Abstract

This thesis focuses on the efficiency of using sampling methods to derive landscape metrics. It also explores what sampling methods are to be preferred for different metrics and how metrics in some cases can be redefined to better suit a sample-based data collection framework. In paper I, a review was conducted to assess previous research in the area of sample-based assessment of landscape metrics. It was found that only rather few studies have been conducted, but these indicate that data acquisition through sampling appears to be a promising alternative to traditional wall-to-wall mapping. In papers II and III, point and line intersect sampling (LIS) methods were used to estimate the metrics Shannon’s diversity and edge density. Monte-Carlo simulation was employed to investigate the statistical properties in terms of bias and root mean square error (RMSE) of the metrics estimators for different sampling designs. Further, the cost (time needed) of data collection using wall-to-wall mapping and sampling was studied. Both bias and RMSE decreased with increasing sample size, to magnitudes small enough to make sampling a competitive alternative to wall-to-wall mapping. As is commonly the case in sampling, systematic designs were found to be superior to simple random designs. In the case of LIS, longer line transects were superior to short ones and a straight line was more efficient than the other configurations considered. Papers IV and V address the contagion metric. In paper IV a new definition of the metric for vector data was developed. The definition is distance dependent and also forms a basis for estimating the contagion metric from point sampling data. It was found that a simple negative exponential function could be used as a good proxy function for the unconditional contagion while no such proxy function was found for the conditional contagion metric. The proxy function for the unconditional contagion was found to be strongly related to the area proportion of different land cover types (Shannon’s diversity) and to the rate of change of the contagion value over different distances. In paper V sampling simulation was performed to evaluate the properties of estimators for different point sampling designs and distances between point pairs. For the unconditional contagion, the sizes of bias and RMSE were fairly small for sample sizes that could be expected in practice, while the conditional contagion was found to require large sample sizes or otherwise the accuracy of the estimates would be poor. A general conclusion from the studies are that sample-based approaches to landscape metrics estimation are promising for several, but not all, of the metrics commonly applied in landscape ecology. Further, by slightly redefining the definitions for some metrics, it is possible to make them better suited for a sample-based data acquisition framework.

Keywords

sampling; methods; landscape; cartography; simulation

Published in

Acta Universitatis Agriculturae Sueciae
2010, number: 2010:74
ISBN: 978-91-576-7519-4
Publisher: Department of Forest Resource Management, Swedish University of Agricultural Sciences

SLU Authors

• Ramezani, Habib

  • Department of Forest Resource Management, Swedish University of Agricultural Sciences
    

UKÄ Subject classification

Forest Science


Permanent link to this page (URI)

https://res.slu.se/id/publ/31101