Introduction

Soils are an integral component of the natural environment, consisting of complex interactions between organic and inorganic constituents as affected by soil-forming factors, namely geology, climate, topography, organisms, and time (Jenny 1941; Birkeland 1999; Adhikari et al. 2012). As such, soils vary vertically with increasing depth (soil profile) and laterally by spatial location. These changes are generally summarized in the form of soil survey reports which have long been utilized as the dominant format for summarizing and mapping soils in Canada, with the first soil survey conducted over a century ago (McKeague and Stobbe 1978). For the Province of New Brunswick, Canada, these surveys have been conducted on a county-by-county basis over the past six decades, with the most recent survey conducted in 2000 (Michalica et al. 2000; Anderson and Smith 2011).

Soil surveys completed within NB summarize soils via landform- and lithology-defined soil associations, and further divide these into drainage-explicit soil associates. Each soil associate is typically assigned one or more soil profiles (horizon sequence from forest floor to parent material) summarizing the depth, chemical, and physical properties of each horizon (McKeague and Stobbe 1978; Anderson and Smith 2011). This information is typically outlined via three sections, with the third section often represented as an appendix at the end of the survey [e.g., description of Poitras soil associate retrieved from Langmaid et al. (1980)]:

Section 1: It comprises an overview of each soil associate within the surveyed area, including information pertaining to parent soil association and soil-forming factors, namely landform, lithology, vegetation, drainage, and topographic surface expression.

Section 2: It provides profile description for sampled soil associates including field-based, horizon-specific identification and measurements including depth, texture, coarse fragment (CF) content, structure, root presence, mottling (if applicable), and pH (Fig. 1).

Fig. 1.

Example of information obtained from soil surveys and utilized in developing the database, including general information (Section 1) (top paragraph) and field-based measurements for each horizon (Section 2) (bottom descriptions) separated by dotted line. The example provided represents the Poitras soil associate retrieved from Langmaid et al. (1980).

Section 3: It outlines the laboratory-measured physical and chemical properties, by horizon, for each soil associate [e.g., horizon description with depth, % carbon, % sand, silt and clay, bulk density (D_b), field capacity (FC), permanent wilting point (PWP)] (Fig. 2).

Fig. 2.

Example of information obtained from soil surveys and utilized in developing the database, including laboratory-measured, horizon-specific soil physical and chemical properties. The example provided represents the Poitras soil associate retrieved from Langmaid et al. (1980).

Soils vary with changing soil-forming factors which vary continuously over landscapes, whereas individual soil surveys are limited geographically, often by administrative boundaries (a provincial county) that are not related to soil-forming factors. Although individual soil surveys are useful as stand-alone documents, they describe the range of soil properties within a limited geographic area that may not adequately capture the full range of variability of soil properties within a soil type. In addition, soils surveys have been collected over decades, a time frame that spans significant changes in analytical procedures, protocols, and methods (Subcommittee on Methods of Analysis of Canada Soil Survey Committee 1978; Guertin et al. 1984; Carter and Gregorich 2006). Therefore, it is important to combine soil surveys into harmonized databases which provide the full variability of soil properties found across environmental gradients while ensuring consistent methods and units of measurement for individual soil properties.

Compiling available soil survey reports for NB revealed numerous inconsistencies in terms of (i) naming of the same soil associations and subsequent associates, (ii) labelling horizon descriptors, and (iii) methods, and units, of measurement for analyzed soil properties. This is, in part, due to changing soil classification and mapping protocols over the past six decades, with the first published soil survey for NB conducted in 1940 (Stobbe 1940).

