Search

Land use/land cover (LULC) is one of the most impactful global change phenomenon. As a result, considerable effort has been devoted to creating large-scale LULC products from remote sensing data, enabling the scientific community to use these products for a wide range of downstream applications. Unfortunately, uncertainty associated with these products is seldom quantified because most approaches are too computationally intensive. Furthermore, uncertainty maps developed for large regions might fail to perform adequately at the spatial scale in which they will be used and might need to be customized to suit the specific applications of end-users.In this study, we describe the class-conditional conformal statistics method, an approach that quantifies uncertainty more uniformly for each class but that requires more calibration data than the conventional conformal method. Using the class-conditional method, we show that it is possible to create customized local uncertainty maps using local calibration data without requiring remote sensing and modeling work and that these local uncertainty maps outperform uncertainty maps calibrated based on global data. We use empirical data from Brazil (i.e., Dynamic World LULC product and Mapbiomas validation data) to demonstrate this methodology. The analysis of these data reveals substantial heterogeneity in observations of the same LULC class between Brazilian states, an indication that national-level data are not representative of the focal state, thus explaining why uncertainty maps calibrated using focal state-level data outperform maps calibrated using national-level data. Importantly, we develop straight-forward approaches to determine the spatial extent over which calibration data are still representative of the area of interest, ensuring that these data can be used to reliably quantify uncertainty. We illustrate the class-conformal methodology by creating uncertainty maps for a selected number of sites in Brazil. Finally, we show how these uncertainty maps can yield valuable insights for LULC map producers.Our methodology paves the way for users to generate customized local uncertainty maps that are likely to be better than uncertainty maps calibrated based on global data while at the same time being more relevant for the specific applications of these users. A tutorial is provided to show how this methodology can be implemented without requiring remote sensing and modeling expertise to generate uncertainty maps.

Abstract Proximity to roads is one of the main determinants of deforestation in the Amazon basin. Determining the construction year of roads (CYR) is critical to improve the understanding of the drivers of road construction and to enable predictions of the expansion of the road network and its consequent impact on ecosystems. While recent artificial intelligence approaches have been successfully used for road extraction, they have typically relied on high spatial‐resolution imagery, precluding their adoption for the determination of CYR for older roads. In this article, we developed a new approach to automate the process of determining CYR that relies on the approximate position of the current road network and a time‐series of the proportion of exposed soil based on the multidecadal remote sensing imagery from the Landsat program. Starting with these inputs, our methodology relies on the Least Cost Path algorithm to co‐register the road network and on a Before‐After Control‐Impact design to circumvent the inherent image‐to‐image variability in the estimated amount of exposed soil. We demonstrate this approach for a 357 000 km2 area around the Transamazon highway (BR‐230) in the Brazilian Amazon, encompassing 36 240 road segments. The reliability of this approach is assessed by comparing the estimated CYR using our approach to the observed CYR based on a time‐series of Landsat images. This exercise reveals a close correspondence between the estimated and observed CYR (rPearson=0.77). Finally, we show how these data can be used to assess the effectiveness of protected areas (PAs) in reducing the yearly rate of road construction and thus their vulnerability to future degradation. In particular, we find that integral protection PAs in this region were generally more effective in reducing the expansion of the road network when compared to sustainable use PAs.

Abstract Background Understanding how to connect habitat remnants to facilitate the movement of species is a critical task in an increasingly fragmented world impacted by human activities. The identification of dispersal routes and corridors through connectivity analysis requires measures of landscape resistance but there has been no consensus on how to calculate resistance from habitat characteristics, potentially leading to very different connectivity outcomes. Methods We propose a new model, called the Time-Explicit Habitat Selection (TEHS) model, that can be directly used for connectivity analysis. The TEHS model decomposes the movement process in a principled approach into a time and a selection component, providing complementary information regarding space use by separately assessing the drivers of time to traverse the landscape and the drivers of habitat selection. These models are illustrated using GPS-tracking data from giant anteaters (Myrmecophaga tridactyla) in the Pantanal wetlands of Brazil. Results The time model revealed that the fastest movements tended to occur between 8 p.m. and 5 a.m., suggesting a crepuscular/nocturnal behavior. Giant anteaters moved faster over wetlands while moving much slower over forests and savannas, in comparison to grasslands. We also found that wetlands were consistently avoided whereas forest and savannas tended to be selected. Importantly, this model revealed that selection for forest increased with temperature, suggesting that forests may act as important thermal shelters when temperatures are high. Finally, using the spatial absorbing Markov chain framework, we show that the TEHS model results can be used to simulate movement and connectivity within a fragmented landscape, revealing that giant anteaters will often not use the shortest-distance path to the destination patch due to avoidance of certain habitats. Conclusions The proposed approach can be used to characterize how landscape features are perceived by individuals through the decomposition of movement patterns into a time and a habitat selection component. Additionally, this framework can help bridge the gap between movement-based models and connectivity analysis, enabling the generation of time-explicit connectivity results.

