Acta Tropica 108 (2008) 234–241 Contents lists available at ScienceDirect Acta Tropica journal homepage: www.elsevier.com/locate/actatropica Schistosomiasis risk estimation in Minas Gerais State, Brazil, using environmental data and GIS techniques Ricardo J.P.S. Guimarães a,b , Corina C. Freitas c , Luciano V. Dutra c , Ana C.M. Moura d , Ronaldo S. Amaral e , Sandra C. Drummond f , Ronaldo G.C. Scholte a,b , Omar S. Carvalho a,∗ a Centro de Pesquisas René Rachou/Fiocruz-MG, Av. Augusto de Lima, 1715 Barro Preto, CEP 30190-002 Belo Horizonte, MG, Brazil Programa de Pós-Graduação da Santa Casa de Misericórdia de Belo Horizonte/MG, Avenida Francisco Sales, 1111 Santa Efigênia, CEP 30150-221 Belo Horizonte, MG, Brazil c Instituto Nacional de Pesquisas Espaciais/INPE, Av dos Astronautas, 1.758 Jd. Granja, CEP 12227-010 São José dos Campos, SP, Brazil d Universidade Federal de Minas Gerais/UFMG, Av. Antônio Carlos, 6627 Pampulha, CEP 31270-901 Belo Horizonte, MG, Brazil e Secretaria de Vigilância em Saúde/MS, Setor Hoteleiro Sul, Q-06, Conjunto A, Bl.C, sala 711, CEP 70 322-915 Brasília, DF, Brazil f Secretaria de Estado de Saúde de Minas Gerais, Rua Rio Grande do Norte, 613, 4 andar, Santa Efigênia, CEP 30130-130 Belo Horizonte, MG, Brazil b a r t i c l e i n f o Article history: Available online 22 July 2008 Keywords: GIS Schistosomiasis Multiple linear regression Prevalence Mixing model MODIS SRTM a b s t r a c t The influence of climate and environmental variables to the distribution of schistosomiasis has been assessed in several previous studies. Also Geographical Information System (GIS), is a tool that has been recently tested for better understanding the spatial disease distribution. The objective of this paper is to further develop the GIS technology for modeling and control of schistosomiasis using meteorological and social variables and introducing new potential environmental-related variables, particularly those produced by recently launched orbital sensors like the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Shuttle Radar Topography Mission (SRTM). Three different scenarios have been analyzed, and despite of not quite large determination factor, the standard deviation of risk estimates was considered adequate for public health needs. The main variables selected as important for modeling purposes was topographic elevation, summer minimum temperature, the NDVI vegetation index, and the social index HDI91 . © 2008 Published by Elsevier B.V. 1. Introduction 1.1. Schistosomiasis mansoni Schistosomiasis mansoni is a major parasitic disease in humans and it is known to be endemic in approximately 54 countries in Americas and Africa (WHO, 1985; Chitsulo et al., 2000). The etiological agent is the trematode Schistosoma mansoni which causes varied symptoms ranging from acute to chronic forms but predominantly with intestinal manifestations. Severe forms of the disease may occur with fibro-obstruction in the liver and portal hypertension, splenomegaly and impairment of the central nervous system. Treatment for schistosomiasis is simple due to the ready availability of drugs that can be administered in a single oral dose (Katz et ∗ Corresponding author at: Centro de Pesquisas René Rachou/Fiocruz-MG, Laboratório de Helmintologia e Malacologia Médica, Av. Augusto de Lima, 1715 Barro Preto, CEP 30190-002 Belo Horizonte, MG, Brazil. Tel.: +55 31 33497745; fax: +55 31 32953115. E-mail address: omar@cpqrr.fiocruz.br (O.S. Carvalho). 0001-706X/$ – see front matter © 2008 Published by Elsevier B.V. doi:10.1016/j.actatropica.2008.07.001 al., 1989). As for prevalence, the disease has been mainly spreading from the periphery to big urban centers and other regions of the country. The disease is socially determined by behavior, spreading from the outskirts of big urban centers to other regions of the country (Graeff et al., 1999). Schistosomiasis is mostly determined by lack of basic sanitation in peripheries of large urban centers, in natura sewage directly released in hydric collections and the development of irrigated agriculture related to the use of hydric resources, which has contributed to the disease spread. Schistosomiasis control comprises public measures of chemotherapic treatments, sanitation supply, drinking water supply, sewage draining, health education and use of molluscicide. The use of molluscicide for snails control has proved to damage the local ecosystem besides being costly. Out of the three S. mansoni intermediate hosts in Brazil, Biomphalaria glabrata, Biomphalaria tenagophila and Biomphalaria straminea, the B. glabrata constitutes the most important transmitter due to its wide geographic distribution, high infection rates and high capacity of schistosomiasis transmission; indeed, its