Prediction of soil moisture regime and its relationship to wood quality in the Boreal forest of Newfoundland
The objective of question nine is to develop a predictive model of soil moisture availability from digital elevation model (DEM) data for various spruce and fir-dominated stands that will then be used to identify the relationship between soil moisture and tree growth across Newfoundland. This research objective was based on recent studies on the predictive value of ecological land-classification (ELC) systems. These studies have suggested that it would be possible to model wood quality attributes, such as wood density and fibre length, of boreal conifers using broad-scale patterns of moisture availability. However, these ELC systems are specific to a given region so a more adaptable approach would be to identify the mechanisms, such as moisture, that determine the patterns of wood quality characteristics and develop a method of classification using this variation. While there are few different methods for modelling soil moisture, there is also a need for simple models that use basic datasets to identify wood quality tendencies.
For this study there will be three phases. The first will involve developing estimates of soil moisture availability for 194 permanent sample plots (PSPs) using both the topographic wetness index (TWI) and stochastic depression analysis (SDA) method on elevation data derived from a provincial imagery product. This elevation data consists of 194 16 km2 tiles with a resolution of 10 meters. The TWI is simply a function of slope and the upstream contribution area per unit width at a right angle to the flow direction. SDA is a tool that takes a stochastic simulation approach to mapping topological depressions in a DEM and is freely available through open source software such as Whitebox. TWI will give us a relative wetness value while SDA will allow us to derive estimates of water table depth. Hence, we will have two different moisture indices to use as indicators of moisture regime for each plot. We will then compare these values to moisture classification based on field data for these sites. The second phase will involve modeling wood quality parameters based on soil moisture indicators using the random forests approach, which was previously employed in studies of ELC systems and wood quality prediction. The third phase looks to evaluate the power of LiDAR-derived DEM data for improving estimates of soil moisture and hence improving models of wood quality. This will be completed by comparing the results grained from a subset of the plots that have both the regular provincial imagery DEM data and the LiDAR-derived DEM data. We anticipate that these models will become useful tools for identifying forest stands characterized by different soil moisture regimes and that particular wood quality properties can be predicted from using these DEM’s and available inventory data.