The use of multistep upscaling approaches in which field data are combined with data from multiple remote sensors that operate at different spatial scales (e.g., UAV LiDAR, GEDI, and Landsat) is becoming increasingly popular. In these approaches, a series of models are fitted linking the information from these different sensors, often resulting in improved predictions over large areas. Quantifying the uncertainty associated with individual models can be challenging as these models may not generate uncertainty estimates (e.g., machine learning models such as random forest), a problem that is further exacerbated if the results from multiple models are combined within a multistep upscaling methodology. In this article, we describe a nonparametric conformal approach to quantify uncertainty. This approach is straight-forward to apply, is computationally inexpensive (differently from bootstrapping), and generates improved predictive intervals. Importantly, this methodology can be used regardless of the number of models adopted in the upscaling approach and the nature of the intermediate models, as long as the final model can generate predictive intervals. We illustrate the improved empirical coverage of the conformalized predictive intervals using simulated data for a two-step upscaling scenario involving field, UAV LiDAR, and Landsat data. This simulation exercise shows how increasing uncertainty in the first stage model (which relates biomass field data to UAV LiDAR data) leads to an increase in the severity of uncertainty underestimation by naïve predictive intervals. On the other hand, conformalized predictive intervals do not exhibit this shortcoming. Finally, we illustrate uncertainty quantification for a multistep upscaling methodology using data from a large-scale carbon project in the Brazilian Amazon. Our validation exercise using these empirical data confirms the improved performance of the conformalized predictive intervals. We expect that the conformal approach described here will be key for uncertainty quantification as multistep upscaling approaches become increasingly more common.

Visceral leishmaniasis (VL) is the second most common protozoosis that affects people around the world. The aim of this study is to understand how environmental and socioeconomic factors, as well as VL control and surveillance interventions, influence the spread and detection of VL cases in Pernambuco state (Brazil). A novel model was developed to analyze cases of VL between 2007 and 2018, enabling the quantification of the association of these variables with two processes: the probability of "invasion" (emergence of new cases) at municipalities by VL, and the probability of detecting cases not reported in municipalities that have already been invaded. Pernambuco state identified 1,410 cases of VL between 2007 and 2018, with an average of 128 cases per year and average incidence of 1.28/100 thousand people. These cases were distributed in 77.1% (142/184) of the municipalities, and 54.8% (773/1,410) of them were autochthonous. Our model reveals that the proportion of agriculture was positively associated with VL invasion probability. We also find that municipalities that are closer to notification centers and/or that have received technical training and support tend to have higher detection rates of VL cases. Taken together, these results suggest that a municipality with almost no agriculture and that received technical training, located close to a notification center, is unlikely to be invaded if no cases have ever been detected. On the other hand, a municipality that is far from the notification center, with no technical training, with a large agricultural area might have already been invaded but the surveillance system might have routinely failed to detect VL cases due to low detection probability. By disentangling the processes of invasion and detection, we were able to generate insights that are likely to be useful for the strategic allocation of VL prevention and control interventions.