The objective of this article was to introduce and describe a framework applied to develop a seamless and terminologically consistent aspatial forest soils database by amalgamating and harmonizing existing soil survey reports for NB as a case study. As such, this study aims to outline the step-by-step process in which the database was created as a framework for application in other geographic locations, whether at a regional, provincial, national, or international scale. This objective was accomplished by

Materials and Methods

Survey report amalgamation

The amalgamation of soil survey reports and harmonization of soil attributes was guided by

1. the standardized soil surveying terminology of the Mapping System Working Group (1981) and the Expert Committee on Soil Survey (1982),

2. sampling and analytical techniques as described by Sub Comiittee on Methods of Analysis (1978) and Guertin et al. (1984),

3. The National Soils Database [“NSDB”, spatial coverage with summary documentation, Canadian Soil Information Service (2000)],

4. the two province-wide soil summary documents and associated distribution maps:

   1. “Soils of New Brunswick: the Second Approximation” [“SNB”, Fahmy et al. (2010)],

   2. “Forest Soils of New Brunswick” [“FSNB”, Colpitts et al. (1995)].

Soil survey reports for NB, in addition to the NSDB, SNB, and FSNB documentation, were retrieved from the publications section of the Canadian Soil Information System (Canadian Soil Information Service 2012). Each survey was assessed to determine the extent of data availability and spatial coverage (Table 1; Fig. 3). Soil survey reports excluded from the database were excluded either due to the omission of horizon-specific data, or insufficient information pertaining to horizon-specific soil-forming processes, e.g., “A₁” instead of “Ae” or “Ah”. Also excluded were survey reports specifically dealing with small sections of agricultural lands due to the influence of ploughing and tillage on soil properties.

Table 1.

Overview of available soil surveys for New Brunswick, Canada, with relevant information regarding utility for inclusion into the database.

Fig. 3.

Spatial coverage of soil surveys utilized for developing the database. Also included is elevation and physiographic regions to represent how coverage varies in each region. Physiographic regions retrieved from Colpitts et al. (1995), elevation with shaded relief retrieved from Furze et al. (2017), and survey outlines retrieved from Canadian Soil Information Service (2012). Figure created in ArcGIS (Environmental Systems Research Institute 2018).

The compilation of soil survey data within the database was conducted manually on a row-by-row basis following a hierarchical framework (Table 2) which first separates the database by the original soil survey from which the data was retrieved. Following survey source, the database was separated by soil association then soil-forming factors for each association, including dominant vegetation type, topography (surface expression, slope position, slope steepness, and aspect), and soil parent materials (lithology and mode of deposition). Next, the database was separated by drainage, stoniness, and rockiness, with individual soil associates assigned to each drainage class. Each soil associate was provided with either single or multiple horizon sequences depending on how often that associate was sampled within each survey. Each soil horizon was also assigned depths for which the horizon begins and ends in addition to measured soil properties for each horizon.

Table 2.

Visual example of database layout for each association within the database utilizing one of the Holmesville association entries.

This compilation resulted in an amalgamated database consisting of 522 soil profiles with 2490 rows of horizon-specific data. Inspecting the amalgamated database, however, revealed inconsistencies between surveys, including

1. naming of soil associations and soil associates,

2. variability and incompleteness in soil drainage and soil order classifications,

3. labelling, and incompleteness, of soil-forming factor entries, mainly parent material, topographic surface expression, vegetative cover, and

4. changes in soil survey methods such as sampling strategies, laboratory analyses, and quantitative units for reporting results (Table 3) (Subcommittee on Methods of Analysis of Canada Soil Survey Committee 1978; Mapping System Working Group 1981; Expert Committee on Soil Survey 1982; Guertin et al. 1984).

Table 3.

Overview of variability in units of measurements for select soil properties from soil surveys utilized in developing the database.

Correcting soil naming, drainage, and classification inconsistencies

Naming

Naming inconsistencies of both soil associations and associates were resolved utilizing the FSNB and SNB reports as guiding authority as follows:

1. Some of the soil associations were referred to as complexes between two associations, e.g., “Baie du Vin – Galloway”, “Barrieau – Buctouche”, and “Parleeville – Tobique” due to similar soil-forming factors and soil properties. For these instances, only one name was retained based on descriptions of parent material lithology and mode of deposition. For example, the Baie du Vin – Galloway complex was assigned a sandstone lithology and a glaciomarine mode of deposition. From this, Galloway was assigned as the soil association because it occurs on glaciomarine deposits, whereas Baie du Vin occurs on glaciofluvial and marine deposits.

