THE ROLE OF MINERAL SOIL TOPOGRAPHY ON THE SPATIAL DISTRIBUTION OF ORGANIC LAYER THICKNESS IN A PALUDIFIED BOREAL LANDSCAPE 

THE ROLE OF MINERAL SOIL TOPOGRAPHY ON THE SPATIAL DISTRIBUTION OF ORGANIC LAYER THICKNESS IN A PALUDIFIED BOREAL LANDSCAPE 

PRÉSENTATION DE L’ORGANISATION DE LA THÈSE:

Cette thèse comporte cinq chapitres sous forme d’articles scientifiques rédigés en anglais. Chacun des cinq chapitres de cette thèse situe les différentes caractéristiques de la zone étudiée, résume la physiographie de la ceinture d’argile et de la zone d’étude et décrit les caractéristiques générales de la couche organique, le sol minéral, le socle rocheux et les différentes espèces du couvert végétal et du sous-bois. Le design expérimental, la collecte des données utilisées et les analyses réalisées y sont décrits en détail. Chacun des chapitres présente l’analyse et l’interprétation des résultats, une discussion à propos des choix des méthodes utilisées, de leurs limites, des améliorations possibles et de leurs applications à l’échelle du paysage ainsi que de la contribution des résultats à un aménagement durable des forêts paludifiés. La dernière section de la thèse consiste en une conclusion globale faisant ressortir les principales contributions, conclusions et recommandations qui se dégagent de la présente étude.

THE USE OF GROUND PENETRA TING RADAR FOR REM OTE SENSING THE ORGANIC LAYERMINERAL SOIL INTERFACE IN PALUDIFIED BOREAL FORESTS :

The black spruce forests of the Clay Belt cover about 125 000 km2 and are considered as a potential source of wood fibre. However, a considerable volume oftimber in this region is located in areas that are prone to paludification (Fenton et al., 2005; Lavoie et al., 2005). Paludification is a natural process that is characterized by graduai accumulation of surface organic matter deposits over time (Payette and Rochefort 2001), which can lead to losses of forest productivity (Simard et al., 2007, 2009). The depth to the top of the mineral soil beneath the thick organic layer can determine where and to what extent paludification will occur in the Clay Belt (Lavoie et al., 2005). Thus, accurate estimates of organic layer thickness (OLT) are important for forest management, as predictions of this parameter will help guide appropriate management practices to control and improve forest productivity and prevent the advance of paludification into the lowland black spruce forests of the region. Y et, determining the organic layer-mineral soil (OL-MS) interface (representing OLT) over such a large area is complicated by the presence of the relatively thick and dense organic layer. OLT estimation at spatially disconnected locations (i.e., plots) throughout the landscape are time consuming, labour intensive, and cannot provide continuous data at either landscape or regional scales. In addition, information regarding the interface between the organic layer and underlying mineral soil and its spatial variability over large areas cannot be measured directly by traditional remote sensing techniques (i.e., multispectral imaging, photogrammetry), des pite demonstrable advantages of the latter in covering very large surfaces (Valeria et al., 2012). In contrast, ground penetrating radar (GPR), a remote sensing technology, can be a viable alternative for delimiting the OL-MS interface and its spatial continuity through a shallow organic layer in paludified sites of the Clay Belt.

Study area and field sites:

The study area is situated in the James Bay Lowlands physiographic region of Quebec and more specifically in the Clay Belt region that spans the border between Ontario and Quebec (Figure 1.1a). The Clay Belt region of Quebec (Figure 1.1b) is part of the western black spruce-feather moss bioclimatic domain (Robitaille and Saucier, 1996). This region has low relief, which was determined by flat-laying clay deposits of glaciolacustrine origin that were left behind by the pro-glacial Lake Ojibway (Veillette, 1994). Mean annual temperature and precipitation are -0.7 ac and 906 mm, respectively (Environment Canada, 2011), as measured at the Matagami weather station located at about 60 km northeast of our study sites.

GPR measurements, processing, and interpretation:

GPR measurements
In the summer of 2009, GPR surveys were conducted at each site (A and B) using the pulse EKKO PRO (Sensors & Software Inc., Mississauga, Ont.), to determine spatial variability in OLT. GPR measurements were performed at 20 cm intervals along three parallel 40 rn transects on each site; at each site the transects were spaced 10 rn apart. The antenna was carried by hand and moved along the transect to be surveyed. The antenna centre frequency that was used was 200 ~’ which is appropriate as the expected depth of the mineral soil was generally B1.5 m. Each collected trace was the result of 32 stacks to improve the signal-to-noise ratio. Figure 1.3 shows a schematic representation of GPR reflected and propagated EM energy travelling through the organic and mineral soillayers.

