Improving pest monitoring networks using a simulation-based approach to contribute to pesticide reduction.

Conventional pest management mainly relies on the use of pesticides. However, the negative externalities of pesticides are now well known. More sustainable practices, such as Integrated Pest Management, are necessary to limit crop damage from pathogens, pests and weeds in agroecosystems. Reducing pesticide use requires information to determine whether chemical treatments are really needed. Pest monitoring networks (PMNs) are key contributors to this information. However, the effectiveness of a PMN in delivering relevant information about pests depends on its spatial sampling resolution and its memory length. The trade-off between the monitoring efforts and the usefulness of the information provided is highly dependent on pest ecological traits, the damage they can cause (in terms of crop losses), and economic drivers (production costs, agriculture product prices and incentives). Due to the high complexity of optimising PMNs, we have developed a theoretical model that belongs to the family of Dynamic Bayesian Networks in order to compare several PMNs performances. This model links the characteristics of a PMN to treatment decisions and the resulting pest dynamics. Using simulation and inference tools for graphical models, we derived the proportion of impacted fields, the number of pesticide treatments and the overall gross margins for three types of pest with contrasting levels of endocyclism. The term "endocyclic" refers to an organism whose development is mostly restricted to a field and highly depends on the inoculum present in the considered field. The presence of purely endocyclic pests at a given time increases the probability of reoccurrence. Conversely, slightly endocyclic pests have a low persistence. The simulation analysis considered ten scenarios: an expected margin-based strategy with a spatial resolution of four PMNs and two memory lengths (one year or eight years), as well as two extreme crop protection strategies (systematic treatments on all fields and systematic no treatment). For purely and mainly endocyclic pests (e.g. soil-borne pathogens and most weeds, respectively), we found that increasing the spatial resolution of PMNs made it possible to significantly decrease the number of treatments required for pest control. Taking past observations into account was also effective, but to a lesser extent. PMN information had virtually no influence on the control of non-endocyclic pests (such as flying insects or airborne plant pathogens) which may be due to the spatial coverage addressed in our study. The next step is to extend the analysis of PMNs and to integrate the information generated by PMNs into sustainable pest management strategies, both at the field and the landscape level.

A spatial Markovian framework for estimating regional and local dynamics of annual plants with dormancy.

Many species have a dormant stage in their life cycle, including seeds for plants. The dormancy stage influences the species dynamics but is often undetectable. One way to include dormancy is to model it as a hidden dynamical state within a Markovian framework. Models within this framework have already been proposed but with different limitations: only presence/absence observations are modelled, the dormancy stage is limited to one year, or colonisation from neighbouring patches is not taken into account. We propose a hidden Markov model that describes the local and regional dynamics of a species that can undergo dormancy with a potentially infinite dormancy time. Populations are modelled with abundance classes. Our model considers the colonisation process as the indistinguishable influence of neighbour non-dormant population states on a dormant population state in a patch. It would be expected that parameter estimation, hidden state estimation and prediction of the next non-dormant populations would have an exponential computational time in terms of the number of patches. However, we demonstrate that estimation, hidden state estimation and prediction are all achievable in a linear computational time. Numerical experiments on simulated data show that the state of dormant populations can easily be retrieved, as well as the state of future non-dormant populations. Our framework provides a simple and efficient tool that could be further used to analyse and compare annual plants dynamics like weed species survival strategies in crop fields.

A general method for estimating seed dormancy and colonisation in annual plants from the observation of existing flora.

In plant ecology, characterising colonisation and extinction in plant metapopulations is challenging due to the non-detectable seed bank that allows plants to emerge after several years of absence. In this study, we used a Hidden Markov Model to characterise seed dormancy, colonisation and germination solely from the presence-absence of standing flora. Applying the model to data from a long-term survey of 38 annual weeds across France, we identified three homogeneous functional groups: (1) species persisting preferentially through spatial colonisation, (2) species persisting preferentially through seed dormancy and (3) a mix of both strategies. These groups are consistent with existing ecological knowledge, demonstrating that ecologically meaningful parameters can be estimated from simple presence-absence observations. These results indicate that such studies could contribute to the design of weed management strategies. They also open the possibility of testing life-history theories such as the dormancy/colonisation trade-off in natura.

Classification method for disease risk mapping based on discrete hidden Markov random fields.