2. In some situations, only the names of the soil associations, instead of individual soil associates, were provided for each profile although explicit drainage classes were assigned to each. For these, new drainage-related soil associate names were assigned based on table 6 (“Correlation of New Brunswick Soil Series/Associations with Forest Soil Units”) of the SNB report, resulting in 97 soil profiles with updated soil associate names.

3. Within some reports, horizon and depth specifications by soil associate (Section 2 of Fig. 1) were inconsistent with their listing at the end of the reports (Section 3, Fig. 2). This was most prevalent in the Northern Victoria and St. Quentin soil surveys. In these situations, the depths provided with the property measurements (Section 3) were retained.

Also, inconsistent was the quantity of data provided for each soil associate. For example, only general information was provided for some soil associates (Sections 1 and 2), whereas measured soil properties (Section 3) were omitted. This resulted in some soil associates lacking horizon-specific property measurements altogether. This was the case for 150 soil profiles within the database.

Drainage

Soil drainage classifications ranged from very poor (wetlands and organic soils) to rapidly and excessively drained (coarse-textured, upper slope positions), as outlined by the Expert Committee on Soil Survey (1982). The procedure in Fig. 4 was used to determine if soil drainage was correctly classified for each soil associate in terms of soil horizon sequence, as well as properly assigning drainage classifications to horizon sequences lacking this information.

Fig. 4.

Visual representation of model developed to assign drainage regime to aspatial database. Resulting drainage classes are bolded. Note that “Mot.” Is abbreviation for mottle, “CF” for coarse fragments, “C” for coarse, “M” for moderate, “LS” for loamy sand, “S” for sand, “SL” for sandy loam, “L” for loam, “SiL” for silty loam.

This procedure ensured that (i) the soil associate names within each association were consistent with the drainage expectations based on table 4 of the SNB report, titled “New Brunswick mineral soil catenas”, and (ii) drainage and slope positions were congruent such that well- to rapidly drained members occur on upper slope and hill-crest positions, imperfect- to moderately well-drained members occur on the lower slopes, and very poor to poorly drained members along toe slopes and in depressions. The derived drainage classifications were generally consistent with the original drainage assignments. In cases where the original drainage classifications provided broad ranges, the middle drainage class was assigned as a median, e.g., “very poor – imperfect” reassigned to “poor”.

Soil classification

Each soil profile was classified by way of the hierarchical Canadian Soil Classification System context by specifying its belonging to a soil order, great group, and subgroup (Soil Classification Working Group 1998). Classifications for each soil profile were assessed by comparing the provided horizon sequences to those found in the Canadian Soil Classification System. The database includes the Podzolic, Brunisolic, Regosolic, Luvisolic, Gleysolic, and Organic orders.

Correcting horizon-specific properties

Horizon descriptions and depths

Considerable effort was placed on ensuring that the soil profile and individual horizon classifications within the surveys were consistent with those outlined in Soil Classification Working Group (1998). All horizons were generalized by master horizon (forest floor, A, B, and C) and by the first subscript for each master horizon, e.g., Ae, Ah, Bf (represents the dominant process influencing the soil). When generalizing by the initial subscript, dominant suffixes followed by the “j” subscript were replaced with the “m” subscript. For example, a “Bf_j” was replaced with “Bm”. Additionally, dominant horizon specifications such as g, c, x, j, and t, were entered into the database in their own individual columns as values ranging from 0 to 1 depending on prominence (0 = absent, 0.5 = partial (“j” suffix), and 1 = present). For example, the presence of gleying in a Bg horizon would receive a value of 1, Bgj a value of 0.5 and B a value of 0 for the “g” column.