GPR data processing
Standard data post-processing was performed on the raw data to enhance them for easier interpretation. EKKO View Deluxe software (Sensors & Software Inc.) was used for data processing. The processing steps included correction for background noise (Dewow filtering) by removing low frequency interferences, zero-time static correction, and the application of spherical and exponential compensation (SEC) gain and automatic gain control (AGC) to enhance weaker reflectors and smoothing across traces. The AGC was suitable for increasing the continuity of reflections within the targeted investigation depth, but it erases all amplitude information (Jol and Bristow, 2003). This was of less importance because amplitude information was not used in the present study.

Identification ofsoillayers and horizon interfaces
In general, the identification and interpretation of interfaces from the GPR profiles depend on how much knowledge the interpreters have about underground structures. Before undertak:ing the interpretation of GPR profiles, it was important to precisely identify the ground wave that represents the top of the surveyed ground surface. We determined the ground wave as the second wave to arrive (after the air wave), as described in Jol and Bristow (2003).

Field soil sam pling and statistical analyses:

Soils were described and classified as organic layer or mineral soil according to the Canadian system of soil classification (Soil Classification Working Group, 1998). This scheme separates the organic layer into three different horizons (Of, Om, and Oh) that are based on the von Post scale (von Post and Granlund, 1926) and that reflect the degree of decomposition of peat. OLT was measured manually using a hand probe to validate the GPR interpretations. Those manual measurements were clone at 1 rn increments along each transect, for a total of 123 measurements at each site. At each sampling point, the hand probe bored through the organic layer until the mineral soil was encountered (Figure 1.4). As those measurements did not provide the thickness of each individual organic horizon, nine soil pits (30 cm 30 cm depth to the mineral soil contact) were dug every 5 rn along each transect for a total of 27 pits per site (Figure 1.4). For each soil pit, the total OLT was measured; the OLMS interface was located, as was each organic horizon. Details of these horizons, together with typical physical properties of the Of, Om, and Oh materials, are provided in Table 1.2.

Direct soil pit-augering information:

Thickness of each horizon, and the total organic layer and OL-MS interface were manually measured along each transect at sites A and B. Our results (Figure 1.5) showed that, between sites A and B, a great degree of variability characterized the depth of the OL-MS interface and the different organic horizons thicknesses and their respective interfaces. Organic layer thicknesses were greater in site B than those in site A. Most Of horizons (93%) in site A were 10-15 cm thick, whereas most Of horizon thicknesses in site B were > 20 cm (96%). The Om and Oh horizons were present in all soil pits in varying amounts (Figure 1.5). Measured EC is higher in the Oh horizon (208-287 11S/cm) than at the OL MS interface (70- 178 11S/cm) (Table 1.3). Such values are characteristic of nonsaline soils and are considered low. The low EC was confmned by a background EM31 conductivity survey that was conducted over the three transects of each site prior to GPR survey. The apparent EC values from this EM31 survey ranged from 13 to 18 11S/m for the 1 rn horizontal dipole mode and from 19 to 26 11S/m for the 1 rn vertical dipole mode (unpublished results).

GPR analysis and interpretation
GPR produced different reflection patterns at the ground surface and belowground on both sites (Figures 1.6a and 1.6b). An effect of ground coupling is evident near surface. This coup ling is caused by the interaction between the EM wave that is radiated from the antennae and the ground, which leads to continuous reflections at the top of each profile. These reflections are not representative of the true signal amplitude, because the GPR receiver antenna is saturated with that signal. The frrst 0.20 rn depth at site A and about the first 0.15 rn at site B correspond to ground coupling (Figures 1.6a and 1.6b). Therefore, they were not considered for retrieving the organic layering information. Below these zones, profiles of each site are characterized by distinct! y different patterns of reflections, which were enhanced by the application of gain (Figure 1.6).