occurrence has been always associated with the disease in endemic areas. Such fact was first reported in 1917 when Lutz remarked that the distribution of schistosomiasis R.J.P.S. Guimarães et al. / Acta Tropica 108 (2008) 234–241 in Brazil remained a public problem to be addressed. The author suggested the possibility of correlation with Planorbis olivaceus (=B. glabrata), which is a snail species from the north living in freshwater collections with aquatic vegetation. The presence of this mollusc species has been recorded in 16 Brazilian states, besides Distrito Federal, and 799 municipalities located between the parallels 0◦ 53 S (Quatipuru, PA) and 29◦ 51 S (Esteio, RS), the meridian 53◦ 44 S (Toledo, PR) and the coast line. B. tenagophila is epidemiologically important in the south of Brazil and has been reported in 404 municipalities as well as in 12 Brazilian states in a quadrant limited by the parallels 10◦ 12 and 33◦ 41 S, the meridian 54◦ 35 W and the coast line. B. straminea has been recorded in 1201 municipalities within 24 Brazilian states and is predominant in the states of Piaui, Ceara, Rio Grande do Norte, Paraiba, Pernambuco, Alagoas, Sergipe and Bahia, in the area between parallels 02◦ 54 and 31◦ 00 S, the meridian 44◦ 43 W and the coast line (Carvalho et al., 2008). The mollusc fauna of the genus Biomphalaria in the state of Minas Gerais is represented by seven species. These molluscs have been found in 12 mesoregions, in 286 (33.1%) municipalities, out of the 853 existing in the state: B. glabrata, B. straminea, B. tenagophila, B. peregrina, B. schrammi, B. intermedia and B. occidentalis. The species B. glabrata, recorded in 192 municipalities, is the main responsible for S. mansoni transmission. In endemic areas, high densities of this intermediate host associated with other risk factors enable for a high prevalence of schistosomiasis. The snail wide distribution accounts for the spreading characteristic of the disease to nonendemic areas of the state of Minas Gerais (Katz and Carvalho, 1983; Carvalho et al., 1987, 1988, 1989, 1994, 1997, 1998). Schistosomiasis distribution in the state of Minas Gerais is not regular, since areas of high prevalence are close to non-endemic regions, even to those areas where the transmission is null. The disease is known to be endemic in the regions named as Norte (zone of Médio São Francisco and Itacambira), Oriental and Central regions (zone of Alto Jequitinhonha, West, Alto São Francisco and Metalurgica). The highest infection rates are found in the northeast and east regions of Minas Gerais, comprising the areas of Mucuri, Rio Doce and Zona da Mata (Pellon and Teixeira, 1950; Katz et al., 1978; Carvalho et al., 1987; Lambertucci et al., 1987). 235 prevalence and risk factors distributions. The joint use of GIS and statistical techniques allows the determination of risk factors and geolocalization of risk areas leading to the optimization of the resources and to the choice of better strategies for controlling the disease (Beck et al., 1997, 2000; Bavia et al., 2001). The prediction of schistosomiasis using GIS was first attempted in the Philippines and the Caribbean by (Cross et al., 1984). The influence of climate and environmental variables to the distribution of schistosomiasis was documented by Brown (1994) and Appleton (1978). The use of GIS for the study of schistosomiasis was also done in several other countries: in Asia (Cross et al., 1996), China (Zhou et al., 2001; Seto et al., 2002; Yang et al., 2005), Ethiopia (Kristensen et al., 2001; Malone et al., 2001), Egypt (Malone et al., 1994, 1997; Abdel-Rahman et al., 2001), Uganda (Kabatereine et al., 2004), Tanzania (Brooker et al., 2001), Chad (Beasley et al., 2002; Brooker et al., 2002). In Brazil, one of the first studies trying to correlate disease distribution with environmental variables was conducted by (Bavia et al., 2001), in Bahia state. Other studies using GIS were held in Brazil, such as, in Pernambuco (Barbosa et al., 2004) and Minas Gerais (Guimarães et al., 2006). 1.3. Objectives The objective of this paper is to further develop the GIS technology for modeling and control of schistosomiasis in the state of Minas Gerais, Brazil, using meteorological and social variables and introducing new potential environmental factors, particularly those produced by recently launched orbital sensors like the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Shuttle Radar Topography Mission (SRTM). 