- Open Access
Evaluating local vegetation cover as a risk factor for malaria transmission: a new analytical approach using ImageJ
© Ricotta et al.; licensee BioMed Central Ltd. 2014
- Received: 8 October 2013
- Accepted: 7 March 2014
- Published: 13 March 2014
In places where malaria transmission is unstable or is transmitted under hypoendemic conditions, there are periods where limited foci of cases still occur and people become infected. These residual “hot spots” are likely reservoirs of the parasite population and so are fundamental to the seasonal spread and decline of malaria. It is, therefore, important to understand the ecological conditions that permit vector mosquitoes to survive and forage in these specific areas. Features such as local waterways and vegetation, as well as local ecology, particularly nocturnal temperature, humidity, and vegetative sustainability, are important for modeling local mosquito behavior. Vegetation around a homestead likely provides refuge for outdoor resting of these insects and may be a risk factor for malaria transmission. Analysis of this vegetation can be done using satellite information and mapping programs, such as Google Earth, but manual quantification is difficult and can be tedious and subjective. A more objective method is required.
Vegetation cover in the environment is reasonably static, particularly in and around homesteads. In order to evaluate and enumerate such information, ImageJ, an image processing software, was used to analyse Google Earth satellite imagery. The number of plants, total amount of vegetation around a homestead and its percentage of the total area were calculated and related to homesteads where cases of malaria were recorded.
Preliminary results were obtained from a series of field trials carried out in South East Zambia in the Choma and Namwala districts from a base at the Macha District Hospital.
This technique is objective, clear and simple to manipulate and has potential application to determine the role that vegetation proximal to houses may play in affecting mosquito behaviour, foraging and subsequent malaria incidence.
- Vector behavior
- Satellite imagery
Mosquito survival and foraging are driving factors in the overall epidemiology of malaria. At times when conditions are harsh and dry, survival of the vectors is limited and local foci of transmission persist because of shrub vegetation that provides mosquitoes with resting sites and refuges . Where malaria transmission is unstable or seasonally restricted by dry, hot seasonal weather, predicting or defining conditions that will sustain pockets of vectors will be of considerable epidemiological importance . On a local scale, foci are crucial for the detection and elimination of small, isolated transmission reservoirs. These “hot spots” are particularly important during periods of low transmission, as they are a primary source of infection for resurgent mosquito populations [3, 4]. As the existence and stability of these foci are a function of climate conditions that enable the mosquitoes to forage, features of the landscape and ecology in a region are likely to play a major role in malaria transmission dynamics by affecting mosquito development, survival, and foraging behavior . Such features include land surface temperature, water bodies, altitude, rainfall, and vegetation . It is important to keep in mind how mosquito behaviour is affected by environmental conditions when evaluating the biological significance of these features and the role they play in the risk of malaria transmission.
However, while these ecological factors are important to know, they are not fine-scale enough to determine the location of site-specific transmission. Instead, by looking at local waterways and vegetation in addition to considering local ecology (specifically nocturnal temperature, humidity, and vegetative sustainability), local mosquito behavior can be predicted [7, 8]. For example, as shrubs and smaller plants can create or alter the microclimate in which mosquitoes can rest outside of the home, vegetation around a homestead is likely to be an important determinant of malaria transmission . This is particularly so during the dry season as resting sites can be sparse and the amount of available vegetation could affect mosquito survivorship and dispersal away from the home, and therefore affect malaria risk in a particular area. Such fine-tuned environmental data can be obtained via satellites, and using aerially-obtained environmental data is an excellent way to evaluate the spatial element of malaria transmission. The choice of satellite data depends on the study design and questions, as well as resource availability.
Current methods for assessing vegetation cover using remote sensing require complex calculations to make the data interpretable for use in epidemiological analyses . The primary method for evaluation of vegetation cover is the normalized difference vegetation index (NDVI) . This is the ratio of near infrared and red spectral bands obtained via SPOT imagery, which can quantify green leaf vegetation that can be mapped and used in statistical models . Here, a method for evaluating vegetation without requiring complex satellite data and intricate calculations was devised. However observing and in some way measuring the extent of vegetation cover from satellite maps is somewhat subjective and difficult to interpret, it can be done more objectively and consistently using a computer program.