Organic layer thickness estimation from the GPR data at site A
A continuous reflection in all of the GPR profiles occurs at about 40 cm below the surface at position 0 rn along the transect (Figure 1.7). Ground truth data from soil pitaugering measurements suggest that this reflection coincides with location of the OL-MS interface. At this interface, substantial decreases in substrate water content are found (Table 1.3) and the abrupt changes in dielectric constant likely lead to this distinct reflection. The mineral-organic interface is nearly continuous and located at about same depth along the three parallel transects, indicating that its position is likely spatially constant over the investigated site. OLT estimates that were obtained with GPR and manual probing were well correlated (Figure 1.8; r = 0.93, P < 0.001). Overall, depths to the OL-MS interface that were obtained from GPR and manual measurements were very similar and differed only by ± 2 cm.

MAPPING PERMANENT AND REVERSIBLE PALUDIFICATION IN BLACK SPRUCE FORESTS, CANADA USING REMOTE SENSING AND GIS-BASED AUTOMATED CLASSIFICATION:

Accumulation of a thick organic layer over time is a characteristic of the northern boreal forest regions such as the interior of Alaska, the Canadian Hudson Bay- James Bay lowlands, and the western Siberian plain. This accumulation is mainly attributed to the natural process of paludification, which is generally thought to crea te wetter conditions that lead over time to a reduction in soil temperature, decomposition rates, microbial activity, nutrient availability, and canopy openings (Crawford et al., 2003; Lavoie et al., 2005; Vygodskaya et al., 2007). Paludification can be regarded as an important factor that causes substantial losses in productivity in the boreal forest and, consequently, potential sources of wood fiber. Paludification is particularly problematic in the forested landscape of the Clay Belt, a region of the Canadian Hudson Bay- James Bay lowlands, where this process has contributed to the transformation of many productive forested areas into non productive forested peatlands. Within the Clay Belt region, ground surface topography and time-since-last fire have been considered as the two major drivers of paludification (Fenton et al., 2009). Consequently, permanent (also called edaphic or paludification ofwet depressions in Payette and Rochefort 2001) and reversible paludification (or successional in Simard et al., 2009) are two types of paludification that can be recognised. Within the landscape, permanent paludification dominates in natural depressions, which have wetter soil conditions that favour organic layer build-up. Reversible paludification occurs on flat or sloping terrain, where a feather mossdominated bryophyte layer is replaced over time by Sphagnum species (Boudreault et al., 2002; Fenton and Bergeron 2006), starting at approximately 100 years after fire (Fenton and Bergeron 2006; Simard et al., 2007). On one hand, reversible paludification may be reversed through natural severe fire or a combination of silvicultural practices and site preparation, as detailed by Fenton et al., (2009). On the other hand, permanent paludification can be considered as an irreversible condition.

Correlations between site index and individual topographie variables:

Derived topographie variables from each of the selected plots were used to evaluate resolution effects on site index (Table 5.3). Among the nine topographie variables that were tested, slope had the highest correlation with site index across 5- to 20-m resolutions (Table 5.3). Slope was consistently and positively related to site index over the landscape, with a correlation that increased from r = 0.27 to r = 0.35 as spatial resolution increased from 5- to 10-m (Table 5.3). Topographie wetness index was significantly (P < 0.05) correlated with site index, ranging from -0.23 to -0.27 at 10-m and 15-m resolutions, respectively. At 20-m resolution, a more significant correlation was found between topographie wetness index and site index (r = -0.31, P < 0.01; Table 5.3), whereas no significant correlations were found at the 5-m resolution (r = -0.18, P > 0.05; Table 5.3). As was the case for slope, the correlation increased with increasing resolution for the topographie wetness index. At 10- to 20-m resolutions, the negative correlation between site index and topographie wetness index, which represents soil moisture variation caused by water accumulation, indicated that forest productivity tended to increase with dryer soil conditions (represented by low TWI values; Figure 5.4d). This result agrees with field observations for which productive forest was usually associated with areas having dryer conditions. Across all resolutions, correlation between site index and aspect index were constantly very weak and not significant (r :S 0.16, P > 0.05; Table 5.3). Nevertheless, this correlation increased for all resolutions when we divided the aspect index into two classes, viz., southwest and northeast (Table 5.3; Figures. 5.4e and 5.4f). For instance, at 5-m resolution, southwest-facing areas exhibited an improved significant correlation coefficient (r = -0.41, P < 0.01; Figure 5.4e). Across all resolutions, southwest-facing plots were all significantly correlated with site index (r ranging from 0.41[5ml to 0.23[5m1 , Table 5.3), whereas northeastfacing plots were only significantly correlated at 5-m resolution (r = -0.45, P < 0.05; Table 5.3). Across all DTM resolutions, correlation between elevation and site index were also consistently very weak and not significant (r = -0.12[5-, 10., 15.m1 and r = -0.11[zo-m], P > 0.05; Table 5.3). This weak correlation indicated that elevation had no influence on forest 162 productivity and could not be used to discrimina te between areas of higher and lower forest productivity over the entire study area (Figure 5.4f). Topographie position index, which is simply the difference between a cell elevation value and the average elevation of the neighbourhood around that cell, was also found not to be significantly correlated with site index across the four resolutions (r <::: 0.23, P > 0.05). All curvature variables (mean, profile and plan) were not correlated with site index for 5- to 15-m; however, correlation was weak but significant (r <::: 0.23, P < 0.05) at 20-m resolution. When we divided each of the curvature variables into two classes, viz., concave and convex, the correlation was not significant across all the resolutions (5- to 20-m), suggesting that curvature variables (mean, profile and plan) did not play a role in controlling forest productivity at any of the investigated resolutions.

