3. Results and discussion

(TOC-Vcph, Shimadzu, Japan) [45]. SMBC was determined using the chloroform fumigation

Field harvest was conducted in mid or late August each year from 1982 to 2011, when the standing biomass reached its maximum. For each harvest in each year, 15 quadrats (1 1 m) were selected along a transect (300 100 m). Aboveground biomass was clipped and dried at 65C to constant weight. Between 1982 and 1992, the degraded grassland recovered rapidly and biomass production increased almost linearly. It was mainly caused by the exclusion of human disturbance, particularly overgrazing. After 1992, grasslands assumed a relatively balanced state with lower variation in productivity and diversity. Further variation in productivity was likely caused primarily by climatic variation. We therefore used the peak aboveground biomass during 1992–2011 to evaluate the impacts of climate variability on grassland productivity. Mean daily temperature and precipitation during 1992–2011 were obtained from

a weather station established in 1982, located only 0.9 km from the surveyed transect.

A one-way analysis of variance (ANOVA) followed by Tukey's HSD test was conducted to determine the effect of grazing exclusion time on grassland root traits (RB, RLD, RSA, plant functional group richness, plant species richness), the differences of root traits (SRL, SRS) and proportion in root community between plant functional groups, and the effects of clipping over time on soil respiration, microbial respiration, root respiration, soil temperature and soil moisture. Differences were considered significant for all statistical tests at P < 0.05. All the statistical analyses were conducted using IBM SPSS 18.0 (IBM, USA). Graphs were created

Partial least squares (PLS) regression was used to analyze the responses of grassland productivity to variation in daily temperature and precipitation during all 365 days of the year based on data for 1992–2011. The two major outputs of PLS analysis are the variable importance in the projection (VIP) and standardized model coefficients. The VIP values reflect the importance of all independent variables for explaining variation in dependent variables. The VIP threshold for considering variables as important is often set to 0.8. The standardized model coefficients indicate the strength and direction of the impacts of each variable in the PLS model. The root-mean-square errors (RMSEs) of the regression analyses were calculated to determine the accuracy of the PLS model. In the PLS analyses, periods with VIP greater than 0.8 and high absolute values of model coefficients represent the relevant phases influencing grassland productivity. Positive model coefficients indicate that increasing temperature or precipitation during the respective period should increase ANPP, while negative model coeffi-

extraction method [46].

2.3. Data analyses

2.3.2. Partial least squares

2.3.1. One-way analysis of variance

with Sigma plot 12.5 (Systat Software, USA).

cients imply negative impacts on productivity.

2.2.5. Biomass data collection

48 Plant Ecology - Traditional Approaches to Recent Trends

### 3.1. Effects of grazing exclusion on grassland root biomass and morphological traits

Results demonstrated that long-term grazing exclusion significantly increased grassland root biomass, root length density and root surface area (P < 0.05) (Table 1). The improved root biomass was mainly due to the increased aboveground productivity driven by the compensatory growth of dominant plant species after grazing removal [11, 18]. In the absence of herbivores, plants produced more roots to explore soil resource for aboveground growth, inducing increases in grassland total root length and surface area [47, 48]. Besides, our results indicated that the response of plant belowground richness to grazing exclusion followed a hump-like pattern, similar with responses of plant aboveground richness and diversity to grazing exclusion [11, 16], but with an earlier peak in the early-restoration stage (site GE05). Possibly long-term grazing exclusion caused a drastic decrease in bud bank size of forbs, followed with the decline or even disappearance of plant species relying on resprouting from bud bank after disturbance [49].