Profiles where soil horizon descriptors were marked by “?” or “or” were re-labelled through cross-referencing with other similar soil profiles. It was also ensured that horizon depths (many originally measured in inches) were entered into the database in metric format. Additionally, some C horizons were provided depths with a “+” sign because the bottom of the horizon was not reached (Table 2 — Holmesville soil associate) and, for such cases, these were retained. A new depth column was added to the database, and the depth to the center of each horizon was provided. When the depth to the bottom of the horizon was unknown, the center depth was left blank.

Soil physical properties

For each mineral horizon, measured values for soil physical properties, namely soil texture, CF content, structure, SOM content, D_b, and water retention (at both FC and PWP) were entered into the database utilizing specific procedures for each property, as outlined below.

Soil texture information, was entered into the database in two forms:

1. texture classes assigned from the soil texture ternary diagram, as outlined in Soil Classification Working Group (1998) (Fig. 5), and

2. proportions of sand, silt, and clay within the fine-earth fraction, measured as percentages with the summation equaling 100% (although not always the case within the database).

Fig. 5.

Texture composition of soil samples for each soil survey utilized in developing the database overlain on texture ternary diagram. Figure was derived in R Studio (RStudio Team 2020) using packages “ggplot2”, “plyr”, and “ggtern” (Wickham 2011, 2016; Hamilton and Ferry 2018).

Some texture descriptions provided broad ranges, e.g., “SiL-LS” (silty loam – loamy sand). For these cases, both texture class and percentage of sand, silt, and clay were re-assigned by choosing the midpoint within these classes on the texture ternary diagram. Additionally, some texture classifications did not fall within the realm of the ternary diagram, e.g., “G” (gravel) and “SG” (sandy gravel). This had occurred within 37 samples to which sand, silt, and clay contents remained absent. Although not in the texture triangle, these classifications were retained and placed in a separate column. Also, most texture classifications were assigned including a modifier, e.g., “vfSL” (very fine sandy loam). These modifiers were retained, but texture classes without modifiers were placed in a separate column.

For some horizons, only a texture class was provided, this was typically the case when specific horizon properties (Section 3) were absent from the survey. With these horizons, the percentage of sand, silt, and clay was left absent. For the cases where only the percentage of sand, silt, and clay were present (without assigned texture class), an automated model (Table 4) was derived to determine the texture class based on these percentages. This was used to fill in the voids where the texture classes were absent.

Table 4.

Logical rule statements applied in ascending order to determine proper soil texture class based on texture ternary diagram as outlined in Soil Classification Working Group (1998).

Coarse fragment content entries within the database varied in format, including ranges, qualitative descriptions, e.g., “few” or “some”, and specific measurements (e.g., 10%). Where ranges were provided, the mean values were assigned. Additionally, CF content was also included as part of the horizon texture description, e.g., “gravelly sandy loam”. For these cases, the suggestions of the Expert Committee on Soil Survey (1982) were adopted as follows: descriptions including the adjective “Non” were assigned <15% CF by volume, descriptions with no adjective were assigned 15%–35% CF content, descriptions including the adjective “very” were assigned 30%–60%, and descriptions including the adjective “extremely” were assigned CF contents >60%. Parent material modes of deposition were reviewed, via pivot tables, to determine their applicability in assigning CF contents, but some modes of deposition lacked CF measurements altogether (Table 5). As a result, some soil associations lack CF content measurements.

Table 5.

Overview of coarse fragment (CF) content for each parent material mode of deposition found within database.

Soil structure was provided as a description with three components, shape, size, and distinctness, with all three components assigned to the database. An overview of soil structures can be found in Expert Committee on Soil Survey (1982). It was noticed that terminology for structureless soils were used interchangeably, namely “single grain”, “loose”, “amorphous”, and “massive”; therefore, these were grouped into two classes (“massive” for amorphous and massive descriptions and “single grain” for the remaining two). With this, binary columns were developed for each structure class and assigned 0 or 1 if absent or present, respectively. Samples assigned as transitions between two structure classes (e.g., subangular blocky – platy) were provided a value of 1 in the columns for each structure class.

