Because it is generally not possible to monitor an entire ecosystem, or an entire population, sampling is essential. But the theory is complex and many environmental specialists are not trained in it.  As a result, many programs suffer from a lack of definition in the problematic, in the formulation of hypotheses, in the elaboration of an adapted sampling plan (design) with, as an immediate consequence, a questionable quality of the data. Inadequate data processing and the use of incorrectly specified models without understanding the underlying theory contribute to the questionability of the results. A bad method has many cascading side effects that can lead to program failure. It is therefore essential to know, on the one hand, some essential theoretical aspects of sampling and data modeling and, on the other hand, some reproducible and robust methods that practitioners can use. The key is to be at least aware of several critical aspects such as the definition of the variables of interest to be monitored, sample size, bias, precision, and the trade-off between sufficient sample size and reasonable financial cost to funders.

The course is divided into two mainstream parts:

• Building survey designs that will help ensure environmental monitoring programmes deliver data with characteristics sticking to rigorous scientific process, involving survey data be trustworthy and fit-for-purpose. Surveys and monitoring programmes must be designed and implemented in such a way that the resulting data are: (i) « representative » of the population under investigation and (ii) information rich so that uncertainty around inferences is reduced as much as survey budgets will allow. Several packages exist for generating spatially balanced designs (Kermorvant et al., 2019) ; they provide a good platform for generating spatially balanced designs, but they have different foci in terms of algorithms, functionality, scope, computation requirements and user interface. Students will use main packages (spsurvey, SDraw…).
• Assessing species distribution and abundance are important goals of a wide range of fields including wildlife research and management. Imperfect detection of individual animals and plants is a ubiquitous source of bias in these assessments. To correct the resulting errors, ecologists and statisticians have proposed a suite of hierarchical models which separate the state process (e.g. occurrence or abundance) and the observation, or detection, process. These models typically require field protocols where a site is surveyed repeatedly, allowing the estimation of probability of detection. Occupancy modelling (MacKenzie et al., 2002; Tyre et al., 2003) is among the most widely used of these hierarchical approaches and will be developped in-depth. Similar hierarchical modelling approaches will be applied to estimate abundance from repeated counts (Royle, 2004) using N-mixture models. Multiple software tools have been developed to facilitate fitting occupancy and abundance models and estimating covariate effects. Students will use the pioneer program PRESENCE (stand-alone software that can be used to fit various model types using maximum likelihood) but emphasis will be on R programming language which has greatly increased among ecologists and wildlife biologists using the specialized R package unmarked (which also uses maximum likelihood; Fiske & Chandler, 2011).

Students in this course will develop the ability to:

·         Develop practical skills in designing a basic sampling plan (emphasis on spatial aspects).  

·         Critique and develop research and biomonitoring methodologies used by freshwater ecologists

• Explain the terminology and the underlying theory: "design-based" and "model-based" approaches, estimation, estimator, likelihood, bias, precision, accuracy, basics of modeling in the frequentist (vs Bayesian) framework

• Take into account the imperfection in species detection during monitoring and integrate this paradigm in hierarchical models of occupancy and abundance.

• Estimate occupancy probability and abundance, given detection probability.

Upon completing this course students will have a general overview of how to derive GLM functions and fit response models, understand the statistics used to assess model fit.

Theoretical: probability and inferential statistics, basics of modeling.  Technical: mastery of the R environment (https://www.r-project.org). R programming is required (see important notice below).

Session 1 : 100% contrôle continu écrit

Session 2 : 100% oral

Important notice:

-       Pre-requisites are courses which must be completed prior to enrollment in the subsequent course to ensure adequate preparation. Basics at Licence levels (L1, L2, L3) in ecology and biology and statistics are mandatory.  Students must understand the meaning and purpose of regression and modeling techniques at least.

-       These days R Programming language is used to many purposes as it has extensive and powerful graphics abilities, that are tightly linked with its analytic abilities. The R system is developing so rapidly that new features and abilities appear every days. So, most important pre-requisites are: i/ solid understanding of statistics in mathematics, ii/ good understanding of various type of graphs for data representation and iii/ prior knowledge of any programming.

Additional resources

• MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, L.L. Bailey, and J. E. Hines (2018). Occupancy Estimation and Modeling - Inferring Patterns and Dynamics of Species Occurrence. 2nd Edition. Elsevier Publishing. https://www.mbr-pwrc.usgs.gov/software/presence.html
• Lohr S. (2020). Sampling: Design and Analysis - 2nd edition, is now published by Chapman & Hall/CRC Press. https://www.sharonlohr.com/sampling-design-and-analysis
• Kermorvant C., Coube S., D’Amico F., Bru N & Caill-Milly N. (2020). Sequential process to choose efficient sampling design based on partial prior information data and simulations. Spatial Statistics, 38, 100439, ISSN 2211-6753, https://doi.org/10.1016/j.spasta.2020.100439.
• Kermorvant C., D’Amico F., Bru N., Caill-Milly N. & Robertson B. (2019). Spatially balanced sampling designs for environmental surveys. Environmental Monitoring and Assessment 191:524 / https://doi.org/10.1007/s10661-019-7666-y
• Fiske, I. & Chandler, R. 2011. Unmarked: An R Package for Fitting Hierarchical Models of Wildlife Occurrence and Abundance. 43:23.
• D’Amico F., Kermorvant C., Sanchez J.M. & Arizaga J. (2020). Optimal sampling design to survey riparian bird populations with low detection probability. Bird Study, DOI: 10.1080/00063657.2020.1784090