Effects of climate change on reduction of internal bond strength of particleboard subjected to various climatic conditions in Japan

Abstract

The effects of climate change on the reduction of internal bond strength (IB) of particleboard subjected to various climatic conditions in Japan were predicted. The temperature increased throughout Japan with increasing greenhouse gas (GHG) emissions, so the climate deterioration index (climate index for predicting the IB reduction of particleboard; a high index indicates a large IB reduction) also increased to decrease IB. The number of low-temperature areas decreased because of climate change. The rupture of bonding points in particleboard caused by outdoor exposure led to a IB reduction, and biodeterioration accelerated this reduction. In high-temperature areas, particleboard was found to be prone to biodeterioration, and the IB significantly reduced. An area wherein there was a significant decrease in the IB spread along the Pacific Ocean side of western Japan and the Sea of Japan in central Japan in relation to increasing GHG emissions. Particleboards are difficult to use outdoors because of climate change.

Introduction

Particleboard is typically used indoors only, but its use needs to be extended outdoors to create an increased demand for the product. Thus, the effect of outdoor exposure on particleboard must be investigated. Although outdoor exposure tests have been conducted on particleboard in North America [1–3], very few tests were conducted in Japan until the early 2000s [4]. Therefore, eight sites that are representative of the Japanese climate were selected, and particleboard was subjected to various climatic conditions from 2004 to 2011 by the Research Working Group on Wood-based Panels of the Japan Wood Research Society. The sites were Asahikawa, Noshiro, Morioka, Tsukuba, Maniwa, Okayama, Shizuoka, and Miyakonojo, which are numbered from 1 to 8, respectively, in Table 1. The results of this research group were recently reported in the following papers, e.g. [5–8].

Table 1 Site numbers, sites, north latitude, east longitude, and temperature grouping at outdoor exposure sites

With regard to the effects of climatic conditions on the strength reduction of particleboard, Kojima et al. [7] reported that temperature, sunshine duration, and precipitation are important factors. In our previous study [9, 10], these three climatic conditions were combined into the first principal component (PC1) in principal component analysis (PCA), and the first principal component score (PC1 score) was closely related to the strength reduction of particleboard. Results showed that the strength of particleboard subjected to outdoor exposure in high-PC1 score areas reduced significantly. Therefore, the PC1 score was determined to be an effective index for predicting strength reduction resulting from outdoor exposure, and it was named the “climate deterioration index (CDI)”.

In general, climate change causes serious problems, and many researchers have predicted effects on human life. In this study, the effects of climate change on the reduction of internal bond strength (IB) of particleboard subjected to various climatic conditions in Japan were predicted. Anthropogenic greenhouse gas (GHG) emissions are mainly driven by population size, economic activity, lifestyle, energy use, land use patterns, technology, and climate policy. The representative concentration pathways (RCPs), which are used to make projections in this study based on these factors, describe four 21st century pathways for GHG emissions and atmospheric concentrations, air pollutant emissions, and land use [11]. The RCPs involved are a stringent mitigation scenario (RCP2.6), two intermediate scenarios (RCP4.5 and RCP6.0), and one scenario with very high GHG emissions (RCP8.5) [11]. Scenarios without additional efforts to constrain emissions (’baseline scenarios’) lead to pathways ranging between RCP6.0 and RCP8.5. RCP2.6 is representative of a scenario that aims to keep global warming below 2 °C above pre-industrial temperatures [11]. Figure 1 shows temperature change in compliance with the three RCP scenarios (RCP2.6, RCP4.5, and RCP8.5) at the eight sites. Temperature change of even the RCP2.6 has widespread impacts on the eight sites. In this study, the CDI was predicted for 2031–2050 using three RCP scenarios (RECP2.6, RCP4.5, and RCP8.5). The effects of climate change on the IB reduction of particleboard subjected to various climatic conditions in Japan were predicted using the CDI based on the three RCP scenarios.

Fig. 1

Temperature change in compliance with three RCP scenarios, i.e., RCP2.6, RCP4.5, and RCP8.5, at eight sites. RCP representative concentration pathway. Details about the sites are provided in Table 1

Experimental

Outdoor exposure and IB tests

Particleboard was subjected to outdoor exposure at eight sites from February 2004 to March 2009. The sites are listed in Table 1, wherein the small numbers indicate northern sites. In general, temperature plays an important role to reduce IB, and the eight sites were classified as low-, middle-, and high-temperature sites (Table 1). The annual climatic conditions of mean temperature, solar radiation, and precipitation at these sites are listed in Table 2.