Soil organic matter content was provided in four formats, % SOM, % carbon, and loss on ignition (LOI) at 450 °C and at 850 °C. Soil surveys for Plaster Rock and Northern Victoria Counties provided both % carbon and LOI at 450 °C (328 samples, 13% of database). Kent County was the only report to record LOI at 850 °C (11 samples, 0.4% of database) and did not record % carbon for comparison. Due to the lack of samples and omission of carbon values for comparison, readings for LOI at 850 °C were omitted. The % carbon readings were converted to % organic matter via eq. 1.

(1)

where %SOM is % soil organic matter, %C is % carbon, and 1.72 is the conversion factor because SOM is composed of 58% carbon (Romano and Palladino 2002; Pollacco 2008; Chaudhari et al. 2013; Poggio et al. 2013). Standardizing these measurements into %SOM resulted in 1202 samples with measurements (48.3% of database).

Soil density measurements were provided for both particle density (D_p) and D_b. Box plots for D_b were used per horizon label to determine the extent to which outliers were skewing the dataset (Fig. 6). For each outlier, the original report was reviewed to determine if a data entry mistake had occurred. It was ensured that the ranges of the density values were generally consistent with soil texture, SOM, and soil depth expectations, with additional considerations to distinguish density in compacted versus non-compacted soils. Data entry with obvious data errors (soils with D_b greater than densities for silicate rocks) were deleted (n = 4).

Fig. 6.

Visual comparison of range in D_b values associated with each master horizon to assess presence (or absence) of outliers. Plot derived in R Studio (RStudio Team 2020).

Water retention measurements were provided in bars (bar), atmospheres (atm), and kilopascals (kPa) with values measured in both volumetric and gravimetric forms. Together, 10 reports provided water retention measurements in gravimetric form, whereas five reports provided measurements in volumetric form. Two reports did not specify whether measurements were gravimetric or volumetric. Gravimetric water retention values were recorded and converted to volumetric form via multiplication with D_b, where applicable. The different units of measurement were then amalgamated and adjusted to represent water retentions in kPa: −33 kPa for water retention at FC, and −1500 kPa for water retention at PWP. Additional moisture measurements included water % at 0 cm, water-holding capacity, maximum water-holding capacity, and water retention at saturation, 10 cm, 50 cm, 100 cm, −100 kPa, −400 kPa, hygroscopic moisture, available water, and moisture percentage. Emphasis was placed on moisture retention at FC and PWP due to the influence of these pressures on rooting. Once combined, water retention at FC had 836 samples measured (34.0% of database), whereas PWP had 743 samples measured (29.8%).

Results

Drainage classifications

Incorrect drainage entries were addressed via the framework highlighted in Fig. 4 resulting in biased representations of different drainage classes throughout the database (Table 6). From this, poorly drained soils represent 16.63% of the database, whereas imperfectly drained soils represent 21.76%, moderately well-drained represent 16.47%, well-drained represent 34.06%, excessively drained represent 10.16%, and <1% with absent drainage classifications.

Table 6.

Representation of drainage classes assigned to soil associates within the database with separations highlighting the dominant drainage classes.

Soil profile classifications

The variability in soil classifications assigned to soil profiles within the database is presented in Table 7. Podzols represent nearly half of the database (46.9% of database) followed by Luvisols (18.8%), Brunisols and Gleysols (both at 12.8%, respectively), Organic (6.4%), then Regosols (2.3%). Once all soil classifications were assessed and updated, abbreviations and rankings for soil classifications, stoniness, rockiness, and drainage were assigned to every soil associate (where applicable).

Table 7.

Overview of soil orders separated by great group and subgroup within the database with overall representation of great groups provided.

Soil parent material

Soil parent materials were updated highlighting the variability in both modes of deposition (Table 8) and primary lithology between soil associations. From this, lithology was used to determine the mineral hardness, dominant grain size, and dominant rock type of each lithological class (Table 9).

Table 8.

