Field Monitoring and Modeling of the Hygrothermal Performance of a Cross-Laminated Timber and Wood Fiber-Insulated Building Located in a Cold Climate ^†

Abstract

The increased complexity of buildings has led to rigorous performance demands from materials and building envelopes. As markets for low-carbon, renewable construction materials grow, cross-laminated timber and wood fiber insulation have emerged as promising alternatives to meet these rigorous demands. However, an investigation into the performance and interaction of materials within high-performance systems is necessary to determine the durability risks associated with increased complexity and the introduction of new materials. This is important in order to ensure that these materials can meet the required functions of the building while taking advantage of their environmental benefits. To do so, this case study investigated a building constructed of cross-laminated timber and wood fiber insulation in a cold climate (Zone 6A) (Belfast, ME, USA). During construction, the building was instrumented with temperature, relative humidity, and moisture content monitoring instrumentation through the envelope, i.e., wall and roof assemblies. The conditions within the envelope were monitored for a two-year period and used to calibrate a hygrothermal model, along with measured material properties. The calibrated model was used to conduct a 5-year simulation and mold risk assessment. Findings demonstrated that there was no moisture or mold risk throughout the monitoring period or simulation. This supports the integration of cross-laminated timber and wood fiber insulation in sustainable building practices, particularly in cold climates where moisture management is critical.

1. Introduction

The strive for total occupant comfort and increased energy efficiency has established demanding performance requirements of building systems. Building and construction design, materials, and techniques have evolved in an effort to meet these requirements. As these complex composite building envelopes are designed, material properties and interactions must be considered to ensure durable, efficient, and healthy buildings.

New building materials give designers more options to meet stakeholder objectives. Current innovations are centered around solutions that meet or exceed the performance of traditional materials, while being environmentally benign. Two such materials are cross-laminated timber (CLT) and wood fiber insulation (WFI). Both are well-suited for North American adoption, as there are ample wood baskets and a history of using wood and wood-based materials in construction.

CLT was developed in Austria during the latter half of the 20th century and was launched in European markets during the 1990s [1]. The focus on bio-based materials and the need for alternative wood-based products and systems was critical to the success of CLT in Europe and led to its introduction to North America in the early 2000s [2,3]. As of early 2025, over 500 CLT buildings have been constructed in North America, with more under development [4]. Throughout the acceptance process, research efforts have determined many benefits of CLT, including dimensional stability, high strength-to-weight ratio, thermal mass, and collective research has worked to provide information on how to implement these systems despite the lack of comprehensive code adoption in the U.S. [5].

WFI was developed in 1908 by the Heraklith company and was used as loose-fill insulation until advancements in the refinement and pressing processes led to greater manufacturing capacity nearly 50 years later [6,7]. North American manufacturing of WFI was not realized until 2023, and therefore, there has been limited investigation into WFI in North America. Previous barriers to acceptance in North American markets have been attributed to three factors: lack of manufacturing capacity, limited knowledge of material properties, and few proven systems within North American climates [8]. Efforts to address these challenges are underway, as North American manufacturing began in 2023, and material properties have recently begun to be compiled [9,10,11,12]. These efforts have been undertaken on a limited number of WFI materials, with limited consideration of performance under real-world conditions. Research has been conducted in European climates; however, climate, building, and construction methods are not directly comparable between the two continents.

North American regions experience a wide range of climatic conditions and typically employ light-frame residential construction, whereas European nations typically employ masonry structures, resulting in differing envelope composition and material interactions [13]. Additionally, code requirements must be navigated and can limit the use of WFI. For instance, WFI is restricted by combustibility requirements. Currently, WFI can be used within Type V construction in the U.S. without limitations; however, fire blocking and encapsulating the WFI are required in Type III or Type IV construction [14]. Further, previous assessment of performance has employed insulation materials of different types (loose-fill and board) and thicknesses, ranging from 20–300 mm, which can make conclusions difficult to relate to general use cases [15,16].

The utilization of insulation materials and their interaction with wooden structural materials can lead to increased building and envelope energy efficiency, thermal performance, and durability. However, the majority of available fossil-based insulation materials, for example, expanded polystyrene (EPS) and extruded polystyrene (XPS), on the market are vapor impermeable and can introduce moisture risks. This is of great importance when considering that failures relating to the building envelope, specifically moisture intrusion, are well documented as a leading cause of construction litigation [17]. Moisture-related issues are associated with reduced durability, material degradation, and mold formation, which have been linked to human health problems [18]. Using vapor-permeable insulation materials, such as WFI, in building envelopes would be a possible solution to effectively manage moisture migration and reduce associated risks.

WFI and CLT are hygroscopic in nature and therefore can absorb and desorb water vapor under real climactic conditions [19]. Therefore, assemblies constructed with entirely wood-based materials are considered vapor open so long as a vapor barrier is not included in the assembly. The ability of the assembly to react to environmental conditions in this manner can prevent the accumulation of moisture at any one location due to the inclusion of an impermeable layer. However, durability and thermal properties are affected by the presence of moisture in these materials, and it is necessary to understand the impact of this relationship on performance. It should be noted that consideration of the moisture risk should be undertaken regardless of the material type, as hygroscopic materials are of no greater risk to moisture intrusion than other material types. Although moisture intrusion is not necessarily more likely to occur in wood-based materials, they are sensitive to the presence of moisture and can support mold growth. Mold growth can occur under specific conditions: on wood surfaces with a MC ranging from 30 to 150 percent, a temperature between 0 and 40 °C, and where nutrients are accessible. It is important to ensure that these conditions are not maintained.

Previous work has established the best practices for moisture considerations and CLT-fossil-based insulation wall and roof systems. The U.S. edition of the CLT Handbook [20] has served as a reference for design and been used to inform building codes. This text does not include the consideration of bio-based insulation materials and only discusses the use of systems with XPS, EPS, polyisocyanurate, and mineral wool. However, useful information for designers is provided when considering the placement of insulation, differences in ventilation strategies, and which climates these strategies are effective in. In these examples, CLT is always insulated to the exterior, which protects it from temperature changes, and therefore, dimensional instability.

