Andrew D. M. Dobson
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Andrew D. M. Dobson has a total of 91 co-authors. Here are the top co-authors by number of shared publications:
Sarah E. Randolph
Department of Zoology, University of Oxford, South Parks Road, Oxford OX1 3PS, UK
3 shared publications- Sarah E. Randolph
Department of Zoology, University of Oxford, Oxford, United Kingdom
3 shared publications Aidan Keane
2 shared publications- Aidan Keane
School of GeoSciences University of Edinburgh Edinburgh UK
2 shared publications - Eleanor Jane Milner-Gulland
University of Oxford
2 shared publications - Harriet Ibbett
School of Natural Sciences, College of Environmental Science and Engineering Bangor University Bangor UK
2 shared publications
Latest Publications
- Open Access
- 1 April 2025
- Published by Springer Nature in Landscape Ecology
- Vol. 40 (4) , 1-19
Combining occupancy and dispersal models to predict the conservation benefits of land-use change
Context: Economic incentives for habitat creation in productive agricultural land are an important tool for halting biodiversity loss. Given inevitable funding constraints, assessing the cost-effectiveness of potential habitat creation strategies should be a priority. Objectives: We make model predictions of changes in the populations of ten woodland-affiliated bird species over time across a set of contrasting woodland creation scenarios in a 25 × 25-km square of primarily arable land in England. Methods: Using an economic agent-based model of landowners who make economically rational decisions on land use in response to financial incentives for woodland creation, we created a suite of potential landscapes with varying spatial configurations (and amounts) of newly planted woodland. We then parameterised an individual-based model for each of ten woodland bird species, and used these models to simulate dispersal and settlement across the altered landscapes. Results: Across all bird species, population increases were greater when woodland creation plots were clumped together than when randomly sited. When woodland plots were adjacent to existing woodland, increases were comparable to those in the clumped arrangements, except at lower planting densities, where wood-adjacent planting was more beneficial for birds. In all cases, benefits were very gradual, typically taking at least thirty years to be fully realised. Conclusions: Our study demonstrates that the cost-effectiveness of woodland creation, in terms of subsequent benefits to biodiversity, is strongly determined by its spatial configuration. Short-term biodiversity targets for woodland birds are unlikely to be well served by woodland creation, given the time required for newly planted trees to reach maturity. We recommend the use of interim targets, but emphasise that current rates of woodland creation may only result in modest changes in woodland bird populations.
- 1 April 2025
- Published by Springer Nature in Landscape Ecology
- Vol. 40 (4) , 1-19
Open AccessCombining occupancy and dispersal models to predict the conservation benefits of land-use change
Context: Economic incentives for habitat creation in productive agricultural land are an important tool for halting biodiversity loss. Given inevitable funding constraints, assessing the cost-effectiveness of potential habitat creation strategies should be a priority. Objectives: We make model predictions of changes in the populations of ten woodland-affiliated bird species over time across a set of contrasting woodland creation scenarios in a 25 × 25-km square of primarily arable land in England. Methods: Using an economic agent-based model of landowners who make economically rational decisions on land use in response to financial incentives for woodland creation, we created a suite of potential landscapes with varying spatial configurations (and amounts) of newly planted woodland. We then parameterised an individual-based model for each of ten woodland bird species, and used these models to simulate dispersal and settlement across the altered landscapes. Results: Across all bird species, population increases were greater when woodland creation plots were clumped together than when randomly sited. When woodland plots were adjacent to existing woodland, increases were comparable to those in the clumped arrangements, except at lower planting densities, where wood-adjacent planting was more beneficial for birds. In all cases, benefits were very gradual, typically taking at least thirty years to be fully realised. Conclusions: Our study demonstrates that the cost-effectiveness of woodland creation, in terms of subsequent benefits to biodiversity, is strongly determined by its spatial configuration. Short-term biodiversity targets for woodland birds are unlikely to be well served by woodland creation, given the time required for newly planted trees to reach maturity. We recommend the use of interim targets, but emphasise that current rates of woodland creation may only result in modest changes in woodland bird populations.
