Statistical mapping of tree species over Europe.

Brus, D.J. and Hengeveld, G.M. and Walvoort, D.J.J. and Goedhart, P.W. and Heidema, A.H. and Nabuurs, G.J. and Gunia, K. (2012) Statistical mapping of tree species over Europe. European Journal of Forest Research, 131 (1). pp. 145-157.

[img] PDF - Published Version
Restricted to ICRISAT researchers only


In order to map the spatial distribution of twenty tree species groups over Europe at 1 km × 1 km resolution, the ICP-Forest Level-I plot data were extended with the National Forest Inventory (NFI) plot data of eighteen countries. The NFI grids have a much smaller spacing than the ICP grid. In areas with NFI plot data, the proportions of the land area covered by the tree species were mapped by compositional kriging. Outside these areas, these proportions were mapped with a multinomial multiple logistic regression model. A soil map, a biogeographical map and bioindicators derived from temperature and precipitation data were used as predictors. Both methods ensure that the predicted proportions are in the interval [0,1] and sum to 1. The regression predictions were iteratively scaled to the National Forest Inventory statistics and the Forest map of Europe. The predicted proportions for the twenty tree species were validated by the Bhattacharryya distance between predicted and observed proportions at 230 plot data separated from the calibration data. Besides, the map with the predicted dominant species was validated by computing the error matrix. The median Bhattacharryya distance in the subarea with NFI plot data was 1.712, whereas in the subarea with ICP-Level-I data, this was 2.131. The scaling did not significantly decrease the Bhattacharryya distance. The estimated overall accuracy of this map was 43%. In areas with NFI plot data, overall accuracy was 57%, outside these areas 33%. This gain was mainly attributable to the much denser plot data, less to the prediction method.

Item Type: Article
Additional Information: This project was carried out within the framework of European FP5 projects CARBO-Europe IP and EFORWOOD IP. We are greatly indebted to the Forest Focus programme and the National Forest Inventory institute’s correspondents. NFI plot data were received from Jacques Rondeux and Martine Waterinckx, Belgium; Juro Cavlovic, Croatia; Veiko Aderman, Estonia; Kari Korhonen, Finland; Thierry Be´louard, France; Heino Polley, Germany; Marino Vignoli, Remo Bertani, Giorgio Dalmasso and Maurizio Teobaldelli, Italy; Andrius Kuliesis, Lithuania; Wim Daamen and Henny Schoonderwoerd, Netherlands; Stein Tomter, Norway; Susanna Barreiro and Margarida Tome´, Portugal; Olivier Bouriaud, Romania; Vladimir Seben, Slovak Republic; Gal Kusar, Slovenia; J. Villanueva and Antoni Trasobar, Spain; Go¨ran Kempe, Sweden; Bill Mason and Shona Cameron, United Kingdom; Igor Buksha, Ukraine. Finally, we like to thank an anonymous reviewer for his expert comments on the statistics, and the suggestion to use Bhattacharryya distance for validation.
Uncontrolled Keywords: Logistic regression-Kriging � Map validation-Bhattacharryya distance � Confusion matrix- � Overall accuracy
Author Affiliation: Alterra, Wageningen University and Research Centre,Netherlands.
Subjects: Statistics and Experimentation
Crop Improvement
Divisions: General
Depositing User: Mr Arbind Seth
Date Deposited: 07 Oct 2012 13:49
Last Modified: 07 Oct 2012 13:49
Official URL:

Actions (login required)

View Item View Item