Additionally, the inclusion of a ventilation gap, often used as a drainage plane between siding and insulation layers, reduces the risk of moisture accumulation within the insulation. This reduces inward vapor drive, which can adversely affect the durability of the structure. Although WFI is not considered in the handbook, these principles can be applied to CLT and WFI systems to reduce risks of interstitial condensation. Traditionally, the efficacy of new materials and their system interaction has been determined based on empirical evidence. However, with the increased system demands, introduction of new materials, changing code requirements, and occurrence of catastrophic moisture-related failures, methods to supplement empirical assessment have been developed [21]. Hygrothermal simulation software programs, such as COMSOL, DELPHIN, and WUFI, can be used to assess the effects of boundary conditions and material selection on an envelope system. These programs have been widely used and validated using laboratory and field studies [21,22,23,24,25]. Utilization of these tools is especially useful during material development and design of the envelope system to understand moisture transfer, the effects of boundary conditions, and to determine potential risks. These software tools can be useful for designers and architects when there are few or no existing examples of a building system. These complex tools use a simple user interface that requires a limited number of standard material properties [21]. Despite validation and complex computational methods being incorporated into the simulation tools, there are still uncertainties stemming from model assumptions, material properties, and boundary conditions [24,26]. This can be further exacerbated when modeling hygroscopic materials, and careful consideration must be taken when using modeling tools to assess performance [27].

Material properties are often cited as one of the most important inputs to be accurately reported when using these modeling tools [27]. Material data sheets can be the first resource for obtaining information; however, these are not always available or provide sufficient information. Direct measurement is regarded as the preferred method to gather these input properties, but this can be costly and time-consuming [15]. Further, although measured properties are often the most representative, shortcomings of the measurement methodologies and uncertainty in measurement can still develop errors within the models [15,22]. Each material is subject to different transport mechanisms, which cannot all be accounted for in a simplified model. Measurement practices and prediction techniques have been found not to be representative of all materials, especially considering hygroscopic materials. For example, the measurement technique for the determination of the moisture storage function, which was developed using mineral-based materials, is not representative of hygroscopic materials, which can lead to an overprediction of the rate of water absorption in wood-based materials. This can be associated with water vapor diffusion being assumed negligible during water uptake measurement and is counted twice in the model [22]. Other limitations have been observed when using Fick’s law to model the hygrothermal behavior of wood, where transient conditions and a retarded vapor sorption at high relative humidity (RH) can be seen [28,29,30].

Ideally, hygrothermal models should be compared to measured data to assess the quality of the model. This is necessary to assess what has the greatest influence on the model and can be used to demonstrate risks or shortcomings of the model and system. Material properties may be required to be adjusted to more accurately describe a material and envelope. Previous work has determined that material inputs may need to be adjusted to improve models, which is not necessarily indicative of the properties being incorrect but can be related to material interaction [22,24,26,28,31].

This study aimed to conduct a hygrothermal assessment of a building envelope comprised of CLT and WFI in a cold North American climate. Due to the fairly recent adoption of CLT and WFI in this region, there are only limited instances where these materials have been used, both on their own or in conjunction with one another. Development of accurate material property datasets can assist in the assessment of the moisture performance and ability to accurately represent the material within hygrothermal modeling tools, thereby helping to increase the adoption of these sustainable materials. As a part of this work, a 92 square meter building located in Belfast, Maine, U.S., was instrumented with 59 temperature, relative humidity (RH), and moisture content (MC) sensors throughout the wall and roof envelope. Monitoring data were collected over a 2-year period. Hygrothermal simulations were performed using on-site weather data and a one-dimensional hygrothermal simulation software tool (WUFI Pro 7.0). The inputs of WFI insulation material were measured by conducting laboratory experiments. The monitored and simulated data were compared and used to assess the mold growth index and durability of the building.

2. Materials and Methods

2.1. Case Study Building: Cornerspring Montessori School Annex

A one-story school building constructed in 2021 and located in Belfast, Maine, USA, was selected for this case study, as shown in Figure 1. This 92-square-meter building contained two classrooms, an entryway, and a bathroom. Panelized construction techniques were used to construct the building. Panels were manufactured using CLT and WFI. Both the CLT and WFI were of European origin, and samples of the WFI were used to determine material properties.

The school was regularly occupied September through June on weekdays from 8:00 am to 4:00 pm by 5–10 persons. The building operated exclusively on electricity, using LED lighting, an electric hot water heater, and two mini-split heat pumps, accompanied by an energy recovery ventilator, to condition the building. The building employed large amounts of insulation, with insulation depths of 240 mm in the walls and 360 mm in the roof, which resulted in wall and roof nominal U-values of 0.12 W/(m²·K) and 0.08 W/(m²·K), respectively.

Wall and roof panels consisted of three- and five-ply CLT, respectively, and multiple layers of 120 mm thick WFI board. The WFI board had a nominal density of 140 kg/m³, did not include wax, and was rated for internal use. Therefore, a water-resistive barrier (WRB) was applied to the exterior surface of WFI in wall assemblies. As part of the fabrication process, wall and roof panels were instrumented with temperature (°C), RH (%), and MC sensors (%) throughout the envelope, as shown in Figure 2. CLT was exposed to the interior and was finished using an oil-based finish.

Sensors were installed at two densities, as shown in Figure 2 and Figure 3. Primary cluster locations within the monitored walls included temperature and RH sensors at each WFI interface and MC sensors at the first and third lamella of the CLT, labeled P1 to P3 from interior to exterior. Secondary cluster locations include sensors at the interface between CLT and WFI and the WRB and WFI, labeled as P1 and P3, respectively. Similarly, the roof locations were instrumented with sensors in between each layer of WFI and in the first, third, and fifth lamella of the CLT, labeled R1 to R4 from interior to exterior. Measured data were collected over a 2-year period, beginning 1 January 2022.

2.2. Monitoring Instruments

Building monitoring instrumentation recorded conditions within the building assembly at 5 min intervals, and weather data were recorded at 15 min intervals. A description of the technical specification of the monitoring equipment is presented in

Table 1. List of technical specifications of monitoring equipment.

Instrument	Parameter(s)	Accuracy	Resolution	Range	Manufacturer	Model
T&RH	Temperature	±0.3 °C	0.1 °C	−40 to 120 °C	Omnisense	A-1 Humisense
	Relative Humidity	±2.0%	1%	0 to 100%
MC	MC	-	0.1%	-		A-2Moisture
Weather station	Temperature (Interior)	±0.3 °C	0.1 °C	0 to 60 °C	Davis	Vantage Vue, Wireless
	Relative Humidity (Interior)	±2%	1%	1 to 100%
	Barometric Pressure	±10 Pa	10 Pa	54,000 to 110,000 Pa
	Temperature (Exterior)	±0.1 °C	0.5 °C	−40 to 65 °C
	Relative Humidity (Exterior)	±2%	1%	1 to 100%
	Dewpoint	±1 °C	1 °C	−76 to 54 °C
	Wind Speed	±1 m/s	0.5 m/s	0 to 89 m/s
	Wind Direction	3°	1°	1 to 360°
	Precipitation	±4%	0.2 mm	0 to 6553 mm, daily

.