- Open Access
Making Messy Data Work for Conservation
- Open Access
- 29 July 2019
- Published by The Royal Society in Philosophical Transactions of the Royal Society B: Biological Sciences
- Vol. 374 (1781) , 20180053
Integrating models of human behaviour between the individual and population levels to inform conservation interventions
Conservation takes place within social–ecological systems, and many conservation interventions aim to influence human behaviour in order to push these systems towards sustainability. Predictive models of human behaviour are potentially powerful tools to support these interventions. This is particularly true if the models can link the attributes and behaviour of individuals with the dynamics of the social and environmental systems within which they operate. Here we explore this potential by showing how combining two modelling approaches (social network analysis, SNA, and agent-based modelling, ABM) could lead to more robust insights into a particular type of conservation intervention. We use our simple model, which simulates knowledge of ranger patrols through a hunting community and is based on empirical data from a Cambodian protected area, to highlight the complex, context-dependent nature of outcomes of information-sharing interventions, depending both on the configuration of the network and the attributes of the agents. We conclude by reflecting that both SNA and ABM, and many other modelling tools, are still too compartmentalized in application, either in ecology or social science, despite the strong methodological and conceptual parallels between their uses in different disciplines. Even a greater sharing of methods between disciplines is insufficient, however; given the impact of conservation on both the social and ecological aspects of systems (and vice versa), a fully integrated approach is needed, combining both the modelling approaches and the disciplinary insights of ecology and social science. This article is part of the theme issue ‘Linking behaviour to dynamics of populations and communities: application of novel approaches in behavioural ecology to conservation’.
- 29 July 2019
- Published by The Royal Society in Philosophical Transactions of the Royal Society B: Biological Sciences
- Vol. 374 (1781) , 20180053
Open AccessIntegrating models of human behaviour between the individual and population levels to inform conservation interventions
Conservation takes place within social–ecological systems, and many conservation interventions aim to influence human behaviour in order to push these systems towards sustainability. Predictive models of human behaviour are potentially powerful tools to support these interventions. This is particularly true if the models can link the attributes and behaviour of individuals with the dynamics of the social and environmental systems within which they operate. Here we explore this potential by showing how combining two modelling approaches (social network analysis, SNA, and agent-based modelling, ABM) could lead to more robust insights into a particular type of conservation intervention. We use our simple model, which simulates knowledge of ranger patrols through a hunting community and is based on empirical data from a Cambodian protected area, to highlight the complex, context-dependent nature of outcomes of information-sharing interventions, depending both on the configuration of the network and the attributes of the agents. We conclude by reflecting that both SNA and ABM, and many other modelling tools, are still too compartmentalized in application, either in ecology or social science, despite the strong methodological and conceptual parallels between their uses in different disciplines. Even a greater sharing of methods between disciplines is insufficient, however; given the impact of conservation on both the social and ecological aspects of systems (and vice versa), a fully integrated approach is needed, combining both the modelling approaches and the disciplinary insights of ecology and social science. This article is part of the theme issue ‘Linking behaviour to dynamics of populations and communities: application of novel approaches in behavioural ecology to conservation’.
