The Effect of Afforestation on Soil Moisture Content in Northeastern China

Widespread afforestation programs sequester carbon from the atmosphere and mitigate the rising of atmospheric carbon dioxide (CO2). Meanwhile, afforestation carbon sequestration may cost soil water. However, changes in soil moisture content (SMC) after large-scale afforestation or reforestation have rarely been quantified. In this study, we measured changes in SMC following afforestation using a paired plots method with data from 757 plots in Northeastern China. We found a marginally significant decline in soil moisture content of the top 1-m soil (SMC0-1m) after afforestation (P = 0.08) at the regional scale. The SMC responses to afforestation also vary across species. For example, significant SMC decrease are found for Populus spp. plantations (P < 0.05) and plantations of Pinus sylvestris var. mongolica (P < 0.05). Splitting the first meter of the soil profile into different depth intervals revealed that SMC declined significantly in shallow layers (0–30 cm) for Populus spp. and Pinus sylvestris var. mongolica. We also found that when SMC in the control exceeded a specific threshold, SMC for all five tree species considered tended to decrease, suggesting that the effects of afforestation on soil hydrology vary across different regions.


Introduction
Afforestation and reforestation programs provide important mechanisms to enhance carbon sequestration capacity of the terrestrial ecosystem [1][2][3] and mitigate the increase of greenhouse gas emission. Such programs have been widely implemented in many regions, including America [4], South America [5], and China [6]. China, currently the largest carbon emitter of the world, is also known for containing more than 25% of the world's planted forests [7]. Important afforestation projects in China include the Three-North Protective Forest Program, the River Protection Forest Program, and the Grain for Green Program [8]. Largely due to this afforestation effort, China's forest coverage increased from 16.0% in the 1980s to 20.4% in 2009 [9], and forest carbon sequestration was about 1.65 Pg C during the same period, with a mean rate of 0.075 Pg C yr -1 [3].
However, afforestation does not come without any environmental cost, which is especially evident in arid and semi-arid regions [10][11]. One of the most striking issues associated with afforestation lies in its influence on hydrological cycles and modification of water resource balance [12][13]. In general, afforestation increases canopy interception and transpiration loss of water, and hence reduces river and stream runoff and water yield [14][15]. A global analysis of 504 annual catchment observations showed that afforestation dramatically decreased stream flow within a few years of forest planting [16]. In China's Loess Plateau where afforestation programs like the Grain for Green Project have transformed 3.84 million hectares of farmland to forest plantations by the 2008 [17], which also bring out negative eco-hydrological effects like aggravated local water scarcity, even soil desiccation [10].
Soil moisture content (SMC) is one of the most important parameters determining the productivity and sustainability of terrestrial ecosystems [10,18,19]. Vegetation cover influences soil moisture through interception, transpiration and surface shading [20]. Because forest tends to have larger leaf area, surface roughness and deeper rooting system than grassland or cropland, vegetation restoration from cropland or grassland to forest could usually result in drastically reduced SMC restored [13], sometimes even to an extent of more than 35% [21], so that irrigation sometimes was performed to improve water availability [22]. Quantitative analyses of the influence on SMC by afforestation have been conducted in some scattered locations or regions [23][24][25], such as the Loess Plateau. However, these scattered results cannot be extrapolated to other forest plantations as such influence is highly heterogeneous in quantity across different regions [26].
Northeastern China is a key region for China's Three-North Protective Forest Program, designed to hold back the expansion of the Gobi Desert. In this study, using pairwise sampling of 757 afforested and control plots, we first quantify the effect of large-scale afforestation on SMC across Northeastern China. Because traits like canopy structure and stomatal properties could vary across different tree species, the capacity of plantation forests in modifying hydrological cycle is also different across different forest types. Thus a second objective of this study is to compare the change of SMC across five tree species widely used in afforestation in Northeastern China. Finally, afforestation-caused changes in SMC could also be heterogeneous between shallow and deep soil layers [21]. Using soil profile data, we also demonstrate the differences of SMC and its change along soil depths. Overall, the results will provide a comprehensive understanding on how different choices of afforestation tree species may influence soil moisture across horizontal and vertical space, which is critical information for assessing the ongoing afforestation programs and optimizing the design of future ones.

Study Area
The sampling sites are located in Northeastern China (34.20-51.80°N, 106.81-133.31°E) covering the provinces of Heilongjiang, Jilin, Liaoning, Hebei, Shanxi, Shaanxi and the Inner Mongolia Autonomous Region (Fig 1). Over the study area, mean annual temperature ranges from -3 to 15°C, and the mean annual precipitation ranges from 355 to 1068 mm yr -1 . Most of the sites are located in state-owned forest plantation farms and afforestation projects have been performed in a systematic way and well documented [27]. The administrative body of all involved forest farms gave permission for our research activities. The detailed information of forest farms are showed in S1 Table. Soil samples were obtained from plantation stands of five most commonly planted tree species (species groups): Pinus koraiensis, Larix gmelinii, Pinus sylvestris var. mongolica, Pinus tabulaeformis and Populus spp.