Temperature and RH sensors were probe-type sensors, and MC sensors were pin-type sensors. An example of these sensors and the methods of installation are presented in Figure 4. Temperature and RH sensors were situated between layers of WFI and/or WFI and CLT (P1, P2, R1, R2, and R3). At the exterior surface (P3 and R4), a small channel was routed in the WFI, which fit the sensor, and a piece of tape was placed over the bottom of the sensor, so as not to block the probe, which ensured that the sensor stayed in place during the panel manufacture and installation. MC screws were placed 3 cm apart, and sensors were calibrated based on lab measurements. Screw length was determined in order to reach the center of the lamella, which was being measured. The location and depth of the screw were determined based on the location and cluster density. Note that after initial installation, MC sensors at primary locations in the middle lamella, which were installed using nails as pins instead of screws, within the walls and roof, were found to be faulty. Where accessible, these were replaced; otherwise, additional MC sensors were installed in the attic. However, as these were not installed at the same location and were not installed until 8 months into the monitoring period, they were not included in the analysis and only used to assess general trends. Roof assemblies did not have this issue, and MC was used for the calibration; however, one of the four roof sensor clusters was damaged during panel installation, and therefore, the data were not recorded at all sensors at that location.

Hourly temperature and RH values were obtained by averaging 5 min readings. Intermittent outages were experienced throughout the 2-year period. Gaps in data were filled using the preceding data, e.g., if data were missing from 12 December 2022, at 2:00 pm, the previous hour would be used. Minimal outages were recorded over the 2 years: 5 h at the north wall, 56 h at the south wall, and 0 h at the roof location.

Belfast, Maine, USA (IECC Climate Zone 6A) is characterized by long, cold winters, with an average of 6230 annual heating degree days and 514 annual cooling degree days recorded from 1997–2024, using a base of 18.3 °C [32]. Cold temperatures were recorded throughout the monitoring period, with an average exterior temperature of 8.9 °C throughout the 2-year period.

Weather data, including both outdoor and indoor information, was recorded on a 15 min basis and averaged to obtain hourly measurements. Outages associated with weather data were more frequent than the measurements within the assemblies. Gaps in measured interior conditions were filled using the same approach as the assembly temperature and RH monitoring, by using the preceding time period. However, gaps in exterior conditions were filled using weather data from the closest available source, an airport approximately 39 km away (Knox County Regional Airport). Additionally, weather data were supplemented with solar radiation values from the National Renewable Energy Laboratory (NREL) National Solar Radiation Database (NSRDB) [33]. The weather data were used as the boundary conditions for the hygrothermal modeling.

2.3. Properties of Materials Used in Hygrothermal Simulations

Material properties were determined through a combination of manufacturer data sheets, laboratory evaluation, and materials that were available in the Fraunhofer IBP WUFI material database. The single condition thermal conductivity (21 °C, 55% RH), porosity, and water vapor permeability of WFI were empirically derived [10]. To support this work, further laboratory testing of WFI was conducted, including thermal conductivity (10 °C, oven-dry), moisture storage function, and water absorption coefficient. These were conducted using the following standards: ASTM C518, Standard [34] Test Method for Steady-State Thermal Transmission Properties by Means of the Heat Flow Apparatus; ASTM C1498 [35], Standard Test Method for Hygroscopic Sorption Isotherms of Building Materials; and ASTM C1794 [36], Standard Test Methods for Determination of the Water Absorption Coefficient by Partial Immersion. Select material properties are presented in

Table 2. List of materials and selected properties.

Material	Density	Thermal Conductivity	Porosity	Water Vapor Diffusion Resistance	Specific Heat Capacity
	(kg/m3)	(W/mK)	(m3/m3)	(-)	(J/kgK)
CLT b	426	0.1013	0.836	291	1237
WFI c	140	0.0346 a	0.905	2.4	2100 e
WRB e	210	2.3	0.001	85	2300
Siding b	426	0.1013	0.836	291	1237
ZIP System d (Roof Decking)	607	0.101	0.950	211	1880
ZIP System d (Roof Membrane)	130	2.3	0.001	2526	2300

^a Laboratory measurement; ^b Glass et al., 2024 [22]; ^c Snow et al., 2024 [10]; ^d Fraunhofer IBP WUFI material database; ^e Manufacturer specification.

, and detailed material specifications are provided within the Supplementary Materials, Tables S1–S12.

2.4. Hygrothermal Simulations and Data Analysis

Hygrothermal simulation was conducted for a two-year period using the one-dimensional WUFI Pro software (version 7.0). The WUFI software has been widely validated for assessment of building envelopes and materials [8,21,22] and is accessible and commonly used by both researchers and practitioners. The cross-section of the wall assembly was created in the software interface, with monitoring positions that matched the monitoring positions’ locations in the case study building. Simulation outputs of temperature, RH, and MC at the selected positions were compared to measured data using the root mean square error (RMSE), Equation (1), and mean absolute error (MAE), Equation (2). The use of both metrics allowed for the determination of the magnitude of the errors in prediction, as RMSE is sensitive to large errors, and MAE is sensitive to small, continual errors [15,24,37]. These values were calculated for both the temperature and RH at each monitoring position, and MC where applicable. Scaled error (SE), Equation (3) (a-wall and b-roof), was also used to assess the goodness of fit across all measurements. This function was developed using the error associated with each measurement: temperature was measured to ±0.3 °C, and RH was measured to ±2%. This type of assessment has been conducted in previous work to assess the goodness of fit for hygrothermal models [32]. The RMSE was calculated at all positions and each measurement type, then divided by the representative error, and added, with equal weighting across all of the monitoring positions.

$$\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}=\sqrt{{\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{N}}{\left({\mathrm{x}}_{\mathrm{i}}-{\widehat{\mathrm{x}}}_{\mathrm{i}}\right)}^2}{\mathrm{N}}}}$$

(1)

$$\mathrm{M}\mathrm{A}\mathrm{E}={\displaystyle \frac{{\sum }_{\mathrm{i}=1}^{\mathrm{N}}\left|{\mathrm{x}}_{\mathrm{i}}-{\widehat{\mathrm{x}}}_{\mathrm{i}}\right|}{\mathrm{N}}}$$