Advances in biologging have increased the understanding of how animals interact with their environment, especially for cryptic species. For example, giant armadillos (Priodontes maximus) are the largest extant species of armadillo but are rarely encountered due to their fossorial and nocturnal behavior. Through the analysis of speed, turning angles, and accelerometer activity counts, we estimated behavioral states, characterized activity budgets, and investigated the state-habitat associations exhibited by individuals monitored with GPS telemetry in the Brazilian Pantanal from 2019 to 2020. This methodology is proposed as a useful framework for the identification of priority habitat. Using the non-parametric Bayesian mixture model for movement (M3), we estimated four latent behavioral states that were named ‘vigilance-excavation’, ‘local search’, ‘exploratory’, and ‘transit’. These states appeared to correspond with behavior near burrows or termite mounds, foraging, ranging, and rapid movements, respectively. The first and last hours of activity presented relatively high proportions of the vigilance-excavation state, while most of the activity period was dominated by local search and exploratory states. The vigilance-excavation state occurred more frequently in regions between forest and closed savannas, whereas local search was more likely in high proportions of closed savanna. Exploratory behavior probability increased in areas with high proportions of both forest and closed savanna. Our results establish a baseline for behavioral complexity, activity budgets, and habitat associations in a relatively pristine environment that can be used for future work to investigate anthropogenic impacts on giant armadillo behavior and fitness. The integration of accelerometer and GPS-derived movement data through our mixture model has the potential to become a powerful methodological approach for the conservation of other cryptic species.

Abstract Background Access to healthcare is important in controlling malaria burden and, as a result, distance or travel time to health facilities is often a significant predictor in modelling malaria prevalence. Adding new health facilities may reduce overall travel time to health facilities and may decrease malaria transmission. To help guide local decision-makers as they scale up community-based accessibility, the influence of the spatial allocation of new health facilities on malaria prevalence is evaluated in Bunkpurugu-Yunyoo district in northern Ghana. A location-allocation analysis is performed to find optimal locations of new health facilities by separately minimizing three district-wide objectives: malaria prevalence, malaria incidence, and average travel time to health facilities. Methods Generalized additive models was used to estimate the relationship between malaria prevalence and travel time to the nearest health facility and other geospatial covariates. The model predictions are then used to calculate the optimisation criteria for the location-allocation analysis. This analysis was performed for two scenarios: adding new health facilities to the existing ones, and a hypothetical scenario in which the community-based healthcare facilities would be allocated anew. An interactive web application was created to facilitate efficient presentation of this analysis and allow users to experiment with their choice of health facility location and optimisation criteria. Results Using malaria prevalence and travel time as optimisation criteria, two locations that would benefit from new health facilities were identified, regardless of scenarios. Due to the non-linear relationship between malaria incidence and prevalence, the optimal locations chosen based on the incidence criterion tended to be inequitable and was different from those based on the other optimisation criteria. Conclusions This study findings underscore the importance of using multiple optimisation criteria in the decision-making process. This analysis and the interactive application can be repurposed for other regions and criteria, bridging the gap between science, models and decisions.

Abstract Understanding and predicting the effect of global change phenomena on biodiversity is challenging given that biodiversity data are highly multivariate, containing information from tens to hundreds of species in any given location and time. The Latent Dirichlet Allocation (LDA) model has been recently proposed to decompose biodiversity data into latent communities. While LDA is a very useful exploratory tool and overcomes several limitations of earlier methods, it has limited inferential and predictive skill given that covariates cannot be included in the model. We introduce a modified LDA model (called LDAcov) which allows the incorporation of covariates, enabling inference on the drivers of change of latent communities, spatial interpolation of results, and prediction based on future environmental change scenarios. We show with simulated data that our approach to fitting LDAcov is able to estimate well the number of groups and all model parameters. We illustrate LDAcov using data from two experimental studies on the long‐term effects of fire on southeastern Amazonian forests in Brazil. Our results reveal that repeated fires can have a strong impact on plant assemblages, particularly if fuel is allowed to build up between consecutive fires. The effect of fire is exacerbated as distance to the edge of the forest decreases, with small‐sized species and species with thin bark being impacted the most. These results highlight the compounding impacts of multiple fire events and fragmentation, a scenario commonly found across the southern edge of Amazon. We believe that LDAcov will be of wide interest to scientists studying the effect of global change phenomena on biodiversity using high‐dimensional datasets. Thus, we developed the R package LDAcov to enable the straightforward use of this model.