### 3.2. Root traits and proportional changes of five plant functional groups after grazing exclusion

Plant SRL and SRS showed significant differences between five plant functional groups (P < 0.05). Grasses had a much higher specific root length and specific root surface area than forbs. In detail, for SRL, PB had the highest value of 11.80 m g<sup>1</sup> , tripling that of SS (3.46 m g<sup>1</sup> ), while PR and AB had similar SRL value, higher than that of PF (Figure 1a). For SRS, there were no marked variations among PR, PF and AB, and they were significantly higher and lower than those of SS and PB, respectively (Figure 1b). Our results indicated that plant functional groups differed significantly in their proportions (P < 0.05) (Figure 2a–d). As the predominant plant functional groups, PB and PF accounted for more than 50% in total. Based on root biomass, proportions of PR and PB significantly increased with a significant decrease in PF proportion after long-term grazing exclusion (P < 0.05), and SS and AB showed little change (P > 0.05) (Figure 2a). Based on root length density and root surface area, grazing exclusion significantly increased PR proportion and decreased PF proportion (P < 0.05), while PB and SS show little responses to grazing exclusion (P > 0.05) (Figure 2b, c). Interestingly, with the prolonged grazing exclusion years,


Different lowercase letters indicate significant differences (P < 0.05) between five study sites.

Table 1. Root biomass, root length density (RLD), root surface areas (RSA), plant functional group (PFG) richness and plant species richness in study sites.

Figure 1. SRL traits (a) and SRS traits (b) of five plant functional groups. Different lowercase letters indicate significant differences (P < 0.05) between plant functional groups.

proportions of PR and PB in plant species richness significantly increased (P < 0.05), and those of PF and AB significantly decreased (P < 0.05), while SS showed little fluctuation (P > 0.05) (Figure 2d).

As the guerrilla plant species, PR had advantages in spatial propagation and exploration of adjacent nutrient patches by increasing rhizome and root length after grazing exclusion [50]. Additionally, dispersal by rhizomes allowed temporal release of PR plants from their natural enemies (i.e., root herbivores and pathogens), which stimulated plant growth in return [51]. The compositional changes of plant functional groups mainly resulted from their different responses to improved soil resources after grazing exclusion [52]. Compared with forbs, grasses had a stronger correlation with soil N [16], and grasses' higher SRL and SRS consolidated their superiority in acquiring soil resources [20]. Given that nitrogen deposition often occurs with accompanying rainfall events, which forms water and nutrient pulses [53], plants with larger root systems (i.e., grasses) gained more benefit than smaller plants at the start of the nutrient pulse [54]. Therefore, our study indicated that the hierarchy of root system size and root traits among five plant functional groups determined grassland root pattern in semiarid grassland after long-term grazing exclusion.

### 3.3. Effect of clipping on soil respiration

Clipping significantly reduced the mean soil respiration by 14.7% (P < 0.001) and 11.4% (P < 0.05) in 2014 and 2015, respectively (Table 2, Figure 3a). Previous research has reported that clipping could decrease the soil respiration in grassland ecosystems, which was most likely due to the restriction of translocation of photosynthate from aboveground plant tissues to roots and rhizosphere microorganisms [31]. In addition, clipping increased soil temperature by 0.6C (P > 0.05)

proportions of PR and PB in plant species richness significantly increased (P < 0.05), and those of PF and AB significantly decreased (P < 0.05), while SS showed little fluctuation (P > 0.05)

Figure 1. SRL traits (a) and SRS traits (b) of five plant functional groups. Different lowercase letters indicate significant

As the guerrilla plant species, PR had advantages in spatial propagation and exploration of adjacent nutrient patches by increasing rhizome and root length after grazing exclusion [50]. Additionally, dispersal by rhizomes allowed temporal release of PR plants from their natural enemies (i.e., root herbivores and pathogens), which stimulated plant growth in return [51]. The compositional changes of plant functional groups mainly resulted from their different responses to improved soil resources after grazing exclusion [52]. Compared with forbs, grasses had a stronger correlation with soil N [16], and grasses' higher SRL and SRS consolidated their superiority in acquiring soil resources [20]. Given that nitrogen deposition often occurs with accompanying rainfall events, which forms water and nutrient pulses [53], plants with larger root systems (i.e., grasses) gained more benefit than smaller plants at the start of the nutrient pulse [54]. Therefore, our study indicated that the hierarchy of root system size and root traits among five plant functional groups determined grassland root pattern in

Clipping significantly reduced the mean soil respiration by 14.7% (P < 0.001) and 11.4% (P < 0.05) in 2014 and 2015, respectively (Table 2, Figure 3a). Previous research has reported that clipping could decrease the soil respiration in grassland ecosystems, which was most likely due to the restriction of translocation of photosynthate from aboveground plant tissues to roots and rhizosphere microorganisms [31]. In addition, clipping increased soil temperature by 0.6C (P > 0.05)

semiarid grassland after long-term grazing exclusion.

