Results

The relationship between predictor and response variables was analyzed using multifactor ANOVA. Statistical blocks were added to the linear model to increase statistical power and to check for interaction effects. The emmeans package in R was used to determine sample statistics due to its advantages in case of unequal sample sizes, unbalanced designs or missing values. Multifactor ANOVA with blocking was chosen to analyze the difference in thinned and unthinned mean soil temperature and soil moisture. Data from each month was analyzed separately to avoid violating the assumption of independent sampling units.

Soil Temperature Analysis

Mean soil temperature difference between treatments was statistically significant (α=0.05) in June and July but not in May as indicated by large F values and small P values (table1). The effect of blocks on soil temperature was found to be highly significant in each month (table 1), this effect was not investigated further as blocks were only included in the model to better understand the effect of treatment. The interaction effects of block/treatment was non-significant, indicating that treatment affected soil temperature in the same direction (but not necessarily with the same magnitude) in all blocks investigated. Calculated means indicate that soil temperature was higher in thinned plots in June and July. (table 1, figure 1). Although interactions between months were not analyzed as part of the ANOVA models due to non-independence, no difference in effect direction was observed between months where a significant difference in treatment means was observed (figure1).

Means of thinned and unthinned monthly average soil temperature values calculated in R using the emmeans function were compared using the contrast function, with the unthinned mean taken as a suitable control and the thinned mean assumed to be larger than the thinned mean. The thinned population mean was determined to be larger than the unthinned population mean for June with 98.72% confidence and for July with 98.06% confidence.

Using a confidence level of α=0.05 and assuming that the thinned mean is larger than the unthinned mean, the thinned soil temperature population mean is at least 0.151°C higher for June and at least 0.102°C higher for July. Because all investigated means are relevant and reported here, no α adjustment for multiple comparisons was used.

Soil Moisture Analysis

Mean volumetric soil moisture difference between treatments was statistically significant (α=0.05) in May but not in June or July as indicated by a large F value and small P value (table1). The effect of blocks on soil temperature was found to be significant in each month (table 1), this effect was not investigated further. The interaction effects of block/treatment was non-significant, indicating that treatment affected soil moisture in the same direction (but not necessarily with the same magnitude) in all blocks investigated. Calculated means indicate that soil moisture was higher in thinned plots in all months. (table 1, figure 1).

Means of thinned and unthinned monthly average soil moisture values were calculated in R using the emmeans function and compared using the contrast function, with the unthinned mean taken as a suitable control and the thinned mean assumed to be larger than the thinned mean. The thinned population mean was determined to be larger than the unthinned population mean with 98.43% confidence.

Using a confidence level of α=0.05 and assuming that the thinned mean is larger than the unthinned mean, the thinned volumetric soil moisture population mean is at least 0.964% higher for May (This refers to 0.964% volumetric soil moisture, not 0.964% of either the thinned or unthinned means). No α level correction was used.

Discussion and Conclusion

These results support the hypothesis that thinning can be used as a tool to gain the benefits of increased soil temperature while avoiding the negative effects of reduced soil moisture. However, it is unclear if a difference of approximately 0.1°C and 1% soil moisture is large enough to affect tree growth or counter the expected effects of climate change. Reich et al. examined soil moisture effects across a gradient of approximately 5% to 25%.

Consistency of thinning effect direction across all blocks (lack of cross-over effect) suggests that similar results would be found across a broader range of sites than if there had been an cross-over effect.

The manipulated experimental design used here increases confidence that thinning caused increased soil moisture and temperature. However, it is still possible that thinning is associated with another, unknown, variable and is only correlated with increased soil moisture and temperature.

Perhaps the most interesting result was the lack of significant difference in both soil temperature or soil moisture in any one month (table 3). We know soil temperature decreases soil moisture soil. It's possible that the positive effect of thinning on soil moisture through reduced evapotranspiration was overwhelmed during June and July by differences in soil temperature. If thinning did not affect soil temperature in May, then this effect could not be overwhelmed and the positive effect on soil moisture could have shown through. In June and July it's possible that the negative effect that soil temperature had on soil moisture cancelled out the positive effect that reduced evapotranspiration had on soil moisture.

If the small increase in soil temperature observed here resulted in a complete cancellation of thinning's positive effect on soil moisture, then there are implications for the use of thinning as a tool to maintain tree growth in a changed climate. In months with higher soil temperature (when compared to unthinned forest) thinned forest did not show increased soil moisture. Therefore it is plausible to expect that thinned forest growing under a changed climate would not gain any benefit from increased soil moisture.

There is also no evidence to suggest that thinned forest growing in a changed climate would have lower soil moisture when compared to unthinned forest. Additionally, choosing a higher α value would have resulted in statistically significant soil temperature increases in the same month as statistically significant soil moisture increases.

The take home messages are therefore:

1. Thinning slightly increases soil temperature for part of the year, although it's unclear if the effect is large enough to affect tree growth.

2. Thinning slightly increases soil moisture for part of the year, although it's unclear if the effect is large enough to affect tree growth.

3. In those months where soil temperature was significantly increased, soil moisture was not found to have significantly increased.

And that:

4. No statistically significant evidence was found to suggest that thinned forest allows trees to take advantage of increased soil temperature while also benefitting from increased soil moisture, although non-statistically significant evidence was found.

Citations:

1.        Reich, P. B. et al. Effects of climate warming on photosynthesis in boreal tree species depend on soil moisture. Nature 562, 263–267 (2018).