Summary of updated parent material modes of deposition (landforms) within the database including the quantity of associations within each mode of deposition.

Table 9.

Overview of lithology (dominant rock types) within the database and associated mineral hardness, and dominant grain size and rock type classifications.

Soil horizon classifications

The variability in soil horizon classifications is outlined in Table 10. Summarizing the variations by master horizon and dominant subscript for the mineral soil horizons (excluding forest floor) represents the range in environmental conditions, and processes, represented by the database (Table 11). From this, B horizons represent 41.78% of the database followed by C horizons (27.60%), A horizons (20.17%), then BC horizons (10.16%).

Table 10.

Variability in soil horizon classification encountered within the database, separated by master horizons, followed by primary subscripts, resulting in 180 unique soil horizons.

Table 11.

Variability in master horizons and dominant subscripts for mineral soil horizons within the database.

Soil physical properties

Correcting inconsistencies within the texture entries (both percentages of sand, silt, and clay, and texture classes) resulted in 86.7% of the database having texture measurements. The variability of texture between soil survey reports is outlined in Fig. 5. Most reports tend to sample soils which fall within the center of the texture ternary diagram (texture class of loam). On the contrary, heavy clays (>80% clay), sandy clays, and pure silts remain unsampled.

CF content entries within the database remained separate from one another depending on the format of the original measurements. For example, measurements provided as a specific percentage were entered into the database apart from those provided as ranges. Combining CF measurements resulted in only 35.5% of the database having CF measurements with many soil profiles and soil associations lacking any CF measurements.

Soil structure information was provided for 73.6% of the database. Of this, structureless soils dominated (1527 samples, 61.3%) followed by granular (643 samples, 25.8%), subangular blocky (442 samples, 17.8%), platy (404 samples, 16.2%), then prismatic (six samples, 0.2%). With this, some samples were labelled as transitions between two structure classes (e.g., platy — subangular blocky), thus, these samples were provided with two structure classes.

D_b measurements within the database were generally sparse (937 samples, 38% of database). These measurements followed theoretical expectations for the most part in that (i) D_b for mineral soil horizons <2.4 g·cm⁻³ and (ii) D_b typically increased with increasing depth (Fig. 6).

Of the physical properties assessed, Table 12 summarizes the average values associated with each master horizon with dominant subscript to provide a broad representation of the variability of soil properties. From these, it is apparent that clay content is higher in gleyed and Bt horizons, CF content and D_b increase with increasing depth, SOM content decreases with increasing depth, is lower in eluviated horizons and higher in illuviated horizons (as expected), decreasing D_b (increasing SOM) increases FC and PWP and that, of the soil colloids, SOM has a stronger impact on FC and PWP than clay content.

Table 12.

Summary of average soil property values associated with each individual master horizon and master horizon with dominant subscript(s).

Discussion

The rationale for a province-wide compilation of soil survey reports stemmed from the need for understanding how underlying soils vary, both locally and regionally, for better land and resource management. Having one harmonized database for all soil information, connected to a spatial map defining soil boundaries, supersedes that of having to review and compare multiple stand-alone soil surveys conducted over a broad period. Such a database allows users to quickly gain access to all available soil information found within the original soil surveys without having to access each survey individually. In addition, compiled soil information can be utilized for modeling the relationships between soil properties and soil-forming factors instead of having to manually compile the data, significantly reducing the pre-processing time required to complete analyses. Finally, with growing concerns around climate change, a harmonized database allows for the determination of carbon storage by soil type at a much larger scale than county-by-county (Carré et al. 2007; Aksoy et al. 2009; Grimm and Behrens 2010; Poggio et al. 2013).