3.3. Effect of clipping on soil respiration

differences (P < 0.05) between plant functional groups.

50 Plant Ecology - Traditional Approaches to Recent Trends

(Figure 2d).

Figure 2. Distribution proportions of five plant functional groups in root biomass (a), root length density (b), root surface area (c) and plant species richness (d) in grazing grassland (GG), grassland with grazing exclusion for 5 years (GE05), 9 years (GE09), 22 years (GE22) and 30 years (GE30), respectively. Different lowercase letters indicate significant differences (P < 0.05) between five plant functional groups for each grassland type, and n.s. indicates no significant difference (P > 0.05) for each plant functional group between five grassland types; \* and \*\* indicate significant differences for each plant functional groups between five grasslands in P < 0.05 level and P < 0.01 level, respectively.


Table 2. P-values of repeated measures ANOVA of total soil respiration (SR), microbial respiration (MR), root respiration (RR), soil temperature (ST) and soil moisture (SM) in a temperate grassland of Loess Plateau.

Figure 3. Seasonal variations of soil respiration and its components in the control and clipping treatments. Values are means, standard deviations (n = 5). Asterisks denote significant difference (P < 0.05) between treatments. Arrows indicate clipping dates.

in 2014 and 1.3C (P < 0.05) in 2015 in our study (Figure 4). We speculated that there was a potential increase in soil respiration driven by soil temperature, because higher soil temperature has been reported to stimulate the activities of plant roots and soil microbes [29]. However, the increase of soil respiration due to elevated soil temperature may not compensate for decrease in soil respiration caused by reduced photosynthesis, leading to the decrease in soil respiration after clipping.

### 3.4. Effect of clipping on root respiration

In the present study, clipping reduced the mean root respiration by 22.1% (P < 0.001) and 13.3% (P > 0.05) in 2014 and 2015, respectively (Table 2, Figure 3b). We found a prompt response in root respiration in the first measurements after two days of clipping treatment, following the sharp reduction of 49.2 and 26.4% within two weeks after treatment in 2014 and 2015, respectively (Figure 3b). We also found that the sharp decrease in root respiration was consistent with the sudden reduction of root production in the same periods (Figures 3b and 5a). Considering the significant correlation between the root production and root respiration (Figure 5a), we attributed the decrease of root respiration after clipping to the limited supply

Modification in Grassland Ecology under the Influence of Changing Climatic and Land Use Conditions http://dx.doi.org/10.5772/intechopen.69478 53

Figure 4. Seasonal variations of soil temperature in the control and clipping treatments. Values are means, standard deviations (n = 5).

of substrate for root growth and production. However, in September–October in 2014 and April–May in 2015, a higher root respiration was observed in the clipping plots (Figure 3b). Previous studies by Wan et al. [55] and Zhou et al. [33] reported that clipping could stimulate root respiration by promoting plant regrowth and root biomass. In our study, the higher root production observed in clipping plots in September–October in 2014 and April–May in 2015 might be responsible for the higher root respiration in the same periods (Figures 3c and 5a).