Abstract Background Mass drug administration and mass-screen-and-treat interventions have been used to interrupt malaria transmission and reduce burden in sub-Saharan Africa. Determining which strategy will reduce costs is an important challenge for implementers; however, model-based simulations and field studies have yet to develop consensus guidelines. Moreover, there is often no way for decision-makers to directly interact with these data and/or models, incorporate local knowledge and expertise, and re-fit parameters to guide their specific goals. Methods We propose a general framework for comparing costs associated with mass drug administrations and mass screen and treat based on the possible outcomes of each intervention and the costs associated with each outcome. We then used publicly available data from six countries in western Africa to develop spatial-explicit probabilistic models to estimate intervention costs based on baseline malaria prevalence, diagnostic performance, and sociodemographic factors (age and urbanicity). In addition to comparing specific scenarios, we also develop interactive web applications which allow managers to select data sources and model parameters, and directly input their own cost values. Results The regional-level models revealed substantial spatial heterogeneity in malaria prevalence and diagnostic test sensitivity and specificity, indicating that a “one-size-fits-all” approach is unlikely to maximize resource allocation. For instance, urban communities in Burkina Faso typically had lower prevalence rates compared to rural communities (0.151 versus 0.383, respectively) as well as lower diagnostic sensitivity (0.699 versus 0.862, respectively); however, there was still substantial regional variation. Adjusting the cost associated with false negative diagnostic results to included additional costs, such as delayed treated and potential lost wages, undermined the overall costs associated with MSAT. Conclusions The observed spatial variability and dependence on specified cost values support not only the need for location-specific intervention approaches but also the need to move beyond standard modeling approaches and towards interactive tools which allow implementers to engage directly with data and models. We believe that the framework demonstrated in this article will help connect modeling efforts and stakeholders in order to promote data-driven decision-making for the effective management of malaria, as well as other diseases.

The proliferation of river infrastructure projects has altered aquatic longitudinal connectivity, posing a growing threat to riverine biodiversity and ecosystem processes worldwide. Effective methods to quantify loss of river connectivity across spatiotemporal scales and in data-limited landscapes are important to understand and inform basin-wide conservation and development planning. Here we introduce a Catchment Area-based Fragmentation Index (CAFI) and its derivative, the Catchment Area- and Rainfall-based Fragmentation Index (CARFI) as new metrics to quantify river fragmentation. These indices use catchment area as a proxy for riverine habitat availability, avoiding the drawbacks of existing metrics that rely on river length and associated derivatives. CAFI/CARFI can be computed across spatiotemporal scales, incorporate barrier passability values, assesses the cumulative impact of multiple barriers, and be applied even in data-limited environments.We first applied CAFI and CARFI to a simulated network to illustrate their properties with respect to the number and location of barriers and compared these results to the widely applied Dendritic Connectivity Index (DCI). While all indices varied with barrier addition, CAFI and CARFI were more sensitive to both barrier number and location. Next, we illustrated the utility of CAFI and CARFI through case studies in two contrasting settings: the Klamath River in California, where dam building has ceased (and dam removals are being considered) and the Netravathi River in India, where dam building is ongoing, with 65 dams proposed for future development. Results indicate that CAFI and CARFI can effectively quantify trends in fragmentation across spatial scales and temporal scenarios of dam development (i.e. descriptive applications) and can aid the prioritization of sites for dam removal, restoration, or conservation (i.e. prescriptive applications). Overall, these indices can quantify the impacts of individual dams and assess a range of development scenarios to inform basin-wide conservation and development planning.

The use of scientific web applications (SWApps) across biological and environmental sciences has grown exponentially over the past decade or so. Although quantitative evidence for such increased use in practice is scant, collectively, we have observed that these tools become more commonplace in teaching, outreach, and in science coproduction (e.g., as decision support tools). Despite the increased popularity of SWApps, researchers often receive little or no training in creating such tools. Although rolling out SWApps can be a relatively simple and quick process using modern, popular platforms like R shiny apps or Tableau dashboards, making them useful, usable, and sustainable is not. These 10 simple rules for creating a SWApp provide a foundation upon which researchers with little to no experience in web application design and development can consider, plan, and carry out SWApp projects.