2. Data acquisition 2.1. Schistosomiasis prevalence data The historical schistosomiasis prevalence data from 189 municipalities (dependent variable) was gathered from Brazilian Health National Foundation and from Health Secretariat of Minas Gerais State Annual Reports. The spatial distribution of these data is presented in Fig. 1a. 1.2. Utilization of the Geographical Information System (GIS) in the study of schistosomiasis 2.2. Remote sensing products Since schistosomiasis is a disease determined in space and time by risk factors, the Geographical Information System is a very powerful tool that might be used for better understanding the disease 2.2.1. Moderate Resolution Imaging Spectroradiometer (MODIS) The MODIS instrument is operating on both the Terra and Aqua spacecraft. It has a viewing swath width of 2330 km and views the Fig. 1. (a) Spatial distribution for prevalence data: no information (white), prevalence (%) 0.001–5.000 (green), 5.001–15.000 (yellow), and above 15.001 (red). (b) Sets: 142 cases for variables definition in blue and 47 cases for model validation in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.) 236 R.J.P.S. Guimarães et al. / Acta Tropica 108 (2008) 234–241 entire surface of the Earth every 1–2 days. Its detectors measure 36 spectral bands between 0.405 m and 14.385 m, and it acquires data at three spatial resolutions—250 m, 500 m, and 1000 m. Data products derived from MODIS observations describe features of the land, oceans and the atmosphere that can be used for studies of processes and trends on local to global scales (Justice et al., 1998). MODIS MOD13 product, which was used in this study, comprises the blue, red, near infrared (NIR) and the middle infrared (MIR) bands, the enhanced vegetation index (EVI) and the normalized difference vegetation index (NDVI) (Huete et al., 1999). The MOD13 product is delivered as a set of image compositions produced globally with 1 km, 500 m and 250 m of resolution, in a 16-day period. 2.2.2. Shuttle Radar Topography Mission data (SRTM) A near-global scale Digital Elevation Model (DEM) was obtained by the Shuttle Radar Topography Mission, generating the most com- plete high-resolution digital topographic database of Earth. SRTM consisted of a specially modified radar system that flew onboard the Space Shuttle Endeavour during an 11-day mission in February of 2000. The ground altitude used in this experiment is given by the value of SRTM DEM for each ground pair of coordinates. The local declivity was derived from SRTM DEM by an appropriate filtering approximation of first-order derivative. 2.2.3. Spectral Linear Mixing Model (SLMM) The SLMM is the image processing algorithm that generates the fraction images with the proportion of each component (vegetation, soil, and shade) inside the pixel which is estimate by minimizing the sum of square of the errors. The proportion values must be nonnegative, and they also must add to unity (Shimabukuro and Smith, 1991). In this case, the pixel response in any given spectral band is assumed to be a linear combination Fig. 2. Prevalence as a function of social and meteorological variables: (a) Observed prevalence and (b) estimated prevalence—no information (white), prevalence (%) 0.001–5.000 (green), 5.001–15.000 (yellow), and above 15.001 (red); (c) summer minimum temperature, 23 ◦ C (red) to 14 ◦ C (blue); (d) 1991 HDI, <0.500 (light red) to >0.701 (dark red); (e) residuals, below −10.001 (magenta), −10.000 to −5.001 (cyan), −5.000 to 5.000 (white), 5.000–10.000 (blue), and above 10.001 (red); and (f) standard error of the estimated prevalence, 0.060–0.100 (white), 0.101–0.150 (green), 0.151–0.200 (yellow), and 0.201–0.320 (red). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.) R.J.P.S. Guimarães et al. / Acta Tropica 108 (2008) 234–241 of the responses of each individual component. For any individual pixel, the linear model can be expressed by ri = n (aij xj ) + ei , k = 1, . . . , p j=1 where ri represents the pixel’s mean spectral reflectance in the ith spectral band, aij is the spectral reflectance of the jth component in the ith spectral band, xj is the proportion of the jth component within the pixel, n is the number of components, ei is the residual for the ith spectral band, and p is the number of spectral bands. The proportions xj are subjected to two constraints: xj = 1 and xj ≥ 0, for all components The least squares techniques can then be applied to estimate the components proportions xj (Shimabukuro and Smith, 1991). 237 In this work the so-called vegetation, soil, and shade fraction images were generated using the MODIS data and the estimated values for the spectral reflectance components also used as an input to the model. 2.2.4. Meteorological variables The meteorological variables consisted of total precipitation and the minimum and maximum temperature for summer and winter seasons, which were obtained from the Brazilian Weather Forecast Center (CPTEC/INPE). 