(2)

$$\begin{array}{ll}{\mathrm{S}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{e}\mathrm{d} \mathrm{E}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r}}_{\mathrm{w}\mathrm{a}\mathrm{l}\mathrm{l}}& =\left({\displaystyle \frac{1}{6}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{T},\mathrm{P}1}}{0.3}}\right)+\left({\displaystyle \frac{1}{6}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{T},\mathrm{P}2}}{0.3}}\right)+\left({\displaystyle \frac{1}{6}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{T},\mathrm{P}3}}{0.3}}\right)+\left({\displaystyle \frac{1}{6}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{R}\mathrm{H},\mathrm{P}1}}{2}}\right)+\left({\displaystyle \frac{1}{6}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{R}\mathrm{H},\mathrm{P}2}}{2}}\right)\\ & +\left({\displaystyle \frac{1}{6}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{R}\mathrm{H},\mathrm{P}3}}{2}}\right)\end{array}$$

(3a)

$$\begin{array}{ll}{\mathrm{S}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{e}\mathrm{d} \mathrm{E}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r}}_{\mathrm{r}\mathrm{o}\mathrm{o}\mathrm{f}}& =\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{T},\mathrm{R}1}}{0.3}}\right)+\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{T},\mathrm{R}2}}{0.3}}\right)+\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{T},\mathrm{R}3}}{0.3}}\right)+\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{T},\mathrm{R}4}}{0.3}}\right)+\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{R}\mathrm{H},\mathrm{R}1}}{2}}\right)\\ & +\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{R}\mathrm{H},\mathrm{R}2}}{2}}\right)+\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{R}\mathrm{H},\mathrm{R}3}}{2}}\right)+\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{R}\mathrm{H},\mathrm{R}4}}{2}}\right)+\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{M}\mathrm{C},\mathrm{I}\mathrm{n}\mathrm{t}}}{1.5}}\right)\\ & +\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{M}\mathrm{C},\mathrm{M}\mathrm{i}\mathrm{d}}}{1.5}}\right)+\left({\displaystyle \frac{1}{11}}\ast {\displaystyle \frac{{\mathrm{R}\mathrm{M}\mathrm{S}\mathrm{E}}_{\mathrm{M}\mathrm{C},\mathrm{E}\mathrm{x}\mathrm{t}}}{1.5}}\right)\end{array}$$

(3b)

A second model was developed in order to perform a mold growth and durability assessment. This simulation was run for a 5-year period and utilized the calibrated material properties from the initial model. Weather conditions were used from typical meteorological year (TMY) data from Belfast, ME. The climate year was selected in accordance with ASHRAE 160, Criteria for Moisture-Control Design Analysis in Buildings.

2.5. Mold Growth Assessment

Mold growth assessment was performed using the WUFI Mold Index VTT software (Version 2.3), in accordance with ASHRAE 160, Criteria for Moisture-Control Design Analysis in Buildings. The model was developed from experimental observations, which developed the calculation of a mold index using temperature, RH, and material characteristics. The mold growth index is separated into three different levels: no or invisible mold growth (mold growth index less than 1), mold growth is visible only under a microscope (mold growth index between 1 and 2), and visible mold growth (mold growth index greater than 2). The mold growth assessment was conducted on both the measured and simulated datasets. The WFI was classified as a sensitive material and assigned a mold index decline coefficient of 0.1. The mold growth index was calculated at locations associated with wall and roof measurement positions, as well as the interior surface of the walls and roof.

3. Results

Measured and simulated results of the north wall monitoring position are presented below. A brief comparison of north and south monitored data is included below; however, all analyses of the south wall and roof monitoring and modeling are included within the Supplementary Materials, Figures S1–S9.

3.1. Monitored Data of Wall and Roof Assemblies

3.1.1. MC of CLT

MC was measured at locations throughout the building; data from the north and south walls, the north corner of the building, and the roof are presented in Figure 5. Recording from the middle lamella of the CLT is presented in the figure for the north and south walls. The north corner is representative of the center of the lamella, which is closest to the exterior of the building. As these sensors are pin-type, the measured data represent the maximum MC along the shank of the pin at a given monitoring location. The monitoring data presented for the roof include all three depths at the top left monitoring position of the roof (position closest to the peak of the roof, to be referred to as the north roof location, as shown in Figure 2). The North and South wall data were not included (dotted lines) for the months where sensor readings were unreliable due to the MC pin type. Limited differences in MC were evident between the monitoring locations and pin depths. In all cases, the weekly averaged MC fell in a narrow range, 7.5 to 10.2%. The highest recorded MC throughout the building was 13.4% in the south wall. MC of all panels was maintained below the 15–18% MC threshold, where fastener corrosion can occur [38,39]. Measured MC trends indicate annual seasonal changes, where the MC of the panels was lower during cooler months and higher during warmer months. Additionally, the MC trends indicate that the building envelope is drying towards an equilibrium, as the first year showed higher MC levels than the second year of monitoring.

3.1.2. Effects of Wall Orientation on Monitored Data of WFI

Measured temperature and RH of north and south wall sensors are presented in Figure 6. Distinct trends in monitoring data are evident. Conditions at the position closest to the exterior (P3) demonstrated greater seasonal variability than those at the positions closer to the interior (P2 and P1) for both temperature and RH. This is attributed to the thermal properties of the insulation and sensors being protected from extreme conditions as the depth of the insulation layer increases. Monitoring location P3 was protected from solar radiation and moisture; however, there was little to no thermal barrier. P2 and P1 were located under 120 and 240 mm of WFI, respectively, therefore reducing the temperature swings at interior positions. This difference in conditions under increasing insulation is represented through the difference between the maximum and minimum temperatures, which were 68, 37.7, and 23.6 °C at P3, P2, and P1, respectively.

This trend did not hold true for RH, where fluctuation was greatest at P3 (45.9% RH), then P1 (33.2% RH), and smallest at P2 (22.1% RH). Measured RH recorded at P3 was closely related to temperature trends, which follow exterior climatic conditions of low RH during warm months and increased RH during cold months. Seasonal variability of RH at P3 was associated with the relative ability of air to hold moisture at a given temperature, and the increase in RH during cold periods is indicative of air being able to hold less moisture than during warm periods. This phenomenon can explain why RH stabilizes across all positions during warm periods, where the temperature at all positions is similar. Additionally, greater variability in RH at interior positions was observed and associated with changes in the interior conditions. The RH at P3 is more closely related to the interior conditions of the building than the exterior conditions. CLT acts as a vapor retarder, which slows the movement of moisture throughout the assembly; however, it does not stop movement. This allows for inward vapor drive. As there is no humidity control in the building, this means that air is heated, and low humidity conditions draw moisture into the building through the envelope.