Abstract Background Bayesian methods have been used to generate country-level and global maps of malaria prevalence. With increasing availability of detailed malaria surveillance data, these methodologies can also be used to identify fine-scale heterogeneity of malaria parasitaemia for operational prevention and control of malaria. Methods In this article, a Bayesian geostatistical model was applied to six malaria parasitaemia surveys conducted during rainy and dry seasons between November 2010 and 2013 to characterize the micro-scale spatial heterogeneity of malaria risk in northern Ghana. Results The geostatistical model showed substantial spatial heterogeneity, with malaria parasite prevalence varying between 19 and 90%, and revealing a northeast to southwest gradient of predicted risk. The spatial distribution of prevalence was heavily influenced by two modest urban centres, with a substantially lower prevalence in urban centres compared to rural areas. Although strong seasonal variations were observed, spatial malaria prevalence patterns did not change substantially from year to year. Furthermore, independent surveillance data suggested that the model had a relatively good predictive performance when extrapolated to a neighbouring district. Conclusions This high variability in malaria prevalence is striking, given that this small area (approximately 30 km × 40 km) was purportedly homogeneous based on country-level spatial analysis, suggesting that fine-scale parasitaemia data might be critical to guide district-level programmatic efforts to prevent and control malaria. Extrapolations results suggest that fine-scale parasitaemia data can be useful for spatial predictions in neighbouring unsampled districts and does not have to be collected every year to aid district-level operations, helping to alleviate concerns regarding the cost of fine-scale data collection.

Abstract Background There is a need for comprehensive evaluations of the underlying local factors that contribute to residual malaria in sub-Saharan Africa. However, it is difficult to compare the wide array of demographic, socio-economic, and environmental variables associated with malaria transmission using standard statistical approaches while accounting for seasonal differences and nonlinear relationships. This article uses a Bayesian model averaging (BMA) approach for identifying and comparing potential risk and protective factors associated with residual malaria. Results The relative influence of a comprehensive set of demographic, socio-economic, environmental, and malaria intervention variables on malaria prevalence were modelled using BMA for variable selection. Data were collected in Bunkpurugu-Yunyoo, a rural district in northeast Ghana that experiences holoendemic seasonal malaria transmission, over six biannual surveys from 2010 to 2013. A total of 10,022 children between the ages 6 to 59 months were used in the analysis. Multiple models were developed to identify important risk and protective factors, accounting for seasonal patterns and nonlinear relationships. These models revealed pronounced nonlinear associations between malaria risk and distance from the nearest urban centre and health facility. Furthermore, the association between malaria risk and age and some ethnic groups was significantly different in the rainy and dry seasons. BMA outperformed other commonly used regression approaches in out-of-sample predictive ability using a season-to-season validation approach. Conclusions This modelling framework offers an alternative approach to disease risk factor analysis that generates interpretable models, can reveal complex, nonlinear relationships, incorporates uncertainty in model selection, and produces accurate predictions. Certain modelling applications, such as designing targeted local interventions, require more sophisticated statistical methods which are capable of handling a wide range of relevant data while maintaining interpretability and predictive performance, and directly characterize uncertainty. To this end, BMA represents a valuable tool for constructing more informative models for understanding risk factors for malaria, as well as other vector-borne and environmentally mediated diseases.

As the world's population grows, global food production will need to increase. While food production efficiency has increased in recent decades through pathogen control, climate change poses new challenges in crop protection against pathogens. Understanding the natural geographical distribution and dispersal likelihood of fungal plant pathogens is essential for forecasting disease plant spread. Here we used cultivation-independent techniques to identify fungal plant pathogens in 1,289 near-surface dust samples collected across the United States. We found that overall fungal pathogen community composition is more related to environmental conditions (in particular soil pH, precipitation and frost) than to agricultural hosts and practices. We also delimited five susceptibility geographical areas in the United States where different sets of pathogens tend to occur.