2.3. Social variables The used social variables consisted of human development index (HDI), HDI-income, HDI-longevity and HDI-education for the years of 1991 and 2000 were obtained from Brazilian SNIU (National System for Urban Indicators). Fig. 3. Prevalence as a function of remote sensing variables: (a) observed prevalence and (b) estimated prevalence—no information (white), prevalence (%) 0.001–5.000 (green), 5.001–15.000 (yellow), and above 15.001 (red); (c) SRTM DEM, 2653 m (brown) to 46 m (blue); (d) winter NDVI, 1 (green) to −1 (purple); (e) residuals, below −10.001 (magenta), −10.000 to −5.001 (cyan), −5.000 to 5.000 (white), 5.000–10.000 (blue), and above 10.001 (red); and (f) standard error of the estimated prevalence, 0.055–0.100 (white), 0.101–0.150 (green), 0.151–0.200 (yellow), and 0.201–0.280 (red). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.) 238 R.J.P.S. Guimarães et al. / Acta Tropica 108 (2008) 234–241 3. Methodology In this paper a relationship between schistosomiasis prevalence and the aforementioned variables is established by using multiple regression models. Since the prevalence data is given on a municipality level, all the input variables were integrated inside the municipalities’ boundaries, using GIS systems (ArcGis, ENVI) and exported to a standard spreadsheet for the statistical analysis and modeling. The final statistical model is built using the data from the 189 municipalities, where prevalence information is available. The fitted model is then used to build the risk map for the entire state of Minas Gerais, by applying the model to the remaining municipalities. 3.1. Types of linear regression models for prevalence risk estimation Three different scenarios were analyzed for risk map estimation. The first scenario establishes a relationship between prevalence of schistosomiasis and social–meteorological variables (Guimarães et al., 2006); the second scenario estimates prevalence from variables derived from remote sensing data, and the third scenario uses all the previous selected explicative variables, through multiple linear regressions. In the first phase of model building process, the schistosomiasis prevalence data, to be used as dependent variable (Fig. 1a), were randomly divided into two sets: one with 142 cases for variables selection and model definition, and another with 47 cases for model validation. The spatial distribution of these sets can be observed in Fig. 1b. The meteorological and remote sensing data, used as independent variables, were observed in two seasons: summer (from 17 January 2002 to 01 February 2002 period) and winter (from 28 July 2002 to 12 August 2002 period). These variables, plus the social variables, were used as input variables to establish a multiple regression model for prevalence risk in each scenario. A logarithmic transformation for the dependent variable (prevalence, denoted by P̂v) was made as it improved the correlation with independent variables. Since multicollinearity effects among the independent variables were detected, a variables selection technique were used in order to choose a set of variables (or transformations of them) that better explain the dependent variable. The variables selection was done by the R2 criterion, using all possible regression procedures (Neter et al., 1996). This selection technique consists of a subset identification with few variables and a coefficient of determination R2 sufficiently closed to that when all variables are used in the model. Interaction effects were also included in the model. After performing the residual analysis on the 142 training points, which gives the mean square error (MSE), the chosen model was then validated using the 47 cases by comparing the MSE with the mean square error of prediction (MSPR) (Neter et al., 1996). If both values are about of the same order the model is considered robust, and the final model is then built using all 189 available prevalence data. 4. Results 4.1. Prevalence as a function of social and meteorological variables Fourteen quantitative independent variables were used in the statistical analysis: three meteorological variables (total precipi- tation, minimum and maximum temperature) in summer and in winter seasons, and four social variables (human development index, HDI-income, HDI-longevity and HDI-education indices) for the years of 1991 and 2000. Besides quantitative variables, two qualitative variables (binary) were also used, to represent three ecosystem types (biomes): savanna, caatinga, and forest (Guimarães et al., 2006). The chosen model, that was statistically significant at 5% level, consisted of a model with four variables: summer minimum temperature (TNs ), 1991 Human Development Index (HDI91 ) and two binary variables representing the biomes: V1 = 1 0 if vegetation is savanna otherwise and V2 = 1 0 if vegetation