Risk mapping in epidemiology enables areas with a low or high risk of disease contamination to be localized and provides a measure of risk differences between these regions. Risk mapping models for pooled data currently used by epidemiologists focus on the estimated risk for each geographical unit. They are based on a Poisson log-linear mixed model with a latent intrinsic continuous hidden Markov random field (HMRF) generally corresponding to a Gaussian autoregressive spatial smoothing. Risk classification, which is necessary to draw clearly delimited risk zones (in which protection measures may be applied), generally must be performed separately. We propose a method for direct classified risk mapping based on a Poisson log-linear mixed model with a latent discrete HMRF. The discrete hidden field (HF) corresponds to the assignment of each spatial unit to a risk class. The risk values attached to the classes are parameters and are estimated. When mapping risk using HMRFs, the conditional distribution of the observed field is modeled with a Poisson rather than a Gaussian distribution as in image segmentation. Moreover, abrupt changes in risk levels are rare in disease maps. The spatial hidden model should favor smoothed out risks, but conventional discrete Markov random fields (e.g. the Potts model) do not impose this. We therefore propose new potential functions for the HF that take into account class ordering. We use a Monte Carlo version of the expectation-maximization algorithm to estimate parameters and determine risk classes. We illustrate the method's behavior on simulated and real data sets. Our method appears particularly well adapted to localize high-risk regions and estimate the corresponding risk levels.

Does clustering of DNA barcodes agree with botanical classification directly at high taxonomic levels? Trees in French Guiana as a case study.

Characterizing biodiversity is one of the main challenges for the coming decades. Most diversity has not been morphologically described, and barcoding is now complementing morphological-based taxonomy to further develop inventories. Both approaches have been cross-validated at the level of species and OTUs. However, many known species are not listed in reference databases. One path to speed up inventories using barcoding is to directly identify individuals at coarser taxonomic levels. We therefore studied in barcoding of plants whether morphological-based and molecular-based approaches are in agreement at genus, family and order levels. We used Agglomerative Hierarchical Clustering (with Ward, Complete and Single Linkage) and Stochastic Block Models (SBM), with two dissimilarity measures (Smith-Waterman scores, kmers). The agreement between morphological-based and molecular-based classifications ranges in most of the cases from good to very good at taxonomic levels above species, even though it decreases when taxonomic levels increase, or when using the tetramer-based distance. Agreement is correlated with the entropy of morphological-based classification and with the ratio of the mean within- and mean between-groups dissimilarities. The Ward method globally leads to the best agreement, whereas Single Linkage can show poor behaviours. SBM provides a useful tool to test whether or not the dissimilarities are structured by the botanical groups. These results suggest that automatic clustering and group identification at taxonomic levels above species are possible in barcoding.

The value of understanding feedbacks from ecosystem functions to species for managing ecosystems.

Ecological systems are made up of complex and often unknown interactions and feedbacks. Uncovering these interactions and feedbacks among species, ecosystem functions, and ecosystem services is challenging, costly, and time-consuming. Here, we ask: for which ecosystem features does resolving the uncertainty about the feedbacks from ecosystem function to species improve management outcomes? We develop a dynamic value of information analysis for risk-neutral and risk-prone managers on motif ecosystems and explore the influence of five ecological features. We find that learning the feedbacks from ecosystem function to species does not improve management outcomes for maximising biodiversity, yet learning which species benefit from an ecosystem function improves management outcomes for ecosystem services by up to 25% for risk-neutral managers and 231% for risk-prone managers. Our general approach provides useful guidance for managers and researchers on when learning feedbacks from ecosystem function to species can improve management outcomes for multiple conservation objectives.

Quantifying the impact of uncertainty on threat management for biodiversity.

With inadequate resources to manage the threats facing biodiversity worldwide, achieving projected management outcomes is critical for efficient resource allocation and species recovery. Despite this, conservation plans to mitigate threats rarely articulate the likelihood of management success. Here we develop a general value of information approach to quantify the impact of uncertainty on 20 threatening processes affecting 976 listed species and communities. To our knowledge, this is the most comprehensive quantification of the impacts of uncertainty on threat management. We discover that, on average, removing uncertainty about management effectiveness could triple the gain in persistence achieved by managing under current uncertainty. Management of fire, invasive animals and a plant pathogen are most impeded by uncertainty; management of invasive plants is least impacted. Our results emphasise the tremendous importance of reducing uncertainty about species responses to management, and show that failure to consider management effectiveness wastes resources and impedes species recovery.

GMDPtoolbox: A Matlab library for designing spatial management policies. Application to the long-term collective management of an airborne disease.