Measurements
Gravimetric soil moisture measurements were made during the summers of 2012 and 2013 at a total of 172 field sites spreading across the study region (Fig 1). Each site contains three tree plantation plots of different ages and one pre-afforestation control plot. The plantation plots are from stands with common local species planted; and the control plots represent initial soil conditions before afforestation. At each sampling site, environmental attributes such as land-use type, aspect, slope degree, elevation, tree species and tree height were recorded. To ensure the consistent soil origin and climate, the distance between any two plantation plots within the same site is always between 50 m and 5 km, and the distance between each paired plantation-control plots is less than 2.5 km. At each plot, three soil pits were dug with a spade and soil samples were taken from depth of 0-5 cm, 5-10 cm, 10-20 cm, 20-30 cm, 30-60 cm, 60-100 cm using a standard 100-cm 3 container. In total, 757 plots were sampled and 11,794 soil samples were collected. The soil samples were handled following standard procedures. Plant detritus were removed before weighing, and soils were sieved through a 2 mm mesh. All samples were weighed in the field, and then weighed again in the laboratory after drying at 105°C till constant weight.

Data analysis
We denote the 0-5 cm, 5-10 cm, 10-20 cm, 20-30 cm, 30-60 cm, 60-100 cm depth intervals as d 1 , d 2 . . .d 6 respectively, SMC (mass of soil water relative to dry-weight, %) of these depths as SMC 1 , SMC 2 ,. . .SMC 6 , and top 1-m SMC as SMC 0-1m . Soil moisture contents are calculated as: The thickness-weighted method is used to calculate the SMC 0-1m (%), which is given by the equation where SMC ij is SMC of the j th layer in the i th profile, w ij is the weighting factor of the j th layer in the i th profile. Both SMC a,0-1m and SMC c,0-1m are calculated with Eq (2). Here the denotations a and c represent afforested and control plots, respectively. ΔSMC 0-1m refers to the difference between SMC a,0-1m and SMC c,0-1m . ΔSMC j refers to the difference between SMC a,j and SMC c,j (j = 1, 2, 3 . . . 6).
The mean and standard deviation (std) of change in top 1-m soil moisture content (ΔSMC 0-1m ) in each site are listed in S2 Table. Statistical methods We performed Jarque-Bera test [29] and Brown-Forsythe's test [30] on the normality and homogeneity of variance, respectively. As it turns out the sample does not conform to normal distribution (P < 0.01), but met the variance homogeneity assumption (P > 0.05). So we use the non-parametric test in this study. A one-tailed paired Wilcoxon signed rank test was performed to detect whether SMC in afforested plots significantly decreased compared to SMC in control plots. The same tests were also performed on SMC of the six depth intervals. The relationship between the changes in top 1-m SMC after afforestation (ΔSMC 0-1m ) and SMC of top 1-m in control plots (SMC c,0-1m ) was fitted using cubic polynomial. A one-tailed Mann Whitney U test was used to compare tree height differences between plots that SMCc, 0-1m are above and below the threshold are significantly higher than zero. Linear mixed model was fitted to evaluate tree height effects on ΔSMC 0-1m within each of the five tree species. Tree species was treated as fixed effects, while tree heights was treated as random effects. All of the comparisons were examined at a significance level of 0.05.

Results
Afforestation-induced change in SMC 0-1m across the study area Soil moisture in control plots (SMC c,0-1m ) has a wide range from 2.5% to 43.3% (Fig  2a), mostly due to a spatial gradient in SMC with the highest values in the northeast and lower values in the southwest. SMC c,0-1m is mostly less than 20% in provinces of Hebei, Shanxi, Shaanxi, and Inner Mongolia, while it is relatively higher (>20%) in Heilongjiang, Jilin, and Liaoning (Fig 2b). On average, afforestation leads to marginally significant decrease in SMC 0-1m (P = 0.08), and the median of ΔSMC 0-1m is about -0.3%. ΔSMC 0-1m varies widely from -28.9% to 34.9% across different afforested plots. 52% of the afforested sites had a negative ΔSMC 0-1m (Fig 3a), mostly in Heilongjiang, Jilin, and Liaoning Provinces (Fig 3b) where SMC c,0-1m is relatively high (Fig 2b).
To further understand how pre-afforestation soil moisture content may affect afforestationcaused changes in SMC, we binned all the 586 afforested plots into 20 groups by their values of SMC c,0-1m (Fig 4). The results suggest that at wet places where SMC c,0-1m is larger than 25%, mean ΔSMC 0-1m is always negative. In relatively dry zones (SMC c,0-1m is between 2% and 24%), both positive and negative values of ΔSMC 0-1m are found, with positive values more dominant (Fig 4).  We then divided the afforested plots into two categories, i.e., plots below and above the species-specific threshold of SMC c,0-1m (Fig 5b). For the plots below the threshold of SMC c,0-1m , afforestation-induced increases in SMC 0-1m are statistically significant (P < 0.05, Fig 5c) for four of the five species, except for Pinus koraiensis. For plots with SMC c,0-1m above the threshold, all five tree species show extremely significant negative ΔSMC 0-1m (P < 0.01, Fig 5d).