Abstract Background Considerable debate has arisen regarding the appropriateness of the test and treat malaria policy broadly recommended by the World Health Organization. While presumptive treatment has important drawbacks, the effectiveness of the test and treat policy can vary considerably across regions, depending on several factors such as baseline malaria prevalence and rapid diagnostic test (RDT) performance. Methods To compare presumptive treatment with test and treat, generalized linear mixed effects models were fitted to data from 6510 children under five years of age from Burkina Faso’s 2010 Demographic and Health Survey. Results The statistical model results revealed substantial regional variation in baseline malaria prevalence (i.e., pre-test prevalence) and RDT performance. As a result, a child with a positive RDT result in one region can have the same malaria infection probability as a demographically similar child with a negative RDT result in another region. These findings indicate that a test and treat policy might be reasonable in some settings, but may be undermined in others due to the high proportion of false negatives. Conclusions High spatial variability can substantially reduce the effectiveness of a national level test and treat malaria policy. In these cases, region-specific guidelines for malaria diagnosis and treatment may need to be formulated. Based on the statistical model results, proof-of-concept, web-based tools were created that can aid in the development of these region-specific guidelines and may improve current malaria-related policy in Burkina Faso.

It is computationally challenging to fit models to big data. For example, satellite imagery data often contain billions to trillions of pixels and it is not possible to use a pixel-level analysis to identify drivers of land-use change and create predictions using all the data. A common strategy to reduce sample size consists of drawing a random sample but this approach is not ideal when the outcome of interest is rare in the landscape because it leads to very few pixels with this outcome. Here we show that a case-control (CC) sampling approach, in which all (or a large fraction of) pixels with the outcome of interest and a subset of the pixels without this outcome are selected, can yield much better inference and prediction than random sampling (RS) if the estimated parameters and probabilities are adjusted with the equations that we provide. More specifically, we show that a CC approach can yield unbiased inference with much less uncertainty when CC data are analyzed with logistic regression models and its semiparametric variants (e.g., generalized additive models). We also show that a random forest model, when fitted to CC data, can generate much better predictions than when fitted to RS data. We illustrate this improved performance of the CC approach, when used together with the proposed bias-correction adjustments, with extensive simulations and a case study in the Amazon region focused on deforestation.

Visceral leishmaniasis (VL) is an important neglected disease caused by a protozoan parasite, and represents a serious public health problem in many parts of the world. It is zoonotic in Europe and Latin America, where infected dogs constitute the main domestic reservoir for the parasite and play a key role in VL transmission to humans. In Brazil this disease is caused by the protozoan Leishmania infantum chagasi, and is transmitted by the sand fly Lutzomyia longipalpis. Despite programs aimed at eliminating infection sources, the disease continues to spread throughout the Country. VL in São Paulo State, Brazil, first appeared in the northwestern region, spreading in a southeasterly direction over time. We integrate data on the VL vector, infected dogs and infected human dispersion from 1999 to 2013 through an innovative spatial temporal Bayesian model in conjunction with geographic information system. This model is used to infer the drivers of the invasion process and predict the future progression of VL through the State. We found that vector dispersion was influenced by vector presence in nearby municipalities at the previous time step, proximity to the Bolívia-Brazil gas pipeline, and high temperatures (i.e., annual average between 20 and 23°C). Key factors affecting infected dog dispersion included proximity to the Marechal Rondon Highway, high temperatures, and presence of the competent vector within the same municipality. Finally, vector presence, presence of infected dogs, and rainfall (approx. 270 to 540mm/year) drove the dispersion of human VL cases. Surprisingly, economic factors exhibited no noticeable influence on disease dispersion. Based on these drivers and stochastic simulations, we identified which municipalities are most likely to be invaded by vectors and infected hosts in the future. Prioritizing prevention and control strategies within the identified municipalities may help halt the spread of VL while reducing monitoring costs. Our results contribute important knowledge to public and animal health policy planning, and suggest that prevention and control strategies should focus on vector control and on blocking contact between vectors and hosts in the priority areas identified to be at risk.

