Infill-sampling design and the Cost of classification errors

Aspie, David; Barnes, Randal J.

doi:10.1007/bf00890117

Cited by 44 publications

(10 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The relation between sampling density/location and error has been addressed in other environmental measurements (Aspie and Barnes, 1990;Williams, 1996;Mac Nally, 1997;Pelto, 2000;Ladson et al, 2006). A productive approach used by some (e.g., Williams, 1996;Mac Nally, 1997;Ladson et al, 2006) and applied here involves ''over-sampling'' (i.e., making field measurements at a density that is thought likely to be greater than necessary for accurate definition of the mean or spatial field of interest) to create a dense dataset of point values, and then selecting subsets of different sizes from the full dataset to evaluate how means and spatial fields based on these subsets differ from each other and from those based on the full dataset.…”

Section: Introductionmentioning

confidence: 99%

Effect of sampling density and design on estimation of streambed attributes

Kennedy

Genereux

Mitášová

et al. 2008

Journal of Hydrology

View full text Add to dashboard Cite

Section: Introductionmentioning

confidence: 99%

Effect of sampling density and design on estimation of streambed attributes

Kennedy

Genereux

Mitášová

et al. 2008

Journal of Hydrology

View full text Add to dashboard Cite

“…In order to contradict d ij ≤ d ik + d kj it must be shown that conditions may exist in which d ij > d ik + d jk . Aspie and Barnes (1990) discussed that prediction errors are determined only by the sample locations and the variogram, which are known, and not on the values of the observed random variables Z(s). To disprove Axiom [4], let the following conditions exist:…”

Section: Dissimilarity Coefficientmentioning

confidence: 99%

Arsenic in Michigan Soil: Spatial Stratification for Environmental Regulatory Purposes

Daniels

Kinnicutt

2009

Soil and Sediment Contamination: An International Journal

View full text Add to dashboard Cite

State environmental regulatory agencies in the U.S. often establish a default background standard for naturally occurring elements in the soil, water, and air. The background standard is determined and then used as a benchmark across the entire jurisdiction. A variety of statistical techniques are used to determine this standard, but often ignore any inherent spatial dependencies within the jurisdiction. If the analysis indicates a specific site exceeds the default standard, additional background sampling and analysis must usually be performed. Frequently, this additional sampling is found to be unnecessary simply because the natural background levels were elevated for this particular site. Conversely, potential contamination may be overlooked in areas where the natural background levels are much lower. Thus, a single default background standard seems inadequate within this context. This paper proposes the use of dissimilarity coefficients based on kriging estimates as a means to regionalize background standards. Along with cluster analysis techniques, these dissimilarity coefficients provide a means to stratify the population into geographic sub-areas. A regulatory agency may now define multiple default background standards based on geographic location. To illustrate, this paper examines a case study concerning residential soil arsenic for 83 Michigan counties.

show abstract

“…Because of the combination of all the potential sampling configurations (choices of S 1 and S 2 ), an exhaustive search of the optimal design is precluded, so that one has to look for a suboptimal approach to solve the sampling design problem. Possible approaches include forward-selection algorithms, in which sampling locations are added sequentially until the required constraints are fulfilled (Lu et al 2000), sequential exchange algorithms (Aspie and Barnes 1990), or Bayesian search theory (Freeze et al 1992;James and Gorelick 1994). Another approach to determine S 1 and S 2 is the recourse to a simulated annealing (SA) algorithm.…”

Section: Sampling Design Algorithm: Simulated Annealingmentioning

confidence: 99%

A geostatistical approach to optimize sampling designs for local forest inventories

Hernández

Emery

2009

Can. J. For. Res.

View full text Add to dashboard Cite

In forest management, it is of interest to obtain detailed inventories such that the local prediction errors on forest attributes are less than a prespecified threshold, while keeping the number of ground samples as low as possible. Given an initial sampling design, we propose an algorithm to determine the additional sample locations. The algorithm relies on two tools: geostatistical simulation, which allows measuring the uncertainty in the values of the attribute of interest, and simulated annealing, which allows finding an infill design that minimizes a given objective function. The proposed approach is applied to a data set from a Prosopis spp. plantation located in the Atacama Desert, in which the measured attribute is the rate of tree survival.Résumé : En gestion des forêts, il est intéressant d'obtenir des inventaires détaillés tels que les erreurs de prédiction locale sur les variables forestières soient inférieures à un seuil prédéfini, tout en réduisant le plus possible le nombre d'échan-tillons de terrain. É tant donné un plan d'échantillonnage initial, nous proposons un algorithme pour déterminer l'emplacement d'échantillons supplémentaires. L'algorithme repose sur deux outils : la simulation géostatistique, qui permet de mesurer l'incertitude sur les valeurs de la variable étudiée, et le recuit simulé, qui permet de trouver un plan d'échantil-lonnage minimisant une fonction objectif donnée. L'approche proposée est appliquée à un jeu de données d'une plantation de Prosopis située dans le désert d'Atacama, où la variable mesurée est le taux de survie des arbres.

show abstract

Infill-sampling design and the Cost of classification errors

Cited by 44 publications

References 10 publications

Effect of sampling density and design on estimation of streambed attributes

Effect of sampling density and design on estimation of streambed attributes

Arsenic in Michigan Soil: Spatial Stratification for Environmental Regulatory Purposes

A geostatistical approach to optimize sampling designs for local forest inventories

Contact Info

Product

Resources

About