The interest in amalgamating soil survey data into a harmonized database also stemmed from the past efforts in combining soil information into harmonized databases for numerous applications. For example, Velmurugan et al. (2009), Dobos et al. (2010), Sulaeman et al. (2013), Kristensen et al. (2019), and Lark et al. (2019) utilized soil profile data from different sources to develop harmonized soil profile databases for application in digital soil mapping (DSM). Alternatively, the Soil Survey Staff within the United States Department of Agriculture’s Natural Resource Conservation Center harmonized soil survey information into two databases, Soil Survey Geographic (SSURGO) Database and Web Soil Survey, each with online spatial applications. Such databases allow users to have full access to all soil survey information for any geographic location (Soil Survey Staff 2020a, 2020b). Additionally, the International Institute for Applied Systems Analysis (IIASA) and the United Nations’ Food and Agriculture Organization (FAO) addressed the need for a global, standardized database representing soils from around the world (Nachtergaele et al. 2010). These studies demonstrate the need and applicability of harmonized soils databases for identifying soils for different land uses and DSM research.

Although the framework introduced in this study is straightforward, it presents a novel framework to harmonizing soil surveys into a multi-purpose database in a Canadian context where soil surveys have yet to be amalgamated into a single database for application. This framework can be applied to other provinces and territories or built upon by incorporating additional soil surveys to develop regional, or a national, database for available soil information. This process is more straightforward if the original soil surveys adhere to the Expert Committee on Soil Survey (1982) and the Soil Classification Working Group (1998). Table 13 outlines the availability of both detailed and reconnaissance soil surveys on a provincial/territorial basis across Canada. From this, it is apparent that there is a substantial amount of information that could be combined into harmonized and standardized databases with many applications.

Table 13.

Overview of available detailed and reconnaissance soil surveys for each Province/Territory across Canada with the range in vintages.

The amalgamated database for NB supersedes that of the SNB and FSNB reports. The limitation with these original soil reports is that they aggregate various soil profile information into individual, generalized profile summaries for each soil associate in each soil association (or forest soil associations for FSNB). Doing so is counterintuitive because it loses much of the inherent variability in soil properties found across soil-forming factors as they vary across the province. Soils are complex in nature, and generalizing this information can lead to misleading interpretations of the data as well as broad spatial delineations of each soil association. Thus, much information is lost with this form of aggregation. Unlike the SNB and FSNB reports, the NSDB maintains multiple soil profiles for each soil associate, but again, this information is summarized losing some of the inherent variability of the soil properties found within each soil associate. For example, the NSDB provides two soil profiles for the Johnville soil associate (imperfectly drained member of the Holmesville soil association), whereas the harmonized database provides 11 soil profiles for the Johnville soil associate. Unlike the SNB, FSNB, and NSDB data, the amalgamated database presented in this study maintains the variability found within each soil association and soil associate and presents the variability in an organized manner. This is important to the end users who, with all soil information made available, may be better able to interpret what is seen on the land base, whereas the aggregated and generalized summaries provided in the SNB and FSNB reports may prevent this.

All the outlined efforts demonstrate the need for standardized protocols for collecting, and recording, soil information. Systematically sampling soils, regardless of topographic position, land use, and drainage class, would provide a much more robust data set for use in future endeavours. With the standardization and harmonization processes applied, there are limitations with the database. First, it is apparent from these efforts that the type, and amount, of data collected at each sampled location remain inconsistent (Table 14). For example, slope steepness and position were only determined for 95 soil associates. Also, horizon-specific measurements were inconsistent in measurement frequencies. D_b measurements were made more frequently than CF content. Second, limitations arise when assigning the middle class to a range in values as well as assigning one value to a categorical description. For example, if the soil texture was assigned as “clay loam” the sand content can vary from 20% to 45%, whereas clay content can vary from 27% to 40% based on the texture ternary diagram. As such, assigning the middle point of 32% for sand and 32.5% for clay may result in misinterpretation of the class, or add error when utilizing the data for analyses. The same holds true for the drainage classes and CF content descriptions within this database. Finally, climate information was unavailable for each soil associate within the original surveys, thus, and although an important soil-forming factor, climate information is not included within the database. This is particularly important because the database houses soil profiles from across the province where climate varies from the lowlands in the east to the highlands in the north and south of the province (Pronk and Allard 2003).