Wall measurements indicate limited variability associated with orientation. The south-oriented wall experienced higher temperatures than the north wall at all positions except for P1. RH measurements demonstrated similar trends, where the south wall had lower relative humidities than the corresponding north wall monitoring positions. This was expected, as the south wall was exposed to greater levels of solar radiation than the north wall. However, this was not true for P1, where recordings in the south wall were lower for temperature and higher for RH than those of the north wall. This was associated with the greater effect of interior conditions at P1 than exterior conditions. The greater variation in RH than that of temperatures can be attributed to seasonal variability and the effects of temperature on RH. When interior and exterior temperatures were similar, there were fewer differences in the RH distribution. This occurred during the summer months, when temperatures at the interior and exterior of the building were similar. During cold months, when the building was heated, the differences in relative humidity were greater. This can be attributed to the temperature dependency of RH, when, as temperatures decrease, so does the ability of air to hold moisture. Additionally, there was no RH control within the building, and during the winter months, the building was heated, further exacerbating the differences in RH. Temperature and RH of the exterior were closely related to P3, which experienced greater seasonal variation than P2 and P1. Instances where rainfall occurred led to an increase in RH, leading to periods where exterior RH is upwards of 20% RH higher than at P3. Trends in interior and exterior temperature are more closely related to monitoring positions within the wall assembly.

3.1.3. Effects of Wall Height on Monitored Data

Monitored temperature and RH at P3 in three vertical positions (top, middle, and bottom) of the north wall are presented in Figure 7. Temperature and RH stratification were minimal across vertical positions. Temperature was consistent across the top and middle positions (average 0.4 °C difference), with a minor decrease at the bottom position (average 1.6 °C difference). Similarly, RH was consistent at the top and middle positions (average 1.5% RH difference), with an increase at the bottom position (average 8.6% RH difference). This can be attributed to ground proximity and contact. Soil conditions are cooler and are constant moisture sinks, which is reflected in the slight difference between the bottom monitoring position. Additionally, P3 sensors were installed under the building’s WRB, which was directly underneath a 40 mm air gap, which acted as a rainscreen. The 40 mm air cavity was protected from insects using a mesh screen and was open to air and moisture exchange at the bottom, and an approximately 13 mm gap was left at the top of the wall between the cladding and fascia boards. This allowed air and water vapor to enter and exit the cavity and can explain the differences in the measured data.

3.1.4. Effects of Roof Slope on Monitored Data of Roof Assembly

The monitored temperature and RH at two roof locations (Figure 2, left) and all depths (R1 through R4) are presented in Figure 8. Trends similar to the wall monitoring data are evident, where the vertical location had a limited impact on the measured temperature and RH. Temperature was consistent across the monitoring positions, with the average difference at R1 being 1 °C, R2 being 0.74 °C, R3 being 0.21 °C, and R4 being 0.35 °C. This trend with low average differences was also evident in the monitored RH, with average differences being 4.3, 4.4, 2.9, and 0.1% RH at monitoring positions R1 through R4, respectively. Interior and exterior temperature and RH trends are similar to the wall monitoring data. Interior and exterior temperatures are closely related to the conditions within the roof assembly than RH.

3.2. Hygrothermal Model Development and Calibration

The north wall’s hygrothermal behavior was simulated using WUFI Pro software (Version 7.0). The results of MC in CLT and temperature and RH in WFI are reported separately. An initial model was first developed, and calibration was conducted to improve the model accuracy in predicting the temperature and RH in WFI.

3.2.1. Hygrothermal Model Development—MC in CLT

Measured MC at wall locations was not captured throughout the entire two-year period and was therefore not included in the SE calculation or calibration considerations. However, a comparison of the available data is presented in Figure 9. The simulation period began on 1 January 2021, and the measurement of MC was first available in August 2022. The RMSE and MAE values presented for MC use the shortened time period (25 August 2022–31 December 2023). The initial and calibrated models refer to different versions of the simulation. The initial model was simulated utilizing the measured material properties, and the calibrated model was simulated utilizing the results of the sensitivity analysis to adjust material properties and interior conditions. The initial model was able to represent trends in MC well, and the RMSE was 2.07% MC. Considering the low error rate and inability to accurately adjust initial parameters of the model in order to improve the error rate, this was excluded from the sensitivity analysis. MC was not considered in the SE, and therefore, limited improvement of the model is evident when considering MC.

3.2.2. Hygrothermal Model Development—Temperature and RH in WFI

Initial model results of the temperature and RH at P1, P2, and P3 are presented in Figure 9, Figure 10, Figure 11 and Figure 12. Overall, good agreement was evident at all positions, with the simulated results following general patterns in measured data at all positions. Temperature predicted aligned more closely with measured data at all positions when compared to RH results. Temperature and RH at P3 were predicted with greater agreement than at P1. This can be attributed to large variability in exterior conditions and limited separation between P3 and exterior conditions, where small errors are evident, and there are a few instances of larger errors. This is followed by the greater separation between interior conditions and P1, which is located at the interface between CLT and WFI, and an overprediction in the temperature and underprediction of RH. As previously noted, increased temperatures cause reduced RH values, which is represented in the model, where greater disagreement between measured and simulated RH was evident when temperature was at greater disagreement.

Observations were substantiated quantitatively through the use of RMSE, MAE, and SE, as shown in

Table 3. North wall’s summary of error between measured (M) and simulated (S) results.

Location	Metric	Data Type	Initial Model			Calibrated Model
			Ave	RMSE	MAE	Ave.	RMSE	MAE
P1	Temperature (°C)	M	19.2	3.38	2.56	19.2	2.32	1.84
		S	21.7			19.5
	Relative Humidity (%)	M	57.2	14.5	13.9	57.2	10.6	9.7
		S	43.4			47.5
P2	Temperature (°C)	M	15.3	1.68	1.31	15.3	1.90	1.51
		S	15.8			14.4
	Relative Humidity (%)	M	66.0	8.92	7.94	66.0	6.08	5.09
		S	59.1			62.4
P3	Temperature (°C)	M	11.4	3.18	2.41	11.4	3.23	2.39
		S	10.4			14.4
	Relative Humidity (%)	M	73.4	4.50	3.51	73.4	4.50	3.74
		S	74.0			75.6
Scaled Error			6.91			5.90

. Calculated RMSE of temperature and RH were greatest at P1, followed by P3, and then P2. This indicated that the greatest error was evident at P1, and P3 was influenced by the constant variability of exterior temperature, which led to errors. Summary of the errors between measured and simulation results align with the qualitative observations presented above. RMSE and MAE are similar at all locations; however, a comparison of these metrics at P3 demonstrated the key differences in what is represented. P3 RH RMSE was 4.50% RH, whereas MAE was 3.51% RH. This is indicative of errors associated with a few instances of greater variability, and not continual small errors. However, the discrepancy between calculated RMSE and MAE was small; therefore, only RMSE was selected to calibrate the model and included in the SE.