Table 2 Climatic conditions and IB at outdoor exposure sites from April 2004 to March 2009

In this study, commercially available phenol–formaldehyde resin-bonded particleboard was used, which has a thickness and density of 12.2 mm and 0.75 g/cm³, respectively. Thirty to forty particleboards measuring 910 mm × 1823 mm were cut into specimens measuring 300 mm × 300 mm. Specimens were selected at random and subjected to outdoor exposure vertically to the south at each site, as shown in Fig. 2. The IB test was conducted in compliance with JIS [13], and the IB data have reported previously [6, 9, 10, 14]. The mean and standard deviation of the initial IB were 0.833 and 0.09 MPa, respectively.

Fig. 2

Outdoor exposure test at site 4

PCA using climatic conditions at eight sites

PCA was conducted using computer software “R. 3.0.2” [15]. Table 2 lists the temperature, precipitation, and solar radiation at the eight sites recorded at exposure times of 1–5 years (2004–2009) [16]. These values at each exposure time at the eight sites were standardized and then analyzed using PCA. Eigenvalues, proportion, cumulative proportion, and eigenvectors resulting from PCA are listed in Table 3.

Table 3 Eigenvalue, proportion, cumulative proportion, and eigenvector calculated using the principal component analysis of climatic conditions

Prediction of CDI at eight sites using three RCP scenarios

To predict the future CDI at the eight sites in compliance with the three RCP scenarios, future temperature, precipitation, and solar radiation for 2031–2050 predicted with MIROC5 (Model for Interdisciplinary Research on Climate) were used [17]. MIROC5 predicted these climatic conditions of original spatial scales (about 125–444 km in latitude) in compliance with the three RCP scenarios. The original spatial scales were very large, so we interpolated these climatic conditions to the third mesh of the national standard grid system (about 1 km in latitude; 1 km × 1 km) using simple linear interpolation method [18]. The future CDI at the eight sites was then predicted using the developed PCA model that was described in “PCA using climatic conditions at eight sites” section in compliance with the three RCP scenarios.

Prediction of CDI nationwide using three RCP scenarios

To calculate the present CDI nationwide, the third mesh climatic conditions (temperature, precipitation, and solar radiation) from the Japan Meteorological Agency were used [19]. The third mesh of the national standard grid system (1 km × 1 km), which was described in “Prediction of CDI at eight sites using three RCP scenarios” section, was used to divide Japan’s area; the area is 377,900 km², so we obtained 378,030 grid squares. Based on the future climatic conditions predicted with MIROC5, future temperature, precipitation, and solar radiation of 378,030 grid squares were spatially interpolated in compliance with the three RCP scenarios. The future CDI of the 378,030 grid squares was predicted using these climatic conditions in compliance with the three RCP scenarios.

Multiple regression analysis

The model used in multiple regression analysis was developed using the present CDI at the eight sites and the exposure time as explanatory variables. The criterion variable was the IB. The IB at the exposure time of 5 years at the eight sites was predicted using the developed multiple regression equation in relation to climate change occurring in compliance with the three RCP scenarios. The IB of the 378,030 grid squares nationwide at the exposure time of 5 years was also predicted using the developed model. IB retention was calculated using the following equation:

$${\text{IB}}\,\,{\text{retention}}\;(\% ) = \frac{{{\text{Predicted IB}}}}{{{\text{Initial IB before outdoor exposure test}}}}~ \times 100$$

Mapping

The distributions of the predicted IB at the exposure time of 5 years were mapped in compliance with the three RCP scenarios of the 378,030 grid squares using QGIS (version 2.2.0) (Geographic Information System).

Results and discussion

Solar radiation used to calculate CDI at eight sites

Solar radiation is generally preferable to sunshine duration for investigating the effects of sunshine on the IB reduction of particleboard. However, solar radiation was not measured at sites 2, 5, 6, and 8; only sunshine duration was measured. In our previous studies [9, 10], sunshine duration was used to calculate the CDI. By contrast, in this study, solar radiation at sites 2, 5, 6, and 8 was predicted using the model that was developed to calculate the CDI [12].

Prior to using predicted solar radiation, the relationship between predicted solar radiation and measured solar radiation was investigated, and the accuracy of the model was confirmed. Solar radiation was measured at sites 1, 3, 4, and 7. Figure 3 shows the relationship between measured and predicted solar radiation at sites 1, 3, 4, and 7. The correlation coefficient was 0.999, showing a strong correlation between measured and predicted solar radiation. Thus, the model was able to predict solar radiation at sites 2, 5, 6, and 8, as listed in Table 2.