Table 14.

Overview of measured soil attributes, including general characteristics and horizon-specific soil properties, within amalgamated database with overall completeness.

An inconsistency which could not be addressed in this amalgamation and harmonization procedure is the inconsistencies in sampling size of different soil associates for developing soil surveys. Most soil associates have more than one profile described, depending on frequency of occurrence within different surveys. For example, the Holmesville soil association occurred in eight soil surveys. As a result, the well-drained soil associate (also called Holmesville) has 18 profiles within the database. Its moderatelywell-drained associate, Johnville, has 12 profiles within the database, followed by the poorly drained associate, Poitras, also with 12 profiles. It is common for some soil associations to occur within different surveys, and therefore, have multiple profiles within the database. In contrast, some of the less-common soil associations lacked a single soil profile altogether (e.g., Aulac, Babineau, Becaguimec, Belledune, Big Bald Mountain, Blackland, Caissie, Bottomland, Catamaran, Clearwater, Escuminac, Jacquet River, Kingston, Research Station, and Tetagouche). To have a fully comprehensive representation of forest soil conditions across NB, soil profiles are needed for these soil associations.

In addition to inconsistencies within the database, many issues became apparent with the spatial representation of soil association boundary delineations across NB, including

Table 15.

Surveyed soil associations not currently represented in the province-wide spatial soil association data set.

The soil database, amalgamated from 17 soil surveys for NB, Canada, is intended to provide a comprehensive overview of forest soil conditions as well as provide a framework for amalgamating and harmonizing soil survey information. Through careful cross-referencing, all data entries were examined to ensure they coincided with soil association and horizon-specific expectations as outlined in the Soil Classification Working Group (1998). Although amalgamated and harmonized, the aspatial database remains incomplete in terms of measurement gaps for horizon-specific physical and chemical properties. Although additional soil properties included within the original surveys were also entered into the database, emphasis was placed on those which are known to have strong impacts on rooting as well as those typically used as predictors in developing pedotransfer functions (PTFs) (Table 14). Future work focuses on standardizing individual soil properties to (i) correct for inconsistent units of measurement and (ii) predict absent soil property values by way of PTF.

Of the available soil surveys (both included and excluded from the database), 53.5% of NB has survey coverage with the spatial coverage of the surveys utilized in this study only representing 39.2% of NB (28 000 km²) (Fig. 3). Thus, additional efforts are required in updating soil representations across the province, either by soil association delineations or digital soil mapping of individual soil properties.

Conclusions

This study demonstrated means to amalgamate and harmonize soil information found within county-based soil surveys for the Province of New Brunswick, Canada, as a case study. This procedure, utilizing 17 soil surveys, has resulted in an amalgamated database containing 106 soil associations, 243 soil associates, and 522 soil horizon sequences (profiles). This framework demonstrates techniques which can easily be adapted to other locations in which soil surveys had been conducted over long periods of time. Such implementation allows for more consistent and standardized soil information at provincial, or even national, scales. This study expressed techniques to address and correct the prominent inconsistencies arising from amalgamating soil surveys. To summarize, this study corrected for

Creation of this database expedites PTF development for gap-filling and summarizing and quantifying both soil association and associates via similarities and differences. Once complete, this will enable spatially re-digitizing the updated database using already-existing soil association delineations, followed by revising these to ensure topographic mapping consistencies.

The amalgamated and harmonized database represents a way forward for large-scale soil studies and continuing with, otherwise discontinued, soil surveys. Many provinces house an array of soil surveys (Table 13) with which the amalgamation outlined in this study can be applied to develop either province-specific or an updated national soils database. Such databases can be used for applications in digital soil mapping, modeling soil property relationships, as well as spatial re-delineation of soil associations and associates, based on updated understanding of soil-forming factors. In combination, this information can assist users in understanding the spatial variability of soils, how soils vary with changing land uses, carbon stock prediction under different climate change scenarios, and asset management in terms of soil erosion and sedimentation modeling.