- Open Access
- 8 June 2019
- Published by Springer Nature in Human Ecology
- Vol. 47 (3) , 449-464
A Framework for Assessing Impacts of Wild Meat Hunting Practices in the Tropics
- 8 June 2019
- Published by Springer Nature in Human Ecology
- Vol. 47 (3) , 449-464
Open AccessA Framework for Assessing Impacts of Wild Meat Hunting Practices in the Tropics
- Open Access
- 20 September 2018
- Published by Wiley in Conservation Biology
- Vol. 33 (3) , 665-675
Detecting deterrence from patrol data
The threat posed to protected areas by the illegal killing of wildlife is countered principally by ranger patrols that aim to detect and deter potential offenders. Deterring poaching is a fundamental conservation objective, but its achievement is difficult to identify, especially when the prime source of information comes in the form of the patrols’ own records, which inevitably contain biases. The most common metric of deterrence is a plot of illegal activities detected per unit of patrol effort (CPUE) against patrol effort (CPUE‐E). We devised a simple, mechanistic model of law breaking and law enforcement in which we simulated deterrence alongside exogenous changes in the frequency of offences under different temporal patterns of enforcement effort. The CPUE‐E plots were not reliable indicators of deterrence. However, plots of change in CPUE over change in effort (ΔCPUE‐ΔE) reliably identified deterrence, regardless of the temporal distribution of effort or any exogenous change in illegal activity levels as long as the time lag between patrol effort and subsequent behavioral change among offenders was approximately known. The ΔCPUE‐ΔE plots offered a robust, simple metric for monitoring patrol effectiveness; were no more conceptually complicated than the basic CPUE‐E plots; and required no specialist knowledge or software to produce. Our findings demonstrate the need to account for temporal autocorrelation in patrol data and to consider appropriate (and poaching‐activity‐specific) intervals for aggregation. They also reveal important gaps in understanding of deterrence in this context, especially the mechanisms by which it occurs. In practical applications, we recommend the use of ΔCPUE‐ΔE plots in preference to other basic metrics and advise that deterrence should be suspected only if there is a clear negative slope. Distinct types of illegal activity should not be grouped together for analysis, especially if the signs of their occurrence have different persistence times in the environment.
- 20 September 2018
- Published by Wiley in Conservation Biology
- Vol. 33 (3) , 665-675
Open AccessDetecting deterrence from patrol data
The threat posed to protected areas by the illegal killing of wildlife is countered principally by ranger patrols that aim to detect and deter potential offenders. Deterring poaching is a fundamental conservation objective, but its achievement is difficult to identify, especially when the prime source of information comes in the form of the patrols’ own records, which inevitably contain biases. The most common metric of deterrence is a plot of illegal activities detected per unit of patrol effort (CPUE) against patrol effort (CPUE‐E). We devised a simple, mechanistic model of law breaking and law enforcement in which we simulated deterrence alongside exogenous changes in the frequency of offences under different temporal patterns of enforcement effort. The CPUE‐E plots were not reliable indicators of deterrence. However, plots of change in CPUE over change in effort (ΔCPUE‐ΔE) reliably identified deterrence, regardless of the temporal distribution of effort or any exogenous change in illegal activity levels as long as the time lag between patrol effort and subsequent behavioral change among offenders was approximately known. The ΔCPUE‐ΔE plots offered a robust, simple metric for monitoring patrol effectiveness; were no more conceptually complicated than the basic CPUE‐E plots; and required no specialist knowledge or software to produce. Our findings demonstrate the need to account for temporal autocorrelation in patrol data and to consider appropriate (and poaching‐activity‐specific) intervals for aggregation. They also reveal important gaps in understanding of deterrence in this context, especially the mechanisms by which it occurs. In practical applications, we recommend the use of ΔCPUE‐ΔE plots in preference to other basic metrics and advise that deterrence should be suspected only if there is a clear negative slope. Distinct types of illegal activity should not be grouped together for analysis, especially if the signs of their occurrence have different persistence times in the environment.
- Top Cited PapersOpen Access
- 24 April 2017
- Published by The Royal Society in Philosophical Transactions of the Royal Society B: Biological Sciences
- Vol. 372 (1722) , 20160117
Lyme disease ecology in a changing world: consensus, uncertainty and critical gaps for improving control
Lyme disease is the most common tick-borne disease in temperate regions of North America, Europe and Asia, and the number of reported cases has increased in many regions as landscapes have been altered. Although there has been extensive work on the ecology and epidemiology of this disease in both Europe and North America, substantial uncertainty exists about fundamental aspects that determine spatial and temporal variation in both disease risk and human incidence, which hamper effective and efficient prevention and control. Here we describe areas of consensus that can be built on, identify areas of uncertainty and outline research needed to fill these gaps to facilitate predictive models of disease risk and the development of novel disease control strategies. Key areas of uncertainty include: (i) the precise influence of deer abundance on tick abundance, (ii) how tick populations are regulated, (iii) assembly of host communities and tick-feeding patterns across different habitats, (iv) reservoir competence of host species, and (v) pathogenicity for humans of different genotypes of Borrelia burgdorferi . Filling these knowledge gaps will improve Lyme disease prevention and control and provide general insights into the drivers and dynamics of this emblematic multi-host–vector-borne zoonotic disease. This article is part of the themed issue ‘Conservation, biodiversity and infectious disease: scientific evidence and policy implications'.