### 3.5. Effect of clipping on microbial respiration

in 2014 and 1.3C (P < 0.05) in 2015 in our study (Figure 4). We speculated that there was a potential increase in soil respiration driven by soil temperature, because higher soil temperature has been reported to stimulate the activities of plant roots and soil microbes [29]. However, the increase of soil respiration due to elevated soil temperature may not compensate for decrease in soil respiration caused by reduced photosynthesis, leading to the decrease in soil respiration after

Figure 3. Seasonal variations of soil respiration and its components in the control and clipping treatments. Values are means, standard deviations (n = 5). Asterisks denote significant difference (P < 0.05) between treatments. Arrows indicate

In the present study, clipping reduced the mean root respiration by 22.1% (P < 0.001) and 13.3% (P > 0.05) in 2014 and 2015, respectively (Table 2, Figure 3b). We found a prompt response in root respiration in the first measurements after two days of clipping treatment, following the sharp reduction of 49.2 and 26.4% within two weeks after treatment in 2014 and 2015, respectively (Figure 3b). We also found that the sharp decrease in root respiration was consistent with the sudden reduction of root production in the same periods (Figures 3b and 5a). Considering the significant correlation between the root production and root respiration (Figure 5a), we attributed the decrease of root respiration after clipping to the limited supply

clipping.

clipping dates.

3.4. Effect of clipping on root respiration

52 Plant Ecology - Traditional Approaches to Recent Trends

Microbial respiration exhibited relatively constant lower values in clipping plots almost throughout the study period in our study. Clipping significantly reduced microbial respiration by 6.0% (P < 0.05) and 9.9% (P < 0.05) in 2014 and 2015, respectively (Table 2, Figure 3c). The main explanation of this result was the reduced supply of labile C for mineralization by soil microorganism after clipping [56]. In the present study, clipping reduced the WSOC by 20.6% (P > 0.05) and 27.1% (P < 0.05) in 2014 and 2015, respectively (Figure 5b). The decrease of WSOC might be responsible for the reduction of SMBC in clipping plots in our study (Figure 5c), because WSOC was one of the main labile C substrates for soil microorganism. In addition, SMBC was reported to be significantly related to microbial respiration in previous research [57], which was similar to our results (R2 =0.88, P < 0.05). Hence, we attributed the decrease of microbial respiration after clipping to the reduction of available C supply for microbial mineralization.

### 3.6. Response of grassland productivity to variation in daily temperature

Between 1992 and 2011, the average harvest date of peak aboveground biomass for grassland at Yunwushan National Nature Reserve was 15th of August. The 365 daily temperature values between the previous September and August of the year of harvest were used as independent variables in the PLS regression. A low root-mean-square error (RMSE) of 8.13 g m<sup>2</sup> for the resulting PLS model indicated that the model was a good fit for the data. Based on the VIP and standardized model coefficients of the PLS analysis, we found that warming during different periods had varied impacts on grassland productivity (Figure 6).

Figure 5. Comparison of root length production (a), water-soluble organic carbon in trenched plots (b) and soil microbial biomass C in trenched plots (c) among treatments. Asterisks and different letters denote significant difference (P < 0.05) between treatments. Arrows indicate clipping dates.

Between 30 March and 30 May, model coefficients for temperature analysis (Figure 6) were always positive and VIP values mostly exceeded 0.8 (the threshold for variable importance), indicating that warming in April and May increases grassland productivity. During 31 May–1 August, model coefficients were consistently negative and VIP values were mostly important, implying that temperature increase in summer (June–July) depressed productivity, forming a striking contrast with the impacts of spring warming. It was of interest that the relevant periods influencing productivity, as identified by PLS regression, were almost the same as the phases of plant growth (i.e., the early and middle stages of the growing season) at our study area. No obvious impacts of temperature variation in August on grassland productivity were apparent. During September–October (the senescence period for vegetation), most model coefficients were negative, indicating that high temperature at that time was unfavorable for productivity of the following year. During 1 November–29 March, the dormancy period, model coefficients were mostly negative, although this phase also included some short intervals with positive coefficients. This variation might indicate that dormancy for grassland is a complex physiological and ecological process. Moreover, it seems possible that the strength of temperature impacts varies throughout the dormancy period. Taking a broader view at model coefficients and aiming at consistency with established phonological phases, we interpreted the entire period (November–March) as another relevant period during which temperature increases appeared to reduce grassland productivity.