The software used for this analysis was ImageJ, a public domain Java image processing program created by the National Institutes of Health . It has image processing capabilities and is used for a wide variety of applications. By measuring pixel values and making simple calculations about pixel size, number, area covered, etc., ImageJ can be used to measure distance between spots, create plots for graphic visualization, and transform images through automated macros, thus reducing measurement bias. In this study, ImageJ was used to analyse freely-available Google Earth images of malaria-endemic locations to identify potential risk factors associated with vegetation cover.
ImageJ is an image analysis software program, which has found wide use in a variety of fields. The program is Java-based and will run on all computer platforms, is freely available and easy to use, and can be customized to support a variety of new functions. Further, it can easily handle large data sets, extending its utility to applications which require uniform analysis over many repetitions. Here, these features were utilized to analyse satellite imagery of homesteads for surrounding vegetation.
ImageJ protocol development and image analysis
Vegetation in Zambia differs by season but consists primarily of grasses, short shrubbery, and patches of trees. In addition, much of the area surrounding homesteads are croplands. Of the different types of vegetation present, short grasses are least visible on satellite data; however, as dry grass provides little shelter for vector mosquitoes, these areas were not included in this analysis. Of the remaining vegetation, short trees and bushes provide more suitable shelter for post-feeding resting mosquitoes and these also produce the largest signal difference in satellite imagery.
Threshold values were analysed in five unit increments to determine the optimum threshold values for area calculation. Representative images were adjusted similarly for brightness and contrast, and thresholds were set (0–60, 0–65, 0–70, etc.) and analysed, as above. The results were compared across threshold values and an optimum threshold value was chosen based on a combination of these results and a visual selection.
All relations were analysed by logistic regression using R version 3.0.0 (2013-04-03) -- “Masked Marvel” (The R Foundation for Statistical Computing) to estimate odds ratios, with malaria case status as the outcome of interest (case = 1, control = 0). 2.4 m/pixel and 10 m/pixel resolution images were pooled to increase the power of the study. Each analysis was repeated for images at radii of 50, 100, and 500 m. Variables were count (number of plants), average size of plants, fraction of the area surrounding the homestead, and total area covered by plants. Quartiles of each variable were also analysed.
The work was carried out under the approval of the Johns Hopkins Bloomberg School of Public Health IRB no: 00002290. All data was de-identified by MIAM staff before transmission to Johns Hopkins. No patient data was retained other than the GPS coordinates of the household, which was approved by this IRB, and therefore no written consent was obtained.
Of the 122 homesteads with available satellite imagery, 79 had usable images (36 case homesteads and 43 control homesteads). Only households that had usable imagery at all three buffer distances were included in the analysis. Reasons for exclusion include poor quality imagery, images combining two different resolutions, and households covered by clouds at the time the image was captured. These are common concerns when evaluating any satellite imagery, especially those from Google Earth, as satellites only pass over locations at certain times and data are not routinely updated. Of note, these images were from June 2007, while homesteads were from 2008–2012. However, as the landscape has not significantly changed in this area (and in fact all homesteads were able to be located on the images), the time discrepancy is not believed to be an issue. This would have to be assessed in each location on a study-by-study basis.