is forest otherwise After variables selection, the significance of several crossproduct interaction effects was tested. The final selected model, with R2 = 0.3568, consisted of the aforementioned variables and the interaction between HDI91 , and V1 , showing that the influence of HDI for the explanation of prevalence is different for savanna when compared to forest and caatinga. The final model for the prevalence is P̂v = e(−0.34+0.21TNs −3.92HDI91 +1.12V2 +0.33HDI91 V1 ) − 1 This model for each biome can be written as • Forest ⇒ P̂v = e(0.78+0.21TNs −3.92HDI91 ) − 1 • Savanna ⇒ P̂v = e(−0.34+0.21TNs −3.49HDI91 ) − 1 • Caatinga ⇒ P̂v = e(−0.34+0.21TNs −3.92HDI91 ) − 1 Fig. 2b shows the estimated prevalence for all Minas Gerais State municipalities using the above estimated regression equation, and Fig. 2c and d the explanatory variables TNs and HDI91 . Fig. 2e shows the plot of the residuals, resulting from the difference between observed (Fig. 2a) and estimated (Fig. 2b) schistosomiasis prevalence. In this Fig. 2e, dark colors (red and blue) represent overestimated values, light colors (cyan and magenta) underestimated ones, and in white are the municipalities where the estimatives differ from the true values by less than 5%, which is considered a good estimative for our purposes. Fig. 2f shows the standard error of the estimated prevalence. The result indicates that during the summer season the risk of contracting schistosomiasis increases, probably due to high concentrations of the snails in the drainage caused mainly by lack of sanitation, high temperature, among other factors, and by the population searching for water bodies, either for drinking or bathing. 4.2. Prevalence as a function of remote sensing variables The remote sensing data used were derived from MODIS and SRTM. MODIS MOD13 products from two dates (one in summer and another in winter) were taken, and for each date nine variables were used: blue, red, near infrared (NIR), middle infrared (MIR), enhanced vegetation index (EVI), normalized difference vegetation index (NDVI), plus vegetation (VEG), soil (SOIL) and shade (SHD) indices derived from the mixture model. Two variables from SRTM elevation data were also used: the digital elevation model itself and the declivity (DEC), derived from the DEM. An analysis of the correlation matrix showed that some variables had non-significative correlations with ln P̂v at 95% confidence level, and also part of them was highly correlated among themselves, indicating that the model can be further simplified. After variables selection using the R2 criterion, the final chosen model was that with two variables (DEM and NDVIw ). R.J.P.S. Guimarães et al. / Acta Tropica 108 (2008) 234–241 239 The final model generated by the regression function based on the 189 municipalities is Table 1 Review of Linear Regression Models types for prevalence risk estimation P̂v = e(0.83−0.002DEM+4.99NDVIw ) − 1 Model Type of input data Variables R2 RMSE 1 Social and meteorological Remote sensing derived Social, meteorological and remote sensing TNs, HDI91 , V2 , HDI91 × V1 DEM, NDVIw DEM, TNs × NDVIw 0.357 0.747 0.286 0.305 0.782 0.772 This model showed a coefficient of determination of 0.2859. Fig. 3b presents the estimated prevalence, and Fig. 4c and d the DEM and NDVIw , respectively. Fig. 3e and f shows the plot of the residuals, and the standard error of the estimated prevalence, respectively. The elevation was negatively correlated with schistosomiasis prevalence, while NDVI was positively correlated. This is consistent with the adequate environmental conditions for the development of the intermediate hosts, since the snails are settled in places with lower altitudes and vegetated areas. 4.3. Prevalence as a function of social, meteorological and remote sensing variables The third scenario was assessed using the previously selected variables, with the exception of the categorical variables representing the biomes. NDVI, which was chosen in the second scenario, should in principle substitute the information provided by biomes specification. The chosen model, which is statistically significant at 5% level, consisted of a model with three variables: summer minimum temperature (TNs ), the digital elevation model and winter normalized difference vegetation index (NDVIw ). After variables selection, the significance of several cross-product interaction effects was tested. The final selected model, with R2 = 0.3052, consisted of the DEM and the interaction between TNs and NDVIw . The final model generated by the regression function based on the 189 municipalities is P̂v = e(0.72−0.002DEM+0.24TNs NDVIw ) − 1 2 3 Fig. 4a shows the prevalence and Fig. 4b shows the estimated prevalence for all Minas Gerais State municipalities. The