Description

Modern forestry requires considerable specialized equipment. First, a feller cuts trees and lays them down on the ground. Then, a skidder pulls those trees to bring them to the road where the branches will be removed by a delimber. The skidders ride on trails where they must make many round trips. These trails are not made of asphalt or gravel, but rather of soil. The intensive use of these trails leads to the formation of ruts and soil compaction on the surface which can be damaging. In this study, FERIC (Forest Engineering Research Institute of Canada/Institut canadien de recherches en génie forestier) with Université du Québec in Abitibi-Témiscamingue, has studied the modifications to two types of soil caused by the traffic of skidders on skid trails.
 

The experimental design is as follows. Two types of soil were studied: clay and sand. For each type of soil, three trails which had never been used were set up where a mechanical cut with regeneration protection was performed a few weeks earlier. Each trail was 30 m long and 4.5 m wide. On each trail a skidder made 15 round trips; going into the forest, the skidder was empty, outbound the skidder was full. Some physical characteristics of the trail were measured before the first round trip as well as after trips one, three, five, seven, eleven, and fifteen.
 

The different measurements taken after each trip previously mentioned are the following. The first is the slippage of the wheels of the skidder, defined as the mean slippage of the four wheels. The slippage of each wheel is a function of the total number of spins and the speed of the skidder. These quantities are determined from the recordings of two video cameras on each side of the trail. Under the weight of the load, the skidder modifies the ground by forming rims and ruts. The rut depth as well as the surface of displaced materials have been computed by taking the average of these measures for three different transversal profiles as follows. For each profile of the width, the height of the ground (with respect to the level before the first trip of the skidder) has been measured at 20 points after the return of the skidder. For a rut, the height is negative whereas it is positive for a rim. The depth of the ruts of a profile is defined as the absolute value of the smallest height among the 20 measured points. The surface of displaced materials for a profile is defined as a function of the absolute value of these 20 heights. Three other characteristics of soil have been measured for each of the trips previously mentioned. The anhydrous density of the soil is considered as an indicator of soil compaction. Measured with a device at depths of 10, 20, and 30 cm under the ruts, only the measurements at 10 cm will be used. Moreover, resistance to penetration (i.e., the strength that must be used to stick a needle in the soil) and the limit of shear resistance (the maximal strength that must be used to provoke a free movement of the soil) are two other parameters measured in prediction models of the effect of traffic on soil. The resistance to penetration is computed by taking means of pressures at depths 5, 10, 15, 20, 25, and 30 cm in mineral soil, but only the measurements at 5 cm will be used. Limit of shear resistance in the ruts is the mean of 10 readings of a shear measurement device sunk in about 10 cm of mineral soil.
 

Preliminary analyses by the researchers have consisted of linear regression analyses for each individual dependent variable (depth of ruts, surface of displaced materials, and density of the soil after the trip) with model selection techniques. Separate analyses were performed for the sand and clay trails. Note that for the density of the soil after the trip, the density of the soil before the trip was not used as a predictor so as to prevent high collinearity. Transformations of variables have been considered. In particular, in the case of sand trails, the difference in rut depths has been used instead of the depth itself since no satisfactory model for that variable was found.
 

The researchers would like to go further in their analysis. The preceding analyses make use of the hypothesis that the errors in the regression models associated to the measurements for each trip are independent. Would it be preferable to consider a model with correlation in the observations of a given trail? Also, many predictors are obtained by interpolation and these interpolations involve measurements which are not available before the prediction can be made. Could we do better by working only with variables measured directly and available before making the prediction? Would nonparametric methods be useful? What are the best models to predict the behavior of rut depth, the surface of displaced materials, and the density of the soil after the trip?
 

We want to thank Mr. Joseph Nader and Mr. Philippe Meek, researchers at FERIC for their help in making this data set available for the Case Studies Session of the Annual Meeting of the SSC.