknowledged.
Received 3 March 2008; revised accepted 22 April 2009
CURRENT SCIENCE, VOL. 97, NO. 1, 10 JULY 2009