Sampling for Natural Resource Monitoring

Abstract

``Can you give me a reference to a book on monitoring?'' is the kind of questionwe are frequently asked by colleagues who are planning to design a monitoringscheme. Although this is a perfectly reasonable question, we became increasinglyuneasy, because we had to admit that, to our knowledge, a book meetingthese researchers' needs did not exist. More frustration built up as monitoringprojects grew in number, diversity and importance in our research environment.Therefore, after much hesitation, we decided to try and write such abook ourselves. But why were we not satisfied with the existing literature? Letus spell this out in some detail, because this explains largely why we wrotethis book the way we did.There are several books and chapters of books dealing with the subjectof monitoring, i.e., sampling in space, time or space--time. These books focuson applications within a particular discipline, for instance hydrology orvegetation science, and of course treat only the methodology relevant to thatdiscipline. In doing so, they normally use the scientific jargon common to thatfield. However, scientists working in other fields may need different monitoringmethods and may also profit from a more generic presentation. Furthermore,we disagree with some of the existing texts on statistical grounds. Our criticismrelates to misconceptions about the differences between the design-basedand the model-based approach to sampling.Broadly speaking, monitoring draws on three distinct statistical methodologies:sampling theory, time-series analysis and geostatistics. Many booksare available on each of these fields, but they lack the methodological scopethat is generally needed in monitoring. Furthermore, they rarely give enoughhelp to practitioners who want to know how to make a proper selection ofmethods from these fields, and how to integrate them into a monitoringscheme. Books on sampling theory are usually written in the abstract andgeneric terms of sampling from finite populations, while it is not always clearhow these methods should be adapted and applied to continuous space, timeor space--time. These books are usually written for statisticians. There are severalgood, practical handbooks on geostatistics, but they do not cover classicalVIII Prefacesampling methodology, nor sampling in time, whereas books on time-seriesanalysis fail to discuss classical sampling methodology and sampling in space.There is a host of research articles on statistical methods for monitoring,scattered over a wide range of journals and proceedings, many of which arehighly specialized or theoretical. It would hardly be possible for a practitionerto extract the relevant knowledge from this type of literature alone. Reviewarticles come closer to our purpose, but limitations of space prevent themfrom offering full coverage of methods, and they often focus on a particularfield of application. Thus we identified a need for a handbook on the statisticalmethodology of monitoring that gives applied scientists sufficient guidance inhow to design a monitoring scheme, rather than presenting a large collectionof methods.Although we focus on practitioners in the field of natural resource monitoringrather than statisticians, basic statistical knowledge is required for aproper understanding of the methodologies described. If one decides, on thebasis of this book, to apply an advanced method of which the theoreticalbasis is not fully fathomed, we advise to consult an applied statistician. Ingeneral, such a consult is useful in the very beginning of a monitoring project.This book may offer the kind of knowledge helpful in an efficient and fruitfulcommunication on statistical aspects of monitoring.The following persons commented on an earlier version of the manuscript:Prof. Jelke Bethlehem (Statistics Netherlands), Dr. Frans van Geer (NetherlandsInstitute of Applied Geoscience TNO), Dr. Gerard Heuvelink (WageningenUniversity, Soil Science Centre), Prof. Alex McBratney (The Universityof Sydney, Agricultural Chemistry and Soil Science Dept.), Dr. Tony Olsen(US Environmental Protection Agency'sWestern Ecology Division), and Prof.Scott Urquhart (Colorado State University, Statistics Dept.). We are gratefulfor their constructive criticism and suggestions. Of course, all shortcomingsof this book remain the responsibility of the authors. We also thank MartinJansen (Alterra) for helping us to produce the figures, and Wybo Dekker(Dutch TeX Users Group) for assistance with LATEX.Financing was received from the Dutch Ministry of Agriculture, NatureManagement and Fisheries via research programs 317, 325, 328, 329, 358, 382,394 and 395, and by Alterra (Wageningen University and Research Centre).Last but not least we thank all colleagues who presented their samplingproblems to us in such a great variety, and who were patient enough to explainwhy our initial perceptions and ideas were too simplistic. This helped us muchto deepen and widen our understanding of sampling.