CONCLUSION GÉNÉRALE:

Approfondir les connaissances sur les relations entre les variables permanentes du site et la productivité forestière de la pessière à épinette noire à l’aide d’approches spatiales est devenue essentielle. Cette thèse a permis d’identifier des relations quantitatives impliquant le degré de paludification, la topographie du sol minéral et de surface et la productivité des forêts à l’échelle locale et du paysage. Les résultats de cette thèse nous ont permis de déterminer différents seuils qui pourraient caractériser, à la fois, des zones paludifiées vs non paludifiés et des zones productives vs improductives. Nous avons été en mesure de cartographier l’étendue spatiale de la paludification (permanente vs réversibles), l’épaisseur de la couche organique (ECO) et la topographie (pente, TWI, TPI etc.) à l’échelle du paysage. Nous avons démontré l’apport significatif de l’utilisation des données terrain et des modèles topographiques multi-résolutions pour quantifier l’effet des échelles sur la productivité forestière. Nos méthodes d’analyse et nos résultats deviennent ainsi un outil essentiel dans la gestion durable des territoires paludifiés. Les aspects innovants de recherche ainsi que les apports scientifiques de chaque chapitre sont détaillés ci-après.

Table des matières

INTRODUCTION GÉNÉRALE 
A.l. Mise en contexte et problématique
A.2. Objectifs de la thèse
A.3. Méthode et matériel
A.3.1. Présentation de la zone d’étude
A.3.2. Dispositif expérimental et collecte des données .
A.3.3. Évaluation de la productivité
A.3.4. Méthodes de télédétection à haute résolution et précision spatiales
A.3.4.1. Méthode Géoradar.
A.3.4.2. Technique LIDAR
A.3.5. Cartographie et délimitation spatiales de la paludification réversible vs
permanente
A.3.6. Analyse des données
A. S. Présentation de l’organisation de la thèse
CHAPITRE I: THE USE OF GROUND PENETRA TING RADAR FOR REMOTE SENSING THE ORGANlC LAYER-MINERAL SOIL INTERFACE IN PALUDIFIED BOREAL FORESTS 
Abstract
RÉSUJ’vlÉ
1.1. Introduction
1.2. I’viaterials and methods
1.2.1. Study area and field sites
1.2.2. GPR measurements, processing, and interpretation
1.2.2.1. GPR measurements
1.2.2.2. GPR data processing
1.2.2.3. Identification of soillayers and horizon interfaces
1.2.3. Field soil sampling and statistical analyses
1.3. Results
1.3.1. Direct soil pit-augering information
1.3.2. GPR analysis and interpretation
1.3.3. Organic layer thickness estimation from the GPR data at site A
1.3.4. Organic layer thickness estimation from the GPR data at site B
1.4. Discussion
1.4.1. Utility and limitations of GPR
1.4.1.1. Estimating the OL-MS interface at site A ..
1.4.1.2. Estimating the OL-MS interface at site B ..
1.4.1.3. Identifying the Of and Om horizons ..
1.4.1.4. Potential use of GPR data at the lands cape and regional scale within
the Clay Belt
1.5. Conclusions
CHAPITRE II: LANDSCAPE-SCALE INFLUENCE OF TOPOGRAPHY ON ORGANIC LAYER ACCUMULATION IN PALUDIFIED BOREAL FORESTS 
Abstract
Résumé
2.1. Introduction
2.2. I’viaterials and methods
2.2.1. Study area
2.2.2. Sampling design and field data collection
2.2.3. LiDAR processing and topographie variable measurements .
2.3.1. Effect of different LiDAR-derived D1M resolutions on the values of
topographie variable
2.3.2. Correlations between OLT and topographie variables based on different