residuals and the standard error of the estimated prevalence are presented in Fig. 4c and d, respectively. The elevation is negatively correlated with schistosomiasis prevalence, while the interaction winter NDVI and summer minimum temperature is positively correlated. This is consistent with the adequate environmental conditions for the development of the intermediate hosts, since the snails are settled in places with lower altitudes and vegetated areas, and in the summer season the risk of contracting schistosomiasis increases, due to high concentrations of the snails, high temperature, and by the population searching for water bodies. Also, the interaction indicates that the prevalence of the schistosomiasis is related with the joint effect of temperature and vegetation. One can note that from the four (4) variables (two from model 1 and two from model 2) used in this alternative scenario the social variable HDI was not selected this time. This is due, probably, because the joint effect of minimum summer temperature and vegetation index is negatively correlated with HDI in Minas Gerais State. Fig. 4. Prevalence as a function of social, meteorological and remote sensing variables: (a) observed prevalence and (b) estimated prevalence—no information (white), prevalence (%) 0.001–5.000 (green), 5.001–15.000 (yellow), and above 15.001 (red); (c) residuals, below −10.001 (magenta), −10.000 to −5.001 (cyan), −5.000 to 5.000 (white), 5.000–10.000 (blue), and above 10.001 (red); and (d) standard error of the estimated prevalence, 0.060–0.100 (white), 0.101–0.150 (green), 0.151–0.200 (yellow), and 0.201–0.300 (red). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.) 240 R.J.P.S. Guimarães et al. / Acta Tropica 108 (2008) 234–241 5. Discussion and conclusions Acknowledgments Variables from different sources (social, meteorological and remote sensing) were used in this study. The prevalence risk was estimated by three models. Table 1 summarizes the results from these models, presenting the variables used in each model, the coefficient of determination and the root mean square error (RMSE). Table 1 shows that vegetation is an important variable, since it is present in all models, either as a qualitative variable (biome) or as a qualitative one (NDVI). It can be seen that the type 1 model (social–meteorological model) presented the higher R2 -value. Type 2 model (remote sensing model) presented the worse result, which shows that temperature is an important variable since it is appears on models 1 and 3. It is also noticeable that NDVI seems not to be a proper variable to substitute the biomes specification because type 3 model, that has all the previous variables but the biome specification, has a lower R2 -value than that obtained from the type 1 model which uses the biomes specification. Some important issues, related to the nature and precision of the prevalence data, need to be considered when looking at the results: the prevalence data were obtained from historical records (most occurring before the broad usage of GPS equipment), and the information is given in a municipality level basis, most of them located at North and Northeast of the State. Therefore, the data might not represent current reality of the disease, and the precise geolocation of the data, which is a characteristic of schistosomiasis disease, might be lost somehow. These issues might have affected and masked the correlations of the prevalence data with explanatory variables. The results, however, led to the determination of significant factors that are related to the disease and the delimitation of risk areas. The importance of the joint use of geographical information systems and remote sensing for illness risk estimation was evidenced. Therefore, even with a low coefficient of determination, it might be concluded that the joint usage of geographical information systems and statistical techniques, allowed the determination of related factors and the delimitation of risk areas for schistosomiasis. The authors would like to acknowledge the support of CNPq (grants# 305546/2003-1; 380203/2004-9; 304274/2005-4), Fapemig (EDP 1775/03; EDT 61775/03; CRA 0070/04) and NIHFogarty (grant# 5D43TW007012). 