Fig. 3

Relationships between predicted and measured solar radiation at sites 1, 3, 4, and 7. Solar radiation at sites 2, 5, 6, and 8 was predicted using model that was developed by Kondo [12]. r correlation coefficient. Details about the sites are provided in Table 1

Table 3 lists the eigenvalue, proportion, cumulative proportion, and eigenvector calculated via PCA using temperature, precipitation, and solar radiation. For PC1, the eigenvalue was 2.34, and the proportion was 77.9%. These values indicated that 77.9% of the information relating to temperature, precipitation, and solar radiation was combined into PC1. Eigenvectors were used as coefficients to calculate the PC score. The PC1 score was calculated using the eigenvectors, and the equation is as follows:

$${\text{PC1 score }}\left( {{\text{CDI}}} \right) = 0.605 \times T + 0.531 \times P + 0.593 \times S$$

T standardized annual temperature, P standardized annual precipitation, and S standardized annual solar radiation.

All eigenvectors of PC1 had positive values for each climatic condition, thereby indicating that the PC1 score increased with increasing temperature, precipitation, and solar radiation. These eigenvectors were almost equal, indicating that the PC1 score was equally affected by temperature, precipitation, and solar radiation. By contrast, the eigenvalues and proportions of PC2 and PC3 were significantly lower than those of PC1, suggesting that PC1 accounted for the majority of the variability caused by climatic conditions. Consequently, PC2 and PC3 scores were not used for analytical purposes in this study. As described in the “Introduction” section, the PC1 score was named the CDI, combined temperature, precipitation, and solar radiation into one explanatory variable. Thus, the CDI (PC1 score) was calculated using temperature, precipitation, and solar radiation at each exposure time (as listed in Table 2).

Climate change at eight sites

Figures 1, 4, and 5 show temperature, precipitation, and solar radiation changes in compliance with the three RCP scenarios at the eight sites, respectively. Temperature increased with increasing GHG emissions at all sites. The increasing rate of temperature at low-temperature sites was higher than that at other sites. Precipitation increased at low-temperature sites and sites 5 and 6. By contrast, solar radiation did not remarkably change at all sites with increasing GHG emissions. Figure 6 shows CDI change in compliance with the three RCP scenarios at the eight sites. The CDI mainly increased at all sites with increasing GHG emissions because of the rise in temperature.

Fig. 4

Precipitation change in compliance with three RCP scenarios, i.e., RCP2.6, RCP4.5, and RCP8.5, at eight sites. RCP representative concentration pathway. Details about the sites are provided in Table 1

Fig. 5

Solar radiation change in compliance with three RCP scenarios, i.e., RCP2.6, RCP4.5, and RCP8.5, at eight sites. RCP representative concentration pathway. Details about the sites are provided in Table 1

Fig. 6

CDI change in compliance with three RCP scenarios, i.e., RCP2.6, RCP4.5, and RCP8.5, at eight sites. CDI climate deterioration index, RCP representative concentration pathway. Details about the sites are provided in Table 1

Distribution changes in temperature, precipitation, and CDI

Figure 7 shows the distribution changes in temperature, precipitation, and CDI of the 378,030 grid squares in compliance with the three RCP scenarios. As shown in Fig. 5, solar radiation did not considerably change at all sites, so the distribution change is not shown. The high peak of the present temperature distribution was approximately 10 °C, with a shoulder at the temperature range of 3–7 °C, as shown in Fig. 7a1. Compared with the present temperature distribution, high peaks in RCP2.6, RCP4.5, and RCP8.5 shifted from 10 to 15 °C, and the mean temperature increased from 10.2 to 15.5 °C. Furthermore, the shoulders of RCP2.6 and RCP4.5 shifted to the high-temperature range and that of RCP8.5 disappeared. A rise in temperature was observed throughout Japan, and the number of low-temperature areas decreased. As shown in Fig. 7b1–7b4, the high peak of the present precipitation distribution was approximately 1100 mm, but the peak shifted from 1100 to 1600 mm. No difference was noted among the distributions of RCP2.6, RCP4.5, and RCP8.5, and the mean precipitation of RCP2.6, RCP4.5, and RCP8.5 was almost the same. In Fig. 7c1, two peaks were found in the present CDI distribution at −3 and −1, and the peak at −3 was lower than that at −1. These peaks then shifted to a higher CDI in the RCP2.6 and RCP4.5 distributions with no difference between them. Furthermore, a shoulder and a peak, and not two peaks, were observed in the RCP8.5 distribution. High peaks were noted in the RCP2.6, RCP4.5, and RCP8.5 distributions, and these peaks shifted from −1 to 1 in comparison with the present CDI distribution. The mean CDI increased from −0.659 to 0.904 with increasing GHG emissions.