- 24 April 2017
- Published by The Royal Society in Philosophical Transactions of the Royal Society B: Biological Sciences
- Vol. 372 (1722) , 20160117
Top Cited PapersOpen AccessLyme disease ecology in a changing world: consensus, uncertainty and critical gaps for improving control
Lyme disease is the most common tick-borne disease in temperate regions of North America, Europe and Asia, and the number of reported cases has increased in many regions as landscapes have been altered. Although there has been extensive work on the ecology and epidemiology of this disease in both Europe and North America, substantial uncertainty exists about fundamental aspects that determine spatial and temporal variation in both disease risk and human incidence, which hamper effective and efficient prevention and control. Here we describe areas of consensus that can be built on, identify areas of uncertainty and outline research needed to fill these gaps to facilitate predictive models of disease risk and the development of novel disease control strategies. Key areas of uncertainty include: (i) the precise influence of deer abundance on tick abundance, (ii) how tick populations are regulated, (iii) assembly of host communities and tick-feeding patterns across different habitats, (iv) reservoir competence of host species, and (v) pathogenicity for humans of different genotypes of Borrelia burgdorferi . Filling these knowledge gaps will improve Lyme disease prevention and control and provide general insights into the drivers and dynamics of this emblematic multi-host–vector-borne zoonotic disease. This article is part of the themed issue ‘Conservation, biodiversity and infectious disease: scientific evidence and policy implications'.
- Open Access
- 1 April 2016
- Published by University of Chicago Press in The American Naturalist
- Vol. 187 (4) , 405-422
Epidemiological Implications of Host Biodiversity and Vector Biology: Key Insights from Simple Models
Models used to investigate the relationship between biodiversity change and vector-borne disease risk often do not explicitly include the vector; they instead rely on a frequency-dependent transmission function to represent vector dynamics. However, differences between classes of vector (e.g., ticks and insects) can cause discrepancies in epidemiological responses to environmental change. Using a pair of disease models (mosquito- and tick-borne), we simulated substitutive and additive biodiversity change (where noncompetent hosts replaced or were added to competent hosts, respectively), while considering different relationships between vector and host densities. We found important differences between classes of vector, including an increased likelihood of amplified disease risk under additive biodiversity change in mosquito models, driven by higher vector biting rates. We also draw attention to more general phenomena, such as a negative relationship between initial infection prevalence in vectors and likelihood of dilution, and the potential for a rise in density of infected vectors to occur simultaneously with a decline in proportion of infected hosts. This has important implications; the density of infected vectors is the most valid metric for primarily zoonotic infections, while the proportion of infected hosts is more relevant for infections where humans are a primary host.
- 1 April 2016
- Published by University of Chicago Press in The American Naturalist
- Vol. 187 (4) , 405-422
Open AccessEpidemiological Implications of Host Biodiversity and Vector Biology: Key Insights from Simple Models
Models used to investigate the relationship between biodiversity change and vector-borne disease risk often do not explicitly include the vector; they instead rely on a frequency-dependent transmission function to represent vector dynamics. However, differences between classes of vector (e.g., ticks and insects) can cause discrepancies in epidemiological responses to environmental change. Using a pair of disease models (mosquito- and tick-borne), we simulated substitutive and additive biodiversity change (where noncompetent hosts replaced or were added to competent hosts, respectively), while considering different relationships between vector and host densities. We found important differences between classes of vector, including an increased likelihood of amplified disease risk under additive biodiversity change in mosquito models, driven by higher vector biting rates. We also draw attention to more general phenomena, such as a negative relationship between initial infection prevalence in vectors and likelihood of dilution, and the potential for a rise in density of infected vectors to occur simultaneously with a decline in proportion of infected hosts. This has important implications; the density of infected vectors is the most valid metric for primarily zoonotic infections, while the proportion of infected hosts is more relevant for infections where humans are a primary host.
