Updating Forest Inventory Data by Remote Sensing or Growth Models to Characterise Maritime Pine Stands at the Management Unit Level

Abstract

In order to get the information required for a decision support system for multifunctional forest ecosystem management, two forest inventory updating methods were developed and applied to a maritime pine forest area. The study area is a public property, Mata Nacional de Leiria, located in the coastal sand dunes of central Portugal, managed for high quality timber production. This forest area has a continuous plot based forest inventory with a high ground sampling intensity (1plot/ha), but the periodicity of the measurements does not allow the characterisation of all management units for the same reference year. To overcome this situation, we tested and compared two methods. One method was based on Landsat TM imagery and the other on the use of growth models. In the first one, stand characteristics were modelled as a function of TM-bands, combinations of TM-bands and stand age, using linear, and multilinear regression models. In the second method, plot-level data was updated for the same reference year using a diameter distribution growth and yield model. Results show that Landsat based volume prediction has a very good performance, specially when stand age is introduced in the models, and that selection of TM-band combinations by stepwise regression is better than the use of traditional vegetation indices, such as NDVI. The comparison of both methods reveals that volume predictions are reasonably similar. Remote sensing based method for prediction of management units volumes has the advantage of requiring a less intensive ground measurement effort, and is therefore more cost-effective.