Fig. 7

Distribution changes in temperature, precipitation, and CDI of 378,030 grid squares in compliance with three RCP scenarios: RCP2.6, RCP4.5, and RCP8.5. CDI climate deterioration index. RCP representative concentration pathway, SD standard deviation

Multiple regression analysis of IB

Table 4 lists the standardized regression coefficient calculated using multiple regression analysis. The standardized regression coefficients of the exposure time and CDI were −0.661 and −0.718, respectively. Both values were negative and almost equal, indicating that the exposure time and CDI contributed to an equal IB reduction. Table 4 lists the coefficient of determination (R ²) and the adjusted coefficient of determination (adjusted R ²); both values were very high. Table 5 lists an analysis of variance table of the multiple regression analysis of the IB. The F value was very high, which was statistically significant at the 0.1% level. These results indicated that multiple regression analysis was able to accurately predict the IB.

Table 4 Standardized regression coefficient, coefficient of determination, and adjusted coefficient of determination of the multiple regression equation

Table 5 Analysis of variance table of multiple regression analysis of IB

Figure 8 shows the IB change of particleboard subjected to the exposure time of 5 years in relation to the three RCP scenarios at the eight sites using the developed multiple regression analysis model. The IB was found to decrease with increasing GHG emissions at all sites, and the reduction in high-temperature sites was higher than that in low- and middle-temperature sites. In particular, the IB at sites 7 and 8 significantly decreased. Climate change largely affected the IB reduction at high-temperature sites, and this reduction indicated other deterioration processes in particleboard because of outdoor exposure. The rupture of bonding points caused by outdoor exposure led to this reduction, and biodeterioration accelerated this reduction [20]. Particleboard was found to be prone to biodeterioration at high-temperature sites [21].

Fig. 8

IB change of particleboard subjected to 5-year outdoor exposure in compliance with three RCP scenarios, i.e., RCP2.6, RCP4.5, and RCP8.5, at eight sites. IB internal bond strength, RCP representative concentration pathway. Details about the sites are provided in Table 1

Although actual IB values were either positive or zero, the predicted IB showed negative values at sites 7 and 8 and was considered a null value. This finding was due to an assumption of linearity in multiple regression analysis, as discussed in the following section.

Map of IB reduction in Japan

Figure 9 shows a map of IB retention predicted using the multiple regression analysis model for the exposure time of 5 years of the 378,030 grid squares in compliance with the three RCP scenarios. At present, the IB reduction was large in the southern areas but small in the northern areas, and the reduction inland was smaller than that along the coastline. In particular, a large reduction was observed in the southern areas along the Pacific Ocean, which is a high-temperature area, and negative values were found for 8880 grid squares. IB retention in the 2.35% nationwide areas showed negative values. Negative IB retention in the high-temperature areas is related to the assumption of linearity in multiple regression analysis, which yields a null value. Therefore, in our previous study, we used an artificial neural network to solve this issue [6]. Low IB retention increased with increasing GHG emissions. For RCP8.5, the low IB retention areas significantly increased; the zone spread out along the Pacific Ocean side of western Japan and the Sea of Japan in central Japan. Negative IB retention was shown for 94,435 grid squares; IB retention in the 25% nationwide area showed negative values.

Fig. 9

Map of IB retention predicted using multiple regression analysis model for the exposure time of 5 years of the 378,030 grid squares in compliance with the three RCP scenarios, i.e., RCP2.6, RCP4.5, and RCP8.5, at eight sites. IB internal bond strength, RCP representative concentration pathway

Conclusions

Temperature throughout Japan increased with increasing GHG emissions because of climate change. Therefore, the CDI increased, and the number of low-temperature areas decreased. The rupture of bonding points caused by outdoor exposure led to IB reduction, and biodeterioration accelerated this reduction. In high-temperature areas, particleboard was significantly prone to biodeterioration. The low IB retention areas significantly increased with increasing GHG emissions; these areas spread out along the Pacific Ocean side of western Japan and the Sea of Japan in central Japan. Particleboards are difficult to use outdoors because of climate change. Negative IB retention in high-temperature areas is related to the assumption of linearity in multiple regression analysis, which yields a null value. This issue needs to be addressed using an artificial neural network or another method.