- Open Access
- 24 February 2015
- Published by Public Library of Science (PLoS) in PLOS Neglected Tropical Diseases
- Vol. 9 (2) , e0003424
Rapid Diagnostic Tests for Dengue Virus Infection in Febrile Cambodian Children: Diagnostic Accuracy and Incorporation into Diagnostic Algorithms
Dengue virus (DENV) infection is prevalent across tropical regions and may cause severe disease. Early diagnosis may improve supportive care. We prospectively assessed the Standard Diagnostics (Korea) BIOLINE Dengue Duo DENV rapid diagnostic test (RDT) to NS1 antigen and anti-DENV IgM (NS1 and IgM) in children in Cambodia, with the aim of improving the diagnosis of DENV infection. We enrolled children admitted to hospital with non-localised febrile illnesses during the 5-month DENV transmission season. Clinical and laboratory variables, and DENV RDT results were recorded at admission. Children had blood culture and serological and molecular tests for common local pathogens, including reference laboratory DENV NS1 antigen and IgM assays. 337 children were admitted with non-localised febrile illness over 5 months. 71 (21%) had DENV infection (reference assay positive). Sensitivity was 58%, and specificity 85% for RDT NS1 and IgM combined. Conditional inference framework analysis showed the additional value of platelet and white cell counts for diagnosis of DENV infection. Variables associated with diagnosis of DENV infection were not associated with critical care admission (70 children, 21%) or mortality (19 children, 6%). Known causes of mortality were melioidosis (4), other sepsis (5), and malignancy (1). 22 (27%) children with a positive DENV RDT had a treatable other infection. The DENV RDT had low sensitivity for the diagnosis of DENV infection. The high co-prevalence of infections in our cohort indicates the need for a broad microbiological assessment of non-localised febrile illness in these children. DENV infection first manifests as an undifferentiated fever before either settling without complications, or progressing to severe disease requiring inpatient admission and careful supportive intravenous fluid management. The ability to differentiate DENV infection from other febrile illnesses, and to predict those at risk of severe disease is likely to be important. We assessed the diagnostic accuracy of a commercially available DENV rapid diagnostic test (RDT) for children admitted with febrile illness to a hospital in Cambodia during the DENV transmission season. We found sensitivity of the DENV RDT to be 58% and specificity to be 85% versus reference assay DENV serology. We then modelled the ability of clinical features, basic laboratory parameters, and DENV RDT result at presentation of the child to distinguish DENV infection from other febrile illness, and determine the need for critical care admission. We found that the DENV RDT did not increase the accuracy with which we diagnosed DENV infection, and was not helpful in deciding which children required critical care admission. Indeed, the relatively high prevalence of serious bacterial disease in the cohort of children indicated a broad microbiological differential diagnosis in all febrile children, regardless of their DENV infection status.