3.2.3. Model Calibration

Additionally, alternative methods were utilized to determine the source of error for calibrating the model, including adjustment of boundary conditions and sensitivity analysis (see detailed discussion in Section 4.1). A series of discussions with the building manager revealed that the two classrooms within the building were not occupied at the same rate. Interior conditions were measured using the temperature and RH sensor of the weather station, which was only representative of a single location in the room where it was placed. Further, the ceilings are sloped, with the ceiling height ranging from 3 to 4 m, which can lead to vertical temperature stratification. Therefore, the measured interior conditions at a single location were insufficient in representing the actual conditions near the wall surfaces at three monitoring positions with varying heights, resulting in a potential source of error. A series of simulations was run using the measured P1 and P3 as boundary conditions and only included the materials that are in between these monitoring positions (see Section 4.1). In order to address this temperature discrepancy effect, temperature shifts were conducted for the interior conditions, based on the P1 average temperatures. The differences between measured and simulated temperatures at P1 led to a difference and then a decrease of 2.5 °C. The absolute humidity was maintained the same during these shifts, which preserved the moisture profile of the assembly. The temperature shift is responsible for 51% of the error reduction at P1, or 15.4% of the 31.6% difference. The adjustment of interior conditions was then identified as a main source of error, and sensitivity analysis was also performed on the model to determine the effects of selected properties.

Another simulation was conducted using the transformed interior conditions in the calibrated model. The temperature and RH results of the calibrated model are presented in Figure 10, Figure 11 and Figure 12. Additionally, different time periods are displayed to provide detailed information about the improvements made to the model. The left column is full of two years of data; the middle and right columns are three-week periods during the summer and winter. This format was adopted for all positions, P1, P2, and P3. Temperature and RH tracks well at all locations; however, minor discrepancies are still evident. During the winter, the temperature is consistently overpredicted. However, during the summer, it is underpredicted. The simulation was able to accurately capture trends at all three positions.

In

Table 4. Summary of error between measured (M) and simulated (S) results using the calibrated model.

Assembly	Location	Metric	Averaged Results		Calibrated Model
			M	S	RMSE	MAE	SE
North Wall	P1	T (°C)	19.2	19.5	2.32	1.84	5.90
		RH (%)	57.2	47.5	10.6	9.7
	P2	T (°C)	15.3	14.4	1.90	1.51
		RH (%)	66.0	62.4	6.08	5.09
	P3	T (°C)	11.4	14.4	3.23	2.39
		RH (%)	73.4	75.6	4.50	3.74
	MC	MC (%)	10.1	8.20	2.07	4.26
South Wall	P1	T (°C)	18.5	18.9	3.09	2.54	8.28
		RH (%)	58.0	48.2	10.7	9.7
	P2	T (°C)	15.8	15.3	2.78	2.14
		RH (%)	62.0	58.5	5.44	4.39
	P3	T (°C)	12.7	15.3	5.87	3.82
		RH (%)	66.6	68.6	4.99	4.05
	MC	MC (%)	11.2	8.39	2.99	4.88
Roof Panel	R1	T (°C)	18.6	18.7	3.72	3.70	5.93
		RH (%)	56.7	57.8	4.6	3.9
	R2	T (°C)	16.5	16.5	3.76	3.66
		RH (%)	60.2	65.1	4.39	3.79
	R3	T (°C)	14.1	14.5	3.86	3.61
		RH (%)	63.8	71.8	4.19	3.58
	R4	T (°C)	11.4	12.4	5.13	3.46
		RH (%)	66.3	75.1	2.87	2.08
	Int	MC (%)	9.9	7.21	1.30	1.10
	Mid	MC (%)	8.7	8.90	0.99	0.81
	Ext	MC (%)	9.9	10.2	1.13	0.92

, greater agreement was evident when considering the RMSE, MAE, and SE. Error at P1 decreased greatly when compared to initial results, with a RMSE of 2.32 °C. RMSE increased at P3 when compared to initial results; however, the magnitude of improvement at P1 led to an overall reduction in SE. Errors associated with P2 remained similar to those of the initial results. Overall, changes to the model were better able to predict the conditions within the wall assembly, and the model was considered sufficiently calibrated for the mold growth assessment. In addition, the acceptance criteria of hygrothermal simulation for field monitoring are discussed in Section 4.2.

3.3. Mold Growth Assessment Using Calibrated Model

Mold growth assessment was conducted using two data sets: the monitoring data and a 5-year simulation. Simulation was conducted using the calibrated materials and a TMY weather file for Belfast, ME. The assessment included all monitoring locations (P1, P2, and P3) as well as the conditions at the interior surface of the CLT, which was in contact with the occupant. As stated in ASHRAE 160, different criteria are specified for surfaces that the occupant has direct contact with and those within the wall envelope. Passing criteria for interior surfaces requires the mold growth index to remain at zero throughout the entire period. Monitoring positions within the wall envelope have a greater tolerance but must not cross a mold growth index of 3 at any point during the time period. Results for the measured and simulated mold growth assessment are presented in

Table 5. Mold growth assessment of the north wall, south wall, and roof panel.

Assembly	Position	Measured (2 Years)		Simulation (5 Years)
		Maximum Mold Growth Index	Pass/Fail	Maximum Mold Growth Index	Pass/Fail
North Wall	Interior Surface	-	-	0.00	Pass
	P1	0.00	Pass	0.00	Pass
	P2	0.00	Pass	0.00	Pass
	P3	0.06	Pass	0.56	Pass
South wall	Interior Surface	-	-	0.00	Pass
	P1	0.00	Pass	0.00	Pass
	P2	0.00	Pass	0.00	Pass
	P3	0.00	Pass	0.05	Pass
Roof	Interior Surface	-	-	0.00	Pass
	R1	0.00	Pass	0.00	Pass
	R2	0.00	Pass	0.00	Pass
	R3	0.00	Pass	0.00	Pass
	R4	0.03	Pass	0.92	Pass

.

Mold growth assessment thresholds were not crossed at any time during the measurement or simulation periods. A mold growth index above zero was recorded at P3 in both instances; however, neither instance caused failure due to not crossing the thresholds, and the short duration, where the index was above 0. The mold growth index above zero at P3 was associated with rapid changes in temperature and RH, where conditions for mold growth were met for short periods of time. Although the mold growth index at P3 was greater than 0 for both measured and simulated data, this correlates with mold growth occurring at levels that are not visible without a microscope. Increases in the mold growth index occurred cyclically during cold months, with a decreasing magnitude, where the greatest potential for mold growth occurred during the first winter and decreased as the assembly reached equilibrium. Additionally, the predicted mold growth index was greater when assessing the simulation data than the measured data. This can be attributed to the overestimation of RH at P3 in the simulation results, as well as the harsh weather conditions associated with the weather file used in the mold growth assessment.