The study of the effect of large-scale drivers (e.g., climate) of human diseases typically relies on aggregate disease data collected by the government surveillance network. The usual approach to analyze these data, however, often ignores a) changes in the total number of individuals examined, b) the bias towards symptomatic individuals in routine government surveillance, and; c) the influence that observations can have on disease dynamics. Here, we highlight the consequences of ignoring the problems listed above and develop a novel modeling framework to circumvent them, which is illustrated using simulations and real malaria data. Our simulations reveal that trends in the number of disease cases do not necessarily imply similar trends in infection prevalence or incidence, due to the strong influence of concurrent changes in sampling effort. We also show that ignoring decreases in the pool of infected individuals due to the treatment of part of these individuals can hamper reliable inference on infection incidence. We propose a model that avoids these problems, being a compromise between phenomenological statistical models and mechanistic disease dynamics models; in particular, a cross-validation exercise reveals that it has better out-of-sample predictive performance than both of these alternative models. Our case study in the Brazilian Amazon reveals that infection prevalence was high in 2004-2008 (prevalence of 4% with 95% CI of 3-5%), with outbreaks (prevalence up to 18%) occurring during the dry season of the year. After this period, infection prevalence decreased substantially (0.9% with 95% CI of 0.8-1.1%), which is due to a large reduction in infection incidence (i.e., incidence in 2008-2010 was approximately one fifth of the incidence in 2004-2008).We believe that our approach to modeling government surveillance disease data will be useful to advance current understanding of large-scale drivers of several diseases.

Large-scale forest conservation projects are underway in the Brazilian Amazon but little is known regarding their public health impact. Current literature emphasizes how land clearing increases malaria incidence, leading to the conclusion that forest conservation decreases malaria burden. Yet, there is also evidence that proximity to forest fringes increases malaria incidence, which implies the opposite relationship between forest conservation and malaria. We compare the effect of these environmental factors on malaria and explore its implications.Using a large malaria dataset (~1,300,000 positive malaria tests collected over ~4.5 million km(2)), satellite imagery, permutation tests, and hierarchical Bayesian regressions, we show that greater forest cover (as a proxy for proximity to forest fringes) tends to be associated with higher malaria incidence, and that forest cover effect was 25 times greater than the land clearing effect, the often cited culprit of malaria in the region. These findings have important implications for land use/land cover (LULC) policies in the region. We find that cities close to protected areas (PA's) tend to have higher malaria incidence than cities far from PA's. Using future LULC scenarios, we show that avoiding 10% of deforestation through better governance might result in an average 2-fold increase in malaria incidence by 2050 in urban health posts.Our results suggest that cost analysis of reduced carbon emissions from conservation efforts in the region should account for increased malaria morbidity, and that conservation initiatives should consider adopting malaria mitigation strategies. Coordinated actions from disparate science fields, government ministries, and global initiatives (e.g., Reduced Emissions from Deforestation and Degradation; Millenium Development Goals; Roll Back Malaria; and Global Fund to Fight AIDS, Tuberculosis and Malaria), will be required to decrease malaria toll in the region while preserving these important ecosystems.

BACKGROUND: A common challenge to the study of several infectious diseases consists in combining limited cross-sectional survey data, collected with a more sensitive detection method, with a more extensive (but biased) syndromic sentinel surveillance data, collected with a less sensitive method. Our article describes a novel modeling framework that overcomes this challenge, resulting in enhanced understanding of malaria in the Western Brazilian Amazon. METHODOLOGY/PRINCIPAL FINDINGS: A cohort of 486 individuals was monitored using four cross-sectional surveys, where all participants were sampled regardless of symptoms (aggressive-active case detection), resulting in 1,383 microscopy and 1,400 polymerase chain reaction tests. Data on the same individuals were also obtained from the local surveillance facility (i.e., passive and active case detection), totaling 1,694 microscopy tests. Our model accommodates these multiple pathogen and case detection methods. This model is shown to outperform logistic regression in terms of interpretability of its parameters, ability to recover the true parameter values, and predictive performance. We reveal that the main infection determinant was the extent of forest, particularly during the rainy season and in close proximity to water bodies, and participation on forest activities. We find that time residing in Acrelandia (as a proxy for past malaria exposure) decreases infection risk but surprisingly increases the likelihood of reporting symptoms once infected, possibly because non-naïve settlers are only susceptible to more virulent Plasmodium strains. We suggest that the search for asymptomatic carriers should focus on those at greater risk of being infected but lower risk of reporting symptoms once infected. CONCLUSIONS/SIGNIFICANCE: The modeling framework presented here combines cross-sectional survey data and syndromic sentinel surveillance data to shed light on several aspects of malaria that are critical for public health policy. This framework can be adapted to enhance inference on infectious diseases whenever asymptomatic carriers are important and multiple datasets are available.