Vertical distribution of afforestation-induced change in SMC among five tree species
We further explore afforestation-induced changes in soil moisture content (ΔSMC) along soil depths (Fig 6). Soil moisture content generally decreases with soil depths (Fig 6a), which is , the text P < 0.05 and P < 0.01 denote significantly and extremely significantly decrease in SMC 0-1m in afforestation plots compared to control plots. In panel (c), the text P < 0.05 and P < 0.01 denote significantly and extremely significantly increase in SMC 0-1m in afforestation plots compared to control plots. In box plots, the whiskers and boxes indicate the minimum, 25 th percentile, median, 75 th percentile and maximum. consistent with early findings [23]. However, ΔSMC does not show a consistent pattern along soil depth (Fig 6b). The largest magnitude of ΔSMC occurred in 10-60 cm soils of Pinus koraiensis, namely plantations, ranging from 1.6% to -2.6%. Non-significant ΔSMC are found in all soil layers of Larix gmelinii and Pinus tabuliformis plantations. Significant negative ΔSMC are also found in top 30-cm soils under plantations of Populus spp. and Pinus sylvestris var. mongolica.
As we showed before, afforestation reduces soil moisture content in wet regions but may increases soil moisture in dry regions. Therefore, for each soil depth, we also identified the species-specific thresholds of SMC c,0-1m (Fig 6c and 6d). As it turns out, when SMC c,0-1m is below the thresholds, ΔSMC at most depth intervals are significantly positive for Larix gmelinii, Pinus sylvestris var. mongolica, Populus spp. and Pinus tabuliformis. The largest SMC increase occurs in Larix gmelinii (Fig 6c). On the contrary, when SMC c,0-1m is above the thresholds, ΔSMC are negative over almost all soil depth intervals. The largest magnitude of ΔSMC is found in the 5-20 cm soils of Pinus tabuliformis plantations (-11.7% --2.0%, Fig 6d).

Impact of tree species on the SMC change after afforestation
Afforestation has been implemented in large areas of arid and semi-arid regions as an effective carbon sequestration and land management tool. There has been a growing knowledge that afforestation can alter hydrological cycle through modifying soil moisture [12] and changing stream flows [16,31,32]. For example, Yang et al. [13] evaluated the hydrological effect of vegetation restoration in Loess Plateau and found that soil moisture content decrease drastically after vegetation restoration. Compared to herbaceous vegetation, forest plantation was found to alter soil moisture even more by increasing rainfall interception and evapotranspiration [33], probably due to their larger leaf area [34] and deeper roots [35]. Consistent with previous studies, our results suggest that afforestation in Northeastern China marginally significantly reduces SMC (-0.3%, P = 0.08), although changes in SMC are not uniform across plantations with different tree species. For example, Populus spp. and Pinus sylvestris var. mongolica showed statistically significant SMC decline, while SMC change was not statistically significant under other tree species, including Pinus koraiensis, Larix gmelinii and Pinus tabuliformis. Such species-dependent impact on soil moisture has rarely been explored before (but sees Zhao [36] for plantations at Gobi).
The difference in canopy evapotranspiration across tree species may largely explain this inter-species difference in afforestation impacts on soil moisture. Many studies have found that Populus spp. plantations have a larger evapotranspiration capacity than most other tree species. For example, in a study over Horqin, among two studied tree species, the evapotranspiration of Populus spp. (558.1mm) is more than 30% larger than Pinus sylvestris var. mongolica plantations (405.0mm) during the growing season [37]. In Loess Plateau, the daily transpiration rate of Populus spp. (0.616 mmol/m 2 /s) is nearly twice more than Pinus tabuliformis (0.318 mmol/ m 2 /s) [38]. Therefore, it is not surprising that afforestation with Populus spp. reduces SMC more than that with other species. Few studies have investigated evapotranspiration capacity of dominant tree species for afforestation in Northeastern China. Such information is required for comprehensive assessment of the ecosystem impacts by those afforestation projects.
Plant structural traits are also associated with regulation of transpiration [39], like canopy characteristics and root systems. Since tree height is a continuous variable, we fit a linear mixed model using tree species, tree height and their interaction as predictive terms to test whether canopy heights can influence the SMC change. As it turns out, tree height impacts on ΔSMC was not statistically significant (P = 0.40). Therefore, tree height differences may not be the main factor driving ΔSMC variations found in our study. We also appreciate other hypothesis concerning other forest traits, such as root depth. 98% roots of Pinus sylvestris var. mongolica distributed in 1.0m depth [40], while Populus spp. roots can extend to 2m [41], which leading to larger root distribution discrepancy between plantation and control plots. Therefore, Populus spp. may maintain their transpiration even water deficit appeared in shallow soil depth. Besides, Populus spp. has larger leaf area than Pinus sylvestris var. mongolica [42]. Thus higher canopy transpiration of Populus spp. would result in more soil moisture loss. However, due to data limitation, we cannot explore them quantitatively in this study.
The variation in growth rate across species may be another possible reason for the inter-species difference in the afforestation-caused SMC decrease [43]. Fast-growing species demands more water for its carbon assimilation [44], and would lead to a higher overall water use per rotation period [31]. For example, Populus spp. trees grows faster than other tree species in this study and also require more water during their life cycles [45]. This high water demand by fastgrowing Populus spp. trees could create a larger difference in soil moisture content between pre-afforested and afforested plots than most other species. In fact, the different water consumption among tree species has already attracted attentions in afforestation species selection under different environment [25,46]. Given the high life-cycle water demand of Populus spp., currently extensive planting of Populus spp. in arid and semi-arid regions over Northeastern China may need to be reconsidered, and large-scale Populus spp. plantations would need to be assessed for their impacts on local water resources.