4. Discussion

Monitored data were collected over a two-year period and demonstrated that there were minor differences between monitoring locations in the walls and roof. This led to the development of a hygrothermal model to assess the performance of the system. The RMSE between the modeled and measured data were reduced to 5.87 °C and 10.7% RH at the greatest across all models. Further, the SE in all models decreased. The largest improvement was evident in the roof model, where SE was reduced from 9.35 to 5.93. At this threshold, we considered the model to be calibrated. The developed model was considered effective in predicting hygrothermal performance and assessing mold growth using historical weather data. However, the authors did not find any existing standardized criteria for assessing model effectiveness. Further discussion is given to the sensitivity analysis of model inputs carried out by the authors and the criteria for model accuracy derived from the literature, aiming to provide stronger evidence of the model’s effectiveness and clarify the criteria applicable for field studies.

4.1. Sensitivity Analysis of Model Inputs

The hygrothermal model inputs can be categorized as (1) material properties of the components in wall and roof assemblies, e.g., permeability of WFI and WRB, and (2) boundary conditions, i.e., interior temperature and exterior solar absorption. Sensitivity analyses can reveal how changes in input variables affect a model’s output accuracy. In this study, the sensitivity analysis was performed to determine the effects of material selection (where other similar materials were available within the Fraunhofer IBP WUFI database) and weather data, including solar absorption and permeability, as shown in Figure 13. This yielded minimal changes in simulation results; however, changes to the exterior solar absorption and permeability of the exterior and interior surfaces were accepted and included in the final model. In most cases, the changes to material properties led to greater error within the model; however, two changes were accepted: an increase in the air change rate of the gap behind the siding and a reduction in the solar absorptivity of the siding. The model was of greatest sensitivity to the air change rate within the air cavity, which affected the direction of moisture movement. Results of the sensitivity analysis are included within the Supplementary Materials.

4.1.1. Effects of Material Properties on Simulation of T, RH, and MC

The hygrothermal simulation tools use basic material properties to predict heat, air, and moisture transfer. Although these methods have been validated, consideration of the material properties and understanding of the shortcomings of these tools needs to be considered. First, the models are based on the assumptions in Fick’s law to model vapor transport and Darcy’s law to model liquid transport under a vapor pressure difference or MC gradient, as well as coefficients that describe the material’s permeability. This can create discrepancies when considering porous materials, especially at high RH and transient conditions [28,29,30]. Researchers have found that sorption is reduced at higher RH, which can lead to overestimation of moisture in wood when not modeled with a dynamic sorption curve. Further, this can impact the influence of predicted RH, as variability in the changes of vapor pressure can also affect the behaviors. The limitations within the assumptions of the modeling tools can explain errors associated with RH during periods with high RH and cold temperatures.

The moisture storage function is one of the most influential material properties to classify. In wood and wood-based materials, the relationships between the equilibrium MC and RH are often referred to as the sorption and desorption isotherms. Wood materials exhibit what is called the hysteresis effect, where the desorption isotherm is greater than that of the adsorption isotherm [40]. Current hygrothermal modeling tools do not consider the effects of hysteresis [41]. In an effort to capture this phenomenon, an averaged isotherm was used in the simulation. Furthermore, the effect of the hysteresis loop was considered in the sensitivity analysis; however, differences between the averaged, adsorption, and desorption isotherms in WFI were found to be negligible on the model results. This aligns with previous works assessing the hygrothermal behavior of wood fiber insulation, where the differences between EMC during adsorption and desorption ranged from 0.85% to 1% MC at 10% and 97% RH, respectively [12]. When implemented, differences of approximately 1% MC have a limited impact on the model.

Although there were shortcomings associated with the modeling tools and ability to represent the building envelope accurately, there were key trends that can help to aid design and consideration when using WFI. Previous works have demonstrated that vapor open assemblies are influenced by interior conditions more than they are influenced by exterior conditions. This is evident through the requirement to shift interior conditions to increase the accuracy of the model. Freimanis et al. found that systems that were vapor open, with a low vapor diffusion resistance, align their RH behavior with temperature profiles. Vapor-tight systems exhibit reduced RH amplitudes as vapor diffusion resistance decreases [23]. This same study found that when both indoor RH and outdoor temperature oscillate, vapor-open systems with low vapor diffusion resistance resemble indoor humidity profiles more closely. Additionally, investigation of cellulose fiber insulation demonstrated that the interior conditions could impact the conditions within the wall in vapor open assemblies, as there is the potential to dry out both the interior and exterior [42]. Connections between the interior conditions of the building and P1 in the wall are evident in the monitoring data. P1 temperature and RH were influenced by the interior conditions more than the positions to the exterior (P2 and P3). This was also evident during the modeling process, where shifts in the interior conditions translated to the greatest source of error across the monitoring positions.

4.1.2. Effects of Boundary Conditions on Simulation of T, RH, and MC

Interior boundary conditions were found to have the greatest impact on improving the accuracy of the model; however, adjustment of solar radiation was necessary in all cases. A solar absorptivity of 0.6, which is representative of grey paint, was selected for the siding material. Additionally, the black metal roof was assumed to have an absorptivity of 0.8. In all cases, reducing the solar absorptivity increased the accuracy of the model. This demonstrates that the exterior boundary conditions can impact the results of the model and have minor implications on the performance of the building. The small differences can be related to the thermal properties of the envelope, which are not sensitive to small changes in climatic conditions. Many of the effects of the boundary conditions are associated with the local climate, where cold temperature extremes occur during the winter and warm temperatures are recorded in the summer. This led to similar trends throughout the two-year period. Temperatures within the assembly are similar at all positions during the summer months, as the temperature difference between the interior and exterior conditions was small. Conversely, large temperature differences between interior and exterior conditions led to a large temperature gradient throughout the wall assembly. This is also mirrored within the RH data.

The high thermal resistance of the envelope can also lead to differences in the hygric performance in response to exterior conditions. The model was able to predict general trends related to MC in the assemblies; however, the assumption of instantaneous equilibrium can explain the discrepancy between monitored and measured results. Wood and wood-based materials can take longer time horizons to reach equilibrium, often weeks or months. This is relevant for the exterior MC monitoring positions, which are in direct contact with the WFI. This position saw the largest discrepancy between measured and simulated results. The exterior MC monitoring position was not subjected to large variations in climate, which led to minimal changes in MC.

