Tuesday, March 29, 2011

Spatializing willow productivity: more resolution and better estimates

In this paper I have presented better estimates for willow yields for bioenergy, using the large dataset of willow commercial plantations in Sweden that I used in previous papers. However, I think this time the study manages to use this information at its full potential. You can read from the following link, and as usual, feedback and ideas to continue this line are very welcome!

MOLA-YUDEGO B. 2011. Predicting and mapping productivity of short rotation willow plantations in Sweden based on climatic data using a non-parametric method. Agricultural and Forest Meteorology. In Press.

Abstract
In this study, estimates for yield of short rotation plantations are provided based on climatic variables using the k nearest neighbour method. The calculations were based on climatic data and yield records from 1790 willow plantations in central and southern Sweden, divided into three categories based on local performance. The chosen neighbours were weighted proportionally to the inverse squared distance measured in the feature space defined by the climatic variables. The climatic variables included monthly averages of maximum, minimum and mean temperatures and precipitation. These were weighted using empirical constants after an optimisation process. The best accuracy was obtained with k = 4 for the group of high performance plantations, and k = 5 for the other groups. The relative RMSE values were 37.9%, 24.4% and 38.9% for the high, medium and low local performance, respectively, and the corresponding relative biases were 2.10%, −0.95% and −1.30%. The method was applied to interpolate the yield values in order to perform maps of potential productivity for the whole area. The results of this approach indicate that it can provide faster and more accurate predictions than previous modelling approaches, and can offer interesting approaches in the field of yield modelling.
Research highlights

► A k-NN method applied to climatic variables is effective for spatialising productivity. ► Relative RMSE were 37.9%, 24.4% and 38.9% for three levels of management performance. ► The method presented can serve to develop site index models for new areas.

Keywords: k-NN methods; Growth and yield models; Bioenergy; Wood fuels


More info!

Agricultural and Forest Meteorology at Science Direct [link]
ResearchGate [link]

No comments:

Post a Comment