Odds ratios of variables when included in regression model as continuous
OR (95% CI)
0.99 (0.97 – 1.02)
0.99 (0.99 – 1.01)
1.01 (0.96 – 1.06)
0.98 (0.93 – 1.04)
0.97 (0.91 – 1.03)
1.00 (0.99 – 1.00)
0.99 (0.99 – 1.00)
0.99 (0.99 – 1.00)
1.00 (0.99 – 1.01)
1.00 (0.99 – 1.01)
0.99 (0.99 – 1.00)
Odds ratios of variables when included in regression models as categorical quartiles
OR (95% CI)
1.34 (0.34 – 4.56)
2.55 (0.72 – 9.73)
1.86 (0.53 – 6.84)
3.50 (0.98 – 13.75)
1.91 (0.53 – 7.31)
2.10 (0.57 – 8.16)
1.83 (0.53 – 6.64)
1.35 (0.38 – 4.91)
1.00 (0.28 – 3.59)
0.97 (0.27 – 3.43)
1.20 (0.34 – 4.24)
1.33 (0.39 – 4.65)
0.92 (0.26 – 3.17)
1.22 (0.34 – 4.46)
1.00 (0.28 – 3.51)
0.75 (0.21 – 2.57)
1.25 (0.35 – 4.57)
0.54 (0.15 – 1.90)
2.85 (0.80 – 11.03)
(614,1.19e + 03]
2.10 (0.57 – 8.16)
(1.19e + 03,2.96e + 03]
2.33 (0.65 – 8.95)
(2.12e + 03,3.57e + 03]
1.00 (0.28 – 3.51)
(3.57e + 03,5.15e + 03]
1.10 (0.31 – 3.93)
(5.15e + 03,1.15e + 04]
1.00 (0.28 – 3.51)
(5.97e + 04,9.36e + 04]
1.00 (0.28 – 3.51)
(9.36e + 04,1.28e + 05]
1.34 (0.38 – 4.89)
(1.28e + 05,2.79e + 05]
0.81 (0.23 – 2.87)
0.38 (0.10 – 1.28)
0.41 (0.10 – 1.49)
0.92 (0.26 – 3.17)
0.36 (0.10 – 4.26)
0.28 (0.07 – 1.02)
0.62 (0.17 – 2.12)
0.82 (0.23 – 5.84)
(881,1.03e + 03]
0.73 (0.20 – 2.57)
(1.03e + 03,1.42e + 03]
0.82 (0.23 – 2.84)
ImageJ provides a novel, free, and easy way to evaluate vegetation around homes for use in epidemiological analysis. This method eliminates the need for complex calculations to quantify vegetation (i.e. NDVI), as the output from the program provides raw numbers that can be imported directly into statistical software for analysis. Additionally, it standardizes how these images are analysed, which removes measurement error introduced when analysing by hand.
The methods developed here use only some of the many capabilities of ImageJ. Further experimentation on a wider range of landscapes and image types will enable implementation of even more practical uses of this freely available software. In this analysis, images were required to be in 8 bit black and white, potentially causing loss of important features. Recent methods have been developed using ImageJ for cell staining to recognize and select areas by colour . This could be beneficial when analysing satellite data, to ensure unwanted features such as bodies of water, livestock, or houses, are not included in the analysis. Specifically important for malaria epidemiology would be a way to select areas based on size; as grasses and large trees do not typically provide suitable shelter for mosquitoes, excluding them from the analysis is important to ensure accuracy when modeling. While this was not evaluated in depth in this paper, ImageJ does have the capacity to select areas based on particle size.
There are many factors to be considered when studying malaria transmission, particularly in hypoendemic regions. The introduction of satellite data to epidemiological analysis has played an important role in advancing knowledge of how the environment affects malaria transmission, but these analyses are still complex and expensive, limiting the number of people who can put it to use. Using ImageJ is an important new tool to enhance satellite imagery analysis, as it is versatile, inexpensive, and provides quantitative output for use in epidemiological modelling.
Grateful thanks are due to Dr. Sungano Mharakurwa and Aniset Kamanga at the Johns Hopkins Malaria Research Institute (JHMRI), Macha, Zambia, as well as the staff at the rural health centers who provided much insight and assistance. Additional thanks are due to Mr. Gabriel Mushinge, who provided advice on remote sensing potential, and Timothy Shields who supplied Hopkins-owned satellite images and assistance with GIS analysis.