4.2. Acceptance Criteria of Hygrothermal Simulation

Acceptance criteria for determining goodness of fit between measured and simulation data were assessed both qualitatively and quantitatively. Previous research has used acceptance criteria that varied greatly, with both qualitative and quantitative comparisons made, which are summarized in

Table 6. Summary of model and measurement error from literature [15,24,32,37,47,48,49,50,51,52,53,54,55,56,57,58,59].

Measurement	Error Type		All	Laboratory	Building
Temperature (°C)	MAE	Min.	0.2	0.2	0.24
		Max.	2.01	0.3	2.01
	RMSE	Min.	0.11	0.11	0.29
		Max.	9.27	1.53	9.27
	Other a	Min.	1.2	1.2	2
		Max.	5	5	3
Relative Humidity (%)	MAE	Min.	0.89	3.38	0.89
		Max.	16.5	8	16.5
	RMSE	Min.	0.97	0.97	1
		Max.	16.7	9.3	16.7
	Other a	Min.	5	5	5
		Max.	30	30	15
Moisture Content (%)	MAE	Min.	-	-	-
		Max.	-	-	-
	RMSE	Min.	0.1	1	0.1
		Max.	4.7	4.7	0.5
	Other	Min.	-	-	-
		Max.	-	-	-

^a Includes difference, percent difference, and percent error.

. None of the researchers explicitly stated their acceptance criteria; however, a comparison of the results was described by most studies. In instances where qualitative statements of good agreement were the only indicator of the ability of the model to predict conditions, no quantitative metrics were provided, exclusive of graphic evidence [43,44,45,46]. In the available literature, RMSE and/or MAE metrics were used the most often and were included in 15 of the papers [15,24,32,37,47,48,49,50,51,52,53,54,55]. The measured values corresponded with temperature, RH, or MC and depended on the scope of the paper. Wide ranges in acceptance criteria were found in RMSE from 0.11–9.27 °C [3,15,24,32,37,47,48,49,50,52,53], 0.97–16.7% RH [15,24,32,37,47,48,49,50,52,53], and 0.1–4.7% MC [32,51,53,54,55]. This demonstrates a wide range of criteria to establish what is considered acceptable. Further, it can be seen that when quantitative measurement was used to determine goodness of fit, there was a range, which often depended on the conditions used for measurement and modeling. Under steady state or laboratory conditions, differences between measured and modeled results were lower than those taken under real-world conditions. Laboratory or steady state conditions were seen in [32,47,48,50,51,54,55], where RMSE ranged from 0.11–1.53 °C, 0.97–8.68% RH, and 0.83–1.6% MC. Studies which were collected under real-world conditions were seen in papers [15,24,37,49,53,56], where RMSE ranged from 0.2–9.27 °C, 1–16.7% RH, and 0.1–0.5% MC. Limited variation was seen in MC in building settings due to the limited use in studies, where only one building-level study included the MC in its analysis [53]. Additionally, the use of differences (percent, maximum, and average) was used in [25,55,56,57,58,59]; however, the basis of measurement was not often discussed. Some papers used the maximum difference in percentage, and others used the average; however, no cohesive measurement was described. Thus, there were no formal acceptance criteria for this study; the MAE and RMSE were used to determine the goodness of fit, and sensitivity analysis was conducted to reduce any error associated with the model.

A review of literature demonstrated that the type of experiment, laboratory, or building, led to differences in the error within the model results. Laboratory experiments utilize well-defined and controlled boundary conditions, and RMSE below 1.53 °C, 9.3% RH, and 4.7% MC, whereas building scale experiments recorded RMSE below 2.01 °C, 16.7% RH, and 0.5% MC. In all cases, excluding MC, the error that was considered allowable was less in laboratory experiments when compared to building-scale experiments.

The parameters of the model were altered in an effort to evaluate the discrepancy in results and determine the ability of the model to predict conditions within WFI. To do so, P1 and P3 monitoring data were used as boundary conditions, with P2 simulated data being compared to monitored data. The model was conducted over the same time period; however, the CLT, air gap, and siding layers were not included. Additionally, wind-driven rain and solar radiation were not considered on the exterior surface. Results demonstrated good agreement with measured data, indicating errors were not associated with the material properties of WFI. The results of the model are included within the Supplementary Materials, as shown in Figures S10–S19. It should be noted that this assessment did not account for potential sources of error associated with material interaction within the wall envelope.

These models demonstrated better agreement between measured and monitored data. The RMSE at 7 of the 11 positions for temperature and 11 of 11 positions for RH was below the laboratory thresholds listed in the literature. Considering this, it was determined that the model was able to accurately predict the conditions within the assembly and represent the materials.

4.3. Effects of Simulation of RH Discrepancy on Mold Growth Assessment

Although the model accurately predicts the temperature and RH profile of the wall assembly, discrepancies between measured and modeled results were evident. These discrepancies, especially in RH, carry into the mold growth assessment, leading to an overprediction in the likelihood of mold growth occurring. In this case, overprediction of RH increases the likelihood of mold growth in the assessment, which would lead to a conservative prediction.

There are periods in the simulations in which the RH was both over- and under-predicted. The RH at P1 was generally underpredicted; however, the measured RH remained below 55% RH for the entire period. This low RH indicates that there was no potential for mold growth throughout the measurement period, and there is little concern with underprediction. The model overpredicted the RH at P3. This trend was beneficial for the mold growth assessment, as the WFI and WRB interface was the greatest likelihood for mold growth. This means that the risk associated with mold growth occurring at this position was overpredicted, leading to a conservative mold growth index prediction.

5. Conclusions

A hygrothermal assessment of building envelopes and materials can identify risks associated with use. A one-story building constructed of CLT and WFI was instrumented with temperature, RH, and MC sensors and monitored over a two-year period. A hygrothermal model was calibrated using measured material properties and monitoring data. Several conclusions can be drawn from this assessment:

• The hygrothermal model was able to sufficiently represent the hygroscopic materials;
• Under conditions within the study, mold growth assessment demonstrated no risk of degradation or attack;
• Vapor open wall assemblies are greatly influenced by the interior conditions of the building;
• Ventilation within an air cavity behind siding can mitigate risks of mold growth at the exterior surface of the insulation layer;
• WFI and CLT show no moisture or mold risk in the cold climate of Belfast, ME;
• A material property dataset of WFI was developed and is included in the Supplementary Materials to act as an aid for future design considerations and research efforts.

Overall, the simulated and monitored data showed that in a cold North American climate, natural materials demonstrate good hygrothermal performance. Future work will include supplementing monitoring equipment within the building to reduce gaps in measured interior conditions. Monitoring of the building will continue to assess performance and long-term durability. The wall assembly considered within this study is representative of a small section of building typologies in North American climates, and further research should be conducted to assess the performance of additional types of WFI, as well as their integration with other structural elements.