Dependence of afforestation impacts on SMC in control plots
Soil moisture content of pre-afforested control plots can serve as the baseline SMC level of a region. Our results suggest that in addition to inter-species difference in afforestation's impacts on soil moisture, baseline SMC also plays a critical role in determining the direction and amplitude of ΔSMC caused by afforestation. Significant soil moisture decrease (negative ΔSMC 0-1m ) is found when the baseline soil moisture (SMC c,0-1m ) is larger than 25% (Fig 4). In wet places with higher water availability, trees tend to grow faster than in drier areas [47], which in return consumes more soil water [48]. Indeed, when we compared tree heights for plots with soil moisture above and below the threshold of SMC c,0-1m , three of the five species show at least significant (P < 0.05) higher trees in plots with higher SMC c,0-1m (S2 Fig), supporting the hypothesis that higher soil moisture promotes tree growth and in turn leads to more depletion of soil moisture.
On the other hand, plants also develop physiological and morphological approaches in adaptation to drought. For instance, plants may respond to drought stress by closing their stomata [49], leading to reduced water consumption from the soil and slow-growing "small aged trees" in dry areas [50][51]. Also, it is worth to note that afforestation may even increase SMC in some arid areas when SMC c,0-1m is between 2% and 24% (Fig 4). This positive change in SMC by afforestation in dry areas may be explained by irrigation and/or precipitation feedbacks. For example, irrigation has been widely employed for most plantation programs located in arid and semi-arid areas [22,52], which can offset or even exceed soil moisture deficit caused by increased canopy evapotranspiration. Also in those water-limited areas, there is a precipitation feedback whereby increased evapotranspiration tends to favor more precipitation [53][54]. It is suggested that even in those dry regions, evapotranspiration of forested lands is still greater than that of non-forested lands [55]. The water vapor from forested lands may modify regional rainfall patterns and trigger more precipitation, and thus increase soil moisture content.
The hypotheses of irrigation and precipitation feedbacks proposed here are currently difficult to test because of the lack of data for the spatial extent of irrigation practice and the complexity in modeling regional precipitation feedback to afforestation. However, our results highlight that complicated influences of afforestation on local hydrological cycle. Afforestation programs often face the challenge of the competing demands between water resources and carbon sequestration, which requires careful selection of tree species and delicately balancing both objects. Current understanding on how to satisfy both water and carbon goals in afforestation programs is still limited. Compared with pair-wise catchment study (e.g. Huang et al [56]), our paired-plot sampling approach is subject to more sources of uncertainties. In particular, although the pair-sampling is always taken at the same day, the magnitude of the soil moisture difference between control-plots and afforestation-plots may be affected by seasonal variations of climate. However, this limitation did not affect our qualitative inference that afforestation has led to marginally significant change on regional hydrological cycle. Improving Earth system modeling and more soil moisture data with continuous monitoring should become the research priorities in future in order to better quantify the afforestation impacts. between plots that top 1m soil moisture content in control plots (SMC c,0-1m ) are above and below the threshold for five tree species. Error bars show confidence interval of tree height difference. The asterisks ( Ã ) denote that tree height differences between plots that SMC c,0-1m are above and below the threshold are significantly higher than zero, Ã P < 0.05, ÃÃ P < 0.01. (TIF) S1 Table. Sample sites information in our experiment. The longitude, latitude and name of forest farms are provided as below.