- 24 February 2015
- Published by Public Library of Science (PLoS) in PLOS Neglected Tropical Diseases
- Vol. 9 (2) , e0003424
Open AccessRapid Diagnostic Tests for Dengue Virus Infection in Febrile Cambodian Children: Diagnostic Accuracy and Incorporation into Diagnostic Algorithms
Dengue virus (DENV) infection is prevalent across tropical regions and may cause severe disease. Early diagnosis may improve supportive care. We prospectively assessed the Standard Diagnostics (Korea) BIOLINE Dengue Duo DENV rapid diagnostic test (RDT) to NS1 antigen and anti-DENV IgM (NS1 and IgM) in children in Cambodia, with the aim of improving the diagnosis of DENV infection. We enrolled children admitted to hospital with non-localised febrile illnesses during the 5-month DENV transmission season. Clinical and laboratory variables, and DENV RDT results were recorded at admission. Children had blood culture and serological and molecular tests for common local pathogens, including reference laboratory DENV NS1 antigen and IgM assays. 337 children were admitted with non-localised febrile illness over 5 months. 71 (21%) had DENV infection (reference assay positive). Sensitivity was 58%, and specificity 85% for RDT NS1 and IgM combined. Conditional inference framework analysis showed the additional value of platelet and white cell counts for diagnosis of DENV infection. Variables associated with diagnosis of DENV infection were not associated with critical care admission (70 children, 21%) or mortality (19 children, 6%). Known causes of mortality were melioidosis (4), other sepsis (5), and malignancy (1). 22 (27%) children with a positive DENV RDT had a treatable other infection. The DENV RDT had low sensitivity for the diagnosis of DENV infection. The high co-prevalence of infections in our cohort indicates the need for a broad microbiological assessment of non-localised febrile illness in these children. DENV infection first manifests as an undifferentiated fever before either settling without complications, or progressing to severe disease requiring inpatient admission and careful supportive intravenous fluid management. The ability to differentiate DENV infection from other febrile illnesses, and to predict those at risk of severe disease is likely to be important. We assessed the diagnostic accuracy of a commercially available DENV rapid diagnostic test (RDT) for children admitted with febrile illness to a hospital in Cambodia during the DENV transmission season. We found sensitivity of the DENV RDT to be 58% and specificity to be 85% versus reference assay DENV serology. We then modelled the ability of clinical features, basic laboratory parameters, and DENV RDT result at presentation of the child to distinguish DENV infection from other febrile illness, and determine the need for critical care admission. We found that the DENV RDT did not increase the accuracy with which we diagnosed DENV infection, and was not helpful in deciding which children required critical care admission. Indeed, the relatively high prevalence of serious bacterial disease in the cohort of children indicated a broad microbiological differential diagnosis in all febrile children, regardless of their DENV infection status.
- Open Access
- 1 January 2014
- Published by Springer Nature in Parasites & Vectors
- Vol. 7 (1) , 231
History and complexity in tick-host dynamics: discrepancies between ‘real’ and ‘visible’ tick populations
Numerical responses of ticks to changes in densities of their hosts can be complex and apparently unpredictable. Manipulations even of deterministic models can produce counter-intuitive results, including tick populations that either rise or fall under increasing host densities, depending on initial conditions.
- 1 January 2014
- Published by Springer Nature in Parasites & Vectors
- Vol. 7 (1) , 231
Open AccessHistory and complexity in tick-host dynamics: discrepancies between ‘real’ and ‘visible’ tick populations
Numerical responses of ticks to changes in densities of their hosts can be complex and apparently unpredictable. Manipulations even of deterministic models can produce counter-intuitive results, including tick populations that either rise or fall under increasing host densities, depending on initial conditions.
- Open Access
- 10 December 2013
- Published by Springer Nature in Parasites & Vectors
- Vol. 6 (1) , 344
Ticks in the wrong boxes: assessing error in blanket-drag studies due to occasional sampling
The risk posed by ticks as vectors of disease is typically assessed by blanket-drag sampling of host-seeking individuals. Comparisons of peak abundance between plots – either in order to establish their relative risk or to identify environmental correlates – are often carried out by sampling on one or two occasions during the period of assumed peak tick activity.
- 10 December 2013
- Published by Springer Nature in Parasites & Vectors
- Vol. 6 (1) , 344
Open AccessTicks in the wrong boxes: assessing error in blanket-drag studies due to occasional sampling
The risk posed by ticks as vectors of disease is typically assessed by blanket-drag sampling of host-seeking individuals. Comparisons of peak abundance between plots – either in order to establish their relative risk or to identify environmental correlates – are often carried out by sampling on one or two occasions during the period of assumed peak tick activity.