### 3.7. Response of grassland productivity to variation in daily precipitation

The 365 daily precipitation values between the previous September and August were also used as independent variables in the PLS analysis. The resulting model still proved to be a good fit for the data, with an RMSE of 6.53 g m<sup>2</sup> . In contrast to the positive effects of higher precipitation in June and July, increasing rainfall during the senescence period (September–October) and the early growing season (April–May) was correlated with low productivity (Figure 6). Similar to temperature effects in August, no significant relationship was found between grassland ANPP and precipitation in August. During the dormancy period, there was no consistent correlation between precipitation and productivity. Positive impacts were almost offset by negative ones.

Modification in Grassland Ecology under the Influence of Changing Climatic and Land Use Conditions http://dx.doi.org/10.5772/intechopen.69478 55

Between 30 March and 30 May, model coefficients for temperature analysis (Figure 6) were always positive and VIP values mostly exceeded 0.8 (the threshold for variable importance), indicating that warming in April and May increases grassland productivity. During 31 May–1 August, model coefficients were consistently negative and VIP values were mostly important, implying that temperature increase in summer (June–July) depressed productivity, forming a striking contrast with the impacts of spring warming. It was of interest that the relevant periods influencing productivity, as identified by PLS regression, were almost the same as the phases of plant growth (i.e., the early and middle stages of the growing season) at our study area. No obvious impacts of temperature variation in August on grassland productivity were apparent. During September–October (the senescence period for vegetation), most model coefficients were negative, indicating that high temperature at that time was unfavorable for productivity of the following year. During 1 November–29 March, the dormancy period, model coefficients were mostly negative, although this phase also included some short intervals with positive coefficients. This variation might indicate that dormancy for grassland is a complex physiological and ecological process. Moreover, it seems possible that the strength of temperature impacts varies throughout the dormancy period. Taking a broader view at model coefficients and aiming at consistency with established phonological phases, we interpreted the entire period (November–March) as another relevant period during which temperature

Figure 5. Comparison of root length production (a), water-soluble organic carbon in trenched plots (b) and soil microbial biomass C in trenched plots (c) among treatments. Asterisks and different letters denote significant difference (P < 0.05)

increases appeared to reduce grassland productivity.

for the data, with an RMSE of 6.53 g m<sup>2</sup>

between treatments. Arrows indicate clipping dates.

54 Plant Ecology - Traditional Approaches to Recent Trends

negative ones.

3.7. Response of grassland productivity to variation in daily precipitation

The 365 daily precipitation values between the previous September and August were also used as independent variables in the PLS analysis. The resulting model still proved to be a good fit

tation in June and July, increasing rainfall during the senescence period (September–October) and the early growing season (April–May) was correlated with low productivity (Figure 6). Similar to temperature effects in August, no significant relationship was found between grassland ANPP and precipitation in August. During the dormancy period, there was no consistent correlation between precipitation and productivity. Positive impacts were almost offset by

. In contrast to the positive effects of higher precipi-

Figure 6. Results of partial least squares (PLS) regression correlating grassland productivity at Yunwushan during 1992– 2011 with 15-day running means of (a) daily mean temperature and (b) daily precipitation from previous September to August. Blue bars in the top row indicate that VIP values are greater than 0.8, the threshold for variable importance. In the middle row, red color means model coefficients are negative and important, while green color indicates important positive relationships between grassland productivity and climate variables. The black lines in the bottom panel stand for daily mean temperature and precipitation, while gray, green and red areas represent the standard deviation of daily climate variables.