Designing management policies in ecology and agroecology is complex. Several components must be managed together while they strongly interact spatially. Decision choices must be made under uncertainty on the results of the actions and on the system dynamics. Furthermore, the objectives pursued when managing ecological systems or agroecosystems are usually long term objectives, such as biodiversity conservation or sustainable crop production. The framework of Graph-Based Markov Decision Processes (GMDP) is well adapted to the qualitative modeling of such problems of sequential decision under uncertainty. Spatial interactions are easily modeled and integrated control policies (combining several action levers) can be designed through optimization. The provided policies are adaptive, meaning that management actions are decided at each time step (for instance yearly) and the chosen actions depend on the current system state. This framework has already been successfully applied to forest management and invasive species management. However, up to now, no "easy-to-use" implementation of this framework was available. We present GMDPtoolbox, a Matlab toolbox which can be used both for the design of new management policies and for comparing policies by simulation. We provide an illustration of the use of the toolbox on a realistic crop disease management problem: the design of long term management policy of blackleg of canola using an optimal combination of three possible cultural levers. This example shows how GMDPtoolbox can be used as a tool to support expert thinking.

Investigating Disease Spread between Two Assessment Dates with Permutation Tests on a Lattice.

ABSTRACT Mapping and analyzing the disease status of individual plants within a study area at successive dates can give insight into the processes involved in the spread of a disease. We propose a permutation method to analyze such spatiotemporal maps of binary data (healthy or diseased plants) in regularly spaced plantings. It requires little prior information on the causes of disease spread and handles missing plants and censored data. A Monte Carlo test is used to assess whether the location of newly diseased plants is independent of the location of previously diseased plants. The test takes account of the significant spatial structures at each date in order to separate nonrandomness caused by the structure at one date from nonrandomness caused by the dependence between newly diseased plants and previously diseased plants. If there is a nonrandom structure at both dates, independent patterns are simulated by randomly shifting the entire pattern observed at the second date. Otherwise, independent patterns are simulated by randomly reallocating the positions of one group of diseased plants. Simulated and observed patterns of disease are then compared through distance-based statistics. The performance of the method and its robustness are evaluated by its ability to accurately identify simulated independent and dependent bivariate point patterns. Additionally, two realworld spatiotemporal maps with contrasting disease progress illustrate how the tests can provide valuable clues about the processes of disease spread. This method can supplement biological investigations and be used as an exploratory step before developing a specific mechanistic model.

Dynamics of Weeds in the Soil Seed Bank: A Hidden Markov Model to Estimate Life History Traits from Standing Plant Time Series.

Predicting the population dynamics of annual plants is a challenge due to their hidden seed banks in the field. However, such predictions are highly valuable for determining management strategies, specifically in agricultural landscapes. In agroecosystems, most weed seeds survive during unfavourable seasons and persist for several years in the seed bank. This causes difficulties in making accurate predictions of weed population dynamics and life history traits (LHT). Consequently, it is very difficult to identify management strategies that limit both weed populations and species diversity. In this article, we present a method of assessing weed population dynamics from both standing plant time series data and an unknown seed bank. We use a Hidden Markov Model (HMM) to obtain estimates of over 3,080 botanical records for three major LHT: seed survival in the soil, plant establishment (including post-emergence mortality), and seed production of 18 common weed species. Maximum likelihood and Bayesian approaches were complementarily used to estimate LHT values. The results showed that the LHT provided by the HMM enabled fairly accurate estimates of weed populations in different crops. There was a positive correlation between estimated germination rates and an index of the specialisation to the crop type (IndVal). The relationships between estimated LHTs and that between the estimated LHTs and the ecological characteristics of weeds provided insights into weed strategies. For example, a common strategy to cope with agricultural practices in several weeds was to produce less seeds and increase germination rates. This knowledge, especially of LHT for each type of crop, should provide valuable information for developing sustainable weed management strategies.

Model-based region-of-interest selection in dynamic breast MRI.

Magnetic resonance imaging (MRI) is emerging as a powerful tool for the diagnosis of breast abnormalities. Dynamic analysis of the temporal pattern of contrast uptake has been applied in differential diagnosis of benign and malignant lesions to improve specificity. Selecting a region of interest (ROI) is an almost universal step in the process of examining the contrast uptake characteristics of a breast lesion. We propose an ROI selection method that combines model-based clustering of the pixels with Bayesian morphology, a new statistical image segmentation method. We then investigate tools for subsequent analysis of signal intensity time course data in the selected region. Results on a database of 19 patients indicate that the method provides informative segmentations and good detection rates.