- Bar-Zeev M, Maibach HI, Khan AA: Studies on the attraction of Aedes aegypti (Diptera: Culicidae) to man. J Med Entomol. 1977, 14: 113-120.View ArticlePubMedGoogle Scholar
- Shiff CJ, Stoyanov C, Choobwe C, Kamanga A, Mukonka VM: Measuring malaria by passive case detection: a new perspective based on Zambian experience. Malar J. 2013, 12: 120-10.1186/1475-2875-12-120.PubMed CentralView ArticlePubMedGoogle Scholar
- Stresman G, Kobayashi T, Kamanga A, Thuma PE, Mharakurwa S, Moss WJ, Shiff C: Malaria research challenges in low prevalence settings. Malar J. 2012, 11: 353-10.1186/1475-2875-11-353. doi: 10.1186/1475-2875-11-353PubMed CentralView ArticlePubMedGoogle Scholar
- Bousema T, Drakeley C, Gesase S, Hashim R, Magesa S, Mosha F, Otieno S, Carneiro I, Cox J, Msuya E, Kleinschmidt I, Maxwell C, Greenwood B, Riley E, Sauerwein R, Chandramohan D, Gosling R: Identification of hot spots of malaria transmission for targeted malaria control. J Infect Dis. 2010, 201: 1764-1774. 10.1086/652456. doi: 10.1086/652456View ArticlePubMedGoogle Scholar
- Afrane YA, Zhou G, Lawson BW, Githeko AK, Yan G: Life-table analysis of Anopheles arabiensis in western Kenya highlands: effects of land covers on larval and adult survivorship. Am J Trop Med Hyg. 2007, 77: 660-666. Available: http://0-www.ncbi.nlm.nih.gov.brum.beds.ac.uk/pubmed/17978067. Accessed 30 January 2013PubMedGoogle Scholar
- Riedel N, Vounatsou P, Miller JM, Gosoniu L, Chizema-Kawesha E, Mukonka V, Steketee RW: Geographical patterns and predictors of malaria risk in Zambia: Bayesian geostatistical modelling of the 2006 Zambia national malaria indicator survey (ZMIS). Malar J. 2010, 9: 37-10.1186/1475-2875-9-37. doi:10.1186/1475-2875-9-37PubMed CentralView ArticlePubMedGoogle Scholar
- Kulkarni MA, Desrochers RE, Kerr JT: High resolution niche models of malaria vectors in northern Tanzania: a new capacity to predict malaria risk?. PLoS One. 2010, 5: 2-doi:10.1371/journal.pone.0009396View ArticleGoogle Scholar
- Machault V, Vignolles C, Borchi F, Vounatsou P, Pages F, Briolant S, Lacaux J, Rogier C: The use of remotely sensed environmental data in the study of malaria. Geospat Health. 2011, 5: 151-168. Available: http://0-www.ncbi.nlm.nih.gov.brum.beds.ac.uk/pubmed/21590665. Acessed 5 July 2013View ArticlePubMedGoogle Scholar
- Afrane YA, Zhou G, Lawson BW, Githeko AK, Yan G: Effects of microclimatic changes caused by deforestation on the survivorship and reproductive fitness of Anopheles gambiae in western Kenya highlands. Am J Trop Med Hyg. 2006, 74: 772-778.PubMedGoogle Scholar
- Machault V, Vignolles C, Pagès F, Gadiaga L, Gaye A, Sokhna C, Trape J, Lacaux J, Rogier C: Spatial heterogeneity and temporal evolution of malaria transmission risk in Dakar, Senegal, according to remotely sensed environmental data. Malar J. 2010, 9: 252-10.1186/1475-2875-9-252. doi:10.1186/1475-2875-9-252PubMed CentralView ArticlePubMedGoogle Scholar
- Schneider CA, Rasband WS, Eliceiri KW: NIH Image to ImageJ: 25 years of image analysis. Nat Methods. 2008, 9: 671-675. doi:10.1038/nmeth.2089View ArticleGoogle Scholar
- Stresman GH, Kamanga A, Moono P, Hamapumbu H, Mharakurwa S, Kobayashi T, Moss WJ, Shiff C: A method of active case detection to target reservoirs of asymptomatic malaria and gametocyte carriers in a rural area in Southern Province, Zambia. Malar J. 2010, 9: 265-doi:10.1186/1475-2875-9-265PubMed CentralPubMedGoogle Scholar
- Midega JT, Mbogo CM, Mwnambi H, Wilson MD, Ojwang G, Mwangangi JM, Nzovu JG, Githure JI, Yan G, Beier JC: Estimating dispersal and survival of Anopheles gambiae and Anopheles funestus along the Kenyan coast by using mark-release-recapture methods. J Med Entomol. 2009, 44: 923-929.View ArticleGoogle Scholar
- Lauber S: ImageJ - threshold analysis. OpenWetWare 2012. Available: http://openwetware.org/index.php?title=Sean_Lauber:ImageJ_-_Threshold_Analysis&oldid=595746. Accessed 15 June 2013
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.