6. Further work A very large database was obtained for this project. This database can be further used for many other modeling studies for schistosomiasis and even for other diseases. In this paper some results were presented for prevalence risk estimation using environmental and other variables that are generalized over municipality area. One interest is to improve accuracy both on geolocation axis and on risk estimation precision. All logging methodology for field prevalence estimation is being modified to attend the more precise geolocation accuracy which is of paramount importance to develop risk models with better spatial resolution. It is expected that this will also improve the coefficient of determination on final model when compared with models generated with municipalities’ level data. Several other variables related to the water usage, such as sanitation, water quality, soil water retention, and existence of intermediate hosts, might be tested as explanatory variables to improve the model. Suitable regionalization of the State territory, taking in account as a criterion the uniformity of the relevant input factors can also improve the overall accuracy as suggested by better result obtained by model type 1. Also intensive studies are being planned to assess local prevalence on sites pointed out, by model, as of high risk. References Abdel-Rahman, M.S., El-Bahy, M.M., Malone, J.B., Thompson, R.A., El Bahy, N.M., 2001. Geographic information systems as a tool for control program management for schistosomiasis in Egypt. Acta Trop. 79, 49–57. Appleton, C.C., 1978. Review of literature on abiotic factors influencing the distribution and life cycles of bilharziasis intermediate host snails. Malacol. Rev. 11, 1–25. Barbosa, C.S., Araújo, K.C., Antunes, L., Favre, T., Piere, O.S., 2004. Spatial distribution of schistosomiasis foci on Itamaracá Island, Pernambuco. Brazil. Mem. Inst. Oswaldo Cruz. 99 (Suppl. I), 79–83. Bavia, M.E., Malone, J.B., Hale, L., Dantas, A., Marroni, L., Reis, R., 2001. Use of thermal and vegetation index data from earth observing satellites to evaluate the risk of schistosomiasis in Bahia, Brazil. Acta Trop. 79, 79–85. Beasley, M., Brooker, S., Ndinaromtan, M., Madjiouroum, E.M., Baboguel, M., Djenguinabe, E., Bundy, D.A.P., 2002. First nationwide survey of the health of schoolchildren in Chad. Trop. Med. Int. Health 7, 625–630. Beck, L.R., Lobitz, B.M., Wood, B.L., 2000. Remote sensing and human health: new sensors and new opportunities. Emerg. Infect. Dis. 6, 217–227. Beck, L.R., Rodriguez, M.H., Dister, S.W., Rodriguez, A.D., Washino, R.K., Roberts, D.R., Spanner, M.A., 1997. Assessment of a remote sensing based model for predicting malaria transmission risk in villages of Chiapas. Mexico Am. J. Trop. Med. Hyg. 56, 99–106. Brooker, S., Beasley, M., Ndinaromtan, M., Madjiouroum, E.M., Baboguel, M., Djenguinabe, E., Hay, S.I., Bundy, D.A.P., 2002. Use of remote sensing and a geographic information system in a national helminth control programme in Chad. Bull. World Health Organ. 80, 783–789. Brooker, S., Hay, S.I., Issae, W., Hall, A., Kihamia, C.M., Lwambo, N.J., Wint, W., Rogers, D.J., Bundy, D.A., 2001. Predicting the distribution of urinary schistosomiasis in Tanzania using satellite sensor data. Trop. Med. Int. Health 6, 998–1007. Brown, D.S., 1994. Freshwater Snails of Africa and their medical Importance. Taylor and Francis, London, United Kingdom. Carvalho, O.S., Amaral, R.S., Dutra, L.V., Scholte, R.G.C., Guerra, M.A.M., 2008. Distribuição espacial de Biomphalaria glabrata, B. straminea e B. tenagophila moluscos hospedeiros intermediários do Schistosoma mansoni no Brasil. In: O.S. Carvalho, P.M.Z. Coelho, H.L Lenzi (Eds.), Schistosoma mansoni e esquistossomose: uma visão multidisciplinar, Editora Fiocruz, Rio de Janeiro, RJ, pp. 393–418. Carvalho, O.S., Massara, C.L., Rocha, R.S., Katz, N., 1989. Esquistossomose mansoni no sudoeste do estado de Minas Gerais (Brasil). Rev. Saude Publ. 23, 341–344. Carvalho, O.S., Massara, C.L., Silveira Neto, H.V., Alvarenga, A.G., Vidigal, T.H.D.A., Chaves, A., Katz, N., 1994. Schistosomiasis mansoni in the region of Triângulo Mineiro—State of Minas Gerais (Brasil). Mem. Inst. Oswaldo Cruz. 89, 509–512. Carvalho, O.S., Massara, C.L., Silveira Neto, H.V., Guerra, H.L., Caldeira, R.L., Mendonça, C.L.F., Vidigal, T.H.D.A., Chaves, A., Katz, N., 1997. Re-evaluation of Schistosomiasis Mansoni in Minas Gerais, Brazil. II. Alto Paranaiba Mesoregion. Mem. Inst. Oswaldo Cruz. 92, 141–142. Carvalho, O.S., Nunes, I.M., Caldeira, R.L., 1998. First report of Biomphalaria glabrata in the state of Rio Grande do Sul, Brazil. Mem. Inst. Oswaldo Cruz. 93, 39–40. Carvalho, O.S., Rocha, R.S., Massara, C.L., Katz, N., 1987. Expansão da esquistossomose mansoni em Minas Gerais. Mem. Inst. Oswaldo Cruz. 82 (Suppl. IV), 295–298. Carvalho, O.S., Rocha, R.S., Massara, C.L., Katz, N., 1988. Primeiros casos autóctones de esquistossomose mansonica em região do noroeste do Estado de Minas Gerais (Brasil). Rev. Saude Publ. 22, 237–239. Chitsulo, L., Engels, D., Montresor, A., Savioli, L., 2000. The global status of schistosomiasis and its control. Acta Trop. 77, 41–51. Cross, E.R., Newcomb, W.W., Tucker, C.J., 1996. Use of weather data and remote sensing to predict the geographic and seasonal