The increased temperature with reduced precipitation in spring (April–May) could improve grassland productivity. Biomass produced in spring is often believed to be limited by cold temperatures at mid or high latitude [58]. Temperature increases early in the growing season may stimulate plant growth directly by raising leaf temperatures or indirectly by increasing water absorption and N mineralization (Figure 7) [40]. Additionally, warmer springs also likely accelerate snowmelt and advance spring greening [59], which might lengthen the growing season and result in increased photosynthesis and carbon acquisition [60]. In contrast to some studies reporting that more precipitation during April–May promoted grassland productivity [39], we found a negative relationship between these variables. To some extent, this discrepancy can be explained by the site hydrology. Frequent winter snow (lasting from November to March) in our study area provides sufficient soil water for plant growth in early spring. The sporadic precipitation during April–May (with an average of 59.5 mm during these two months between 1992 and 2011) may not have important direct impacts on productivity. In contrast, low air and soil temperature, as well as limited solar radiation caused by frequent rain events in May, might partially explain the negative correlations between spring rainfall and grassland productivity.

Warming in summer coinciding with drought can generate physiological stress for plant growth (Figure 7) [61], which can explain the reduced productivity in our study area. Moreover, increases in summer temperature can also lower ANPP, perhaps by reducing soil moisture through increased evapotranspiration. Decrease in precipitation amounts and lengthening of intervals between precipitations events during the past 20 years further reduced soil water

Figure 7. Potential relationships between grassland productivity and climate variability during (a) April–May, (b) June– July and (c) November–March at Yunwushan.

availability in our study region. This is in line with the hypothesis that impacts of climate variation and change on plant productivity might occur via variability in soil moisture [36]. Continuous warming and drought in summer could also affect N mineralization negatively and limit soil resource availability, thereby reducing productivity.

PLS regression did not detect a response of grassland productivity to climatic variation in August. Compared to climate variation during June–July, August shows more variable temperature and precipitation in our study region, although August is cooler on average than July. For instance, the coefficient of variation (CV) of precipitation in August between 1992 and 2011 was 53.3%, while it was only 33.5% for June–July. It is also worth noting, however, that in our study biomass was mostly harvested around the 15th of August, so that the vegetation was only exposed to half a month of August conditions.

Increases in temperature and precipitation during September–October in the previous year were negatively correlated with productivity in the current year, which can be partially explained by the widely reported delays of senescence caused by warming and wetting later in the year [62]. Delay in the senescence period may be related to some extent to increased soil nutrient and water depletion. This would imply that fewer resources may have been available for biomass production in the following year.

While some studies reported that weather during the dormancy period had limited impacts on grassland productivity [38], such effects may become more important, as temperature in winter further increases. Our results indicated that high temperatures during the dormancy period were negatively correlated with productivity. This is consistent with warming experiments in two limestone grasslands in the UK, which showed that winter heating combined with drought reduced the biomass of both communities [63]. Warmer winter can lead to some unanticipated consequences (Figure 7). The most direct impacts have been a shortening of the snow season and a reduction in snow cover, which have been observed in our study area. Declines in the area and depth of snow cover may expose the land surface to more frequent freezing events, exerting negative effects on plant growth. This is supported by observations in northern Scandinavia where extensive areas of vegetation died due to loss of snow cover after extreme winter warming in December 2007 [64]. Increased demands of soil nutrients and water due to accelerated root and microorganism metabolism caused by winter warming might also contribute to the productivity reduction. Finally, variation in spring phenology can also help explain this phenomenon. The timing of spring phenology in most temperate plants results from the interplay of winter cold and spring heat. Plants that evolved in temperate climates fall dormant in autumn to protect themselves from winter freezes and will only resume growth in spring when they have been exposed sufficiently to cold conditions [65]. Temperature increases in spring can advance spring phenology (e.g., greening for grassland), but warming in winter may delay the fulfillment of chilling requirements and thus lead to a slowdown in the advance of spring events or even later onset of spring phenology [65, 66]. The advancing trend in spring greening still dominates climate change responses of plants in our study region so far, since chilling requirements for vegetation are easily satisfied in all winters under the present cold climate with a mean temperature of 2.6C for the dormancy period. As global warming progresses, especially when rates and effects of warming in winter exceed those in spring, advances in greening might be slowed or even turn into delays. We therefore recommend increased scientific attention to impacts of winter warming on grassland productivity and the timing of spring phenology events.