DTM resolutions
2.3.3. Regression Tree-Based Lands cape Segmentation .
2.3.3.1. Areas with Slope <::: 2.3%. .
2.3.3.2. Areas with Slope > 2.3%.
2.4. Discussion
2.4.1. Relationships between OLT and topographie variables at different
DTM resolutions
2.4.2. Landscape Segmentation
2.4.3. Management implications and future research
2.5. Conclusions
CHAPITRE III: THE ROLE OF MINERAL SOIL TOPOGRAPHY ON THE SPATIAL DISTRIBUTION OF ORGANIC LAYER THICKNESS IN A PALUDIFIED BOREAL LANDSCAPE 
Abstract
Résumé
3.1. Introduction
3.2. Methods and materials
3.2.1. Study area
3.2.2. Sampling design and field data collection
3.2.3. Mineral soil topography .
3.2.4. Relating topography variables and OLT
3.3. Results
3.3.1. Importance of individual predictor variables
3.3.2. Regression tree-based model evaluations
3.3.2.1. Regression tree-based modell..
3.3.2.2. Regression tree-based model 2
3.4. Discussion
3.4.1. Individual relationship trends
3.4.2. Regression tree-based modelling approach
3.4.3. Management implications
3.5. Conclusions
CHAPITRE IV: MAPPING PERMANENT AND REVERSIBLE PALUDIFICATION
IN BLACK SPRUCE FORESTS, CANADA USING REMOTE SENSING AND
GIS-BASED AUTOMATED CLASSIFICATION
Abstract
Résumé
4.1. Introduction
4.2. Study area
4.3. Materials and methods
4.3.1. Digital terrain madel and derived topographie variables
4.3.2. Field datasets collection
4.3.3. Topographie position index and classification
4.3.4. Topographie wetness index and classification
4.3. 5. Assignment of paludification types to topographie position classes and
theirs validations
4.3.6. Statistical analyses
4.4. Results and discussion
4.4.1. Topographie position classification
4.4.2. Relationship between topographie position classes and individual
topographie variables
4.4.3. Assigning topographie position classes to paludification types
4.4.3.1. Deep depression class
4.4.3.2. Lower slope depressions class
4.4.3.3. Flat surfaces class
4.4.3.4. M id-slope, upper slope and hilltop classes
4.4.4. Topographie position classification performance
4.4.5. Validation of the produced maps
4.4.6. Forest management implications
4.5. Conclusions
CHAPITRE V: EFFECTS OF TOPOGRAPHY AND THICKNESS OF ORGANIC
LAYER ON PRODUCTIVITY OF BLACK SPRUCE BOREAL FORESTS OF
NORTHWESTERN QUEBEC 
Abstract
Résumé
5.1. Introduction  144
5.2. I’vlaterials and methods
5.2.1. Study area
5.2.2. Sampling design and field data collection
5.2.3. Forest productivity assessment
5.2. 5. Collection of predictor variables explaining forest productivity
5.2. 5.1. LiDAR derived topographie variables
5.2.5.2. Organic layer thickness measurements
5.2.6. Statistical analyses
5.2.7. Model validation
5.3. Results
5.3.1. Vegetation characteristics and their relationships to site index at the plot-scale
5.3.2. Correlations between site index and individual topographie variables
5.3.3. Forest productivity modelling results
5.3.3.1. Forest productivity modelled using model 1
5.3.3.2. Forest productivity modelled using model 2
5.3.4. Construction and validation of predictive site index maps
5.4. Discussion
5.4.1. Individual relationships between topographie variables and site index
5.4.2. Regression tree model 1
5.4.3. Regression tree model 2
5.4.4. Model performance
5.4. 5 I’vlanagement implications
5.5. Conclusions GÉNÉRALE

Télécharger le rapport complet