distribution of Phlebotomus papatasi in southwest Asia. Am. J. Trop. Med. Hyg. 54, 530–536. Cross, E.R., Sheffield, C., Perrine, R., Pazzaglia, G., 1984. Predicting areas endemic for schistosomiasis using weather variables and a Landsat data base. Mil. Med. 149, 542–544. Graeff, T.C., Anjos, C.B., Oliveira, V.C., Velloso, C.F.P., Fonseca, M.B.S., Valar, C., Moraes, C., Garrido, C.T., Amaral, R.S., 1999. Identification of a transmission focus of Schistosoma mansoni in the southernmost Brazilian State, Rio Grande do Sul. Mem. Inst. Oswaldo Cruz. 94, 9–10. Guimarães, R.J.P.S., Freitas, C.C., Dutra, L.V., Moura, A.C.M., Amaral, R.S., Drummond, S.C., Guerra, M., Scholte, R.G.C., Freitas, C.R., Carvalho, O.S., 2006. Analysis and estimative of schistosomiasis prevalence for Minas Gerais state, Brazil, using multiple regression with social and environmental spatial data. Mem. Inst. Oswaldo Cruz. 101 (Suppl. I), 91–96. Huete, A.R., Justice, C., van Leewen, W., 1999. MODIS Vegetation Index (MOD13) Algorithm Theoretical Basis Document, Version 3. University of Arizona, Tucson. Justice, C.O., Vermote, E., Townshend, J.R.G., Defries, R., Roy, P.D., Hall, D.K., Salomonson, V., Privette, J.L., Riggs, G., Strahler, A., Lucht, W., Myneni, B., Knyazikhin, Y., R.J.P.S. Guimarães et al. / Acta Tropica 108 (2008) 234–241 Running, W.S., Nemani, R.R., Wan, Z., Huete, A.R., Leeuwen, W.V., Wolfe, R.E., Giglio, L., Muller, J.P., Lewis, P., Barnsley, M., 1998. The moderate resolution imaging spectroradiometer (MODIS): land remote sensing for global change research. IEEE Trans. Geosci. Rem. Sens. 36, 1228–1247. Kabatereine, N.B., Brooker, S., Tukahebwa, E.M., Kazibwe, F., Onapa, A.W., 2004. Epidemiology and geography of Schistosoma mansoni in Uganda: implications for planning control. Trop. Med. Int. Health 9, 372–380. Katz, N., Carvalho, O.S., 1983. Introdução recente da esquistossomose mansoni no sul do estado de Minas Gerais, Brasil. Mem. Inst. Oswaldo Cruz. 78, 281–284. Katz, N., Dias, E.P., Souza, C.P., Bruce, J.I., Coles, G.C., 1989. Rate of action of schistosomicides in mice infected with Schistosoma mansoni. Rev. Soc. Bras. Med. Trop. 22, 183–186. Katz, N., Mota, E., Oliveira, V.B., Carvalho, E.F., 1978. Prevalência da esquistossomose em escolares no estado de Minas Gerais. In: Proceedings of the XIV Congresso da Sociedade Brasileira de Medicina Tropical e III Congresso da Sociedade Brasileira de Parasitologia, João Pessoa, Brazil, p. 102 (Abstract book). Kristensen, T.K., Malone, J.B., Mccarroll, J.C., 2001. Use of satellite remote sensing and geographic information systems to model the distribution and abundance of snail intermediate hosts in Africa: a preliminary model for Biomphalaria pfeifferi in Ethiopia. Acta Trop. 79, 73–78. Lambertucci, J.R., Rocha, R.S., Carvalho, O.S., Katz, N., 1987. A esquistossomose mansoni em Minas Gerais. Rev. Soc. Bras. Med. Trop. 20, 47–52. Malone, J.B., Abdel-Rahman, M.S., El Bahy, M.M., Huh, O.K., Shafik, M., Bavia, M., 1997. Geographic information systems and the distribution of Schistosoma mansoni in the Nile delta. Parasitol. Today 13, 112–119. 241 Malone, J.B., Huh, O.K., Fehler, D.P., Wilson, P.A., Wilensky, D.E., Holmes, R.A., Elmagdoub, A.L., 1994. Temperature data from satellite imagery and distribution of schistosomiasis in Egypt. Am. J. Trop. Med. Hyg. 51, 714–722. Malone, J.B., Yilma, J.M., Mccarroll, J.C., Erko, B., Mukaratirwa, S., Zhou, X., 2001. Satellite climatology and the environmental risk of Schistosoma mansoni in Ethiopia and east Africa. Acta Trop. 79, 59–72. Neter, J., Kutner, M.H., Nachtssheim, C.J., Wasserman, W., 1996. Applied Linear Statistical models. McGraw-Hill, Boston. Pellon, A.B., Teixeira, I., 1950. Distribuição geográfica da esquistossomose mansônica no Brasil. Divisão de Organização Sanitária, Rio de Janeiro. Seto, E., Xu, B., Liang, S., Gong, P., Wu, W., Davis, G.M., Qui, D., Gu, X., Spear, R., 2002. The use of remote sensing for predictive modeling of schistosomiasis in China. Photogramm Eng. Rem. Sens. 68, 167–174. Shimabukuro, Y.E., Smith, J.A., 1991. The least-squares mixing models to generate fraction images derived from remote sensing multispectral data. IEEE Trans. Geosci. Rem. Sens. 29, 16–20. World Health Organization (WHO), 1985. The Control of Schistosomiasis. WHO Technical Series Report 728. WHO, Geneva, Switzerland. Yang, G.-J., Vounatsou, P., Xiao-Nong, Z., Utzinger, J., Tanner, M., 2005. A review of geographic information system and remote sensing with applications to the epidemiology and control of schistosomiasis in China. Acta Trop. 96, 117–129. Zhou, X.N., Malone, J.B., Kristensen, T.K., Bergquist, N.R., 2001. Application of geographic information systems and remote sensing to schistosomiasis control in China. Acta Trop. 79, 97–106.