The experimental field site (86∘12′ E, 41∘36′ N; 886 m above sea level) is located in one of the oases scattered on the alluvial plain of the Kaidu–Kongqi River (a tributary of the Tarim River) basin, north of the Taklamakan Desert in the Xinjiang Uygur Autonomous Region of northwest China. This region has a temperate continental climate, with a mean annual precipitation of 60 mm, mean annual temperature of 11.48 ∘C, and mean annual water surface evaporation of 2788 mm, as measured using a 20 cm diameter pan. The annual sunshine duration is 3036 h, which is favourable for cotton (Gossypium hirsutum L.) growth. The experimental field covers an area of 3.48 ha. The major soil texture in the field is silt loam, the contents of the sand, silt and clay separately are 32.8, 62.4, and 4.8 %, respectively, and the soil bulk density ranges from 1.4 to 1.64 g cm-3 in the 1.5 m soil profile. The soil porosity is 0.42, which was directly determined in the laboratory using undisturbed soil columns collected in the experimental field.

Cotton is usually sown in April and harvested during October and November; i.e. the growing season is from approximately DOY (day of the year) 100 to 300. The planting style is “one film, one drip pipe beneath the film, and four rows of cotton above the film”, as depicted in Fig. 1. The plastic film (0.008 mm thick) was white and made of dense and airtight transparent polyethylene film. The width of the film was 1.1 m, and the inter-film zone was 0.4 m. Before sowing, small square holes (2 cm length and width) were cut in the plastic film in rows at 0.1 m intervals for germination, seeds were placed in the holes, and each hole was covered with soil. The planting density was approximately 160 000 plants ha-1. The basic fertilizer that was applied annually before sowing included 173 kg ha-1 of compound fertilizers (14 % N, 16 % P2O5, and 15 % K2O), 518 kg ha-1 of calcium superphosphate (18 % N and 40 % P2O5), and 288 kg ha-1 of diammonium phosphate (P2O5 > 16 %). Supplemental fertilizers applied during the growing season included approximately 292 kg ha-1 of urea (46 % N), 586 kg ha-1 of drip compound fertilizer (13 % N, 18 % P2O5, and 16 % K2O), and foliar fertilizer (P2O5 > 52 % and K2O > 34 %). Drip irrigation usually began on 12 June in the bud stage, with an approximate amount of 20–50 mm during each application and 9–12 applications per growing season. The annual irrigation amount was 500–600 mm.

This study focuses on the growing season, as soil respiration in the non-growing season is extremely low. The mulched and non-mulched treatments were arranged in a randomized block design with three replicates in the same field with the same fertilization and irrigation scheme from the year 2014 to 2016. The plastic mulch had been covered until the seed germinated in the non-mulched treatment to protect the germinating seeds. The experiments began approximately before the bud stage, when the cotton began to grow faster. The dates of the beginning of the experiments were DOY 184, 175, and 167, and the lengths of the measured periods were 95, 60, and 100 days, respectively. Soil respiration measurements were carried out using an LI-8100A (LI-COR, Inc., Lincoln, Nebraska) on a day between two irrigation events. Therefore, soil respiration was measured approximately every week during the cotton-growing season. The automated soil CO2 flux measurement system consisted of two parts, PVC collars (10 cm in diameter and 5 cm in height) and a measuring chamber. The PVC collars were inserted 2–3 cm into the soil by removing the living plants and litter within the collars at least 1 day before the measurements. Data were recorded using the data logger in the LI-8100A system.

The soil respiration was measured in the following areas: the furrow and ridge in the non-mulched treatment and the furrow, planting hole, and plastic mulch in the mulched treatment in 2016 (see Fig. 1 for the experimental configuration). Soil respiration was measured in the furrow in the mulched treatment and in the ridge in the non-mulched treatment in 2014, and it was measured in the furrow in the mulched treatment and in both the furrow and ridge in the non-mulched treatment in 2015. The measurements were performed every 2 h during the experimental day from 08:00 to 24:00 UTC+8. To measure the soil respiration at the soil surface without film covering (i.e. the furrows in the mulched and non-mulched fields and the non-mulched ridges), the PVC collars were inserted directly into the soil. Before measuring the CO2 emissions through the planting holes, the PVC collars were inserted into the soil covering two planting holes, and Scotch tape was used to seal the interspaces between the plastic mulch and collar to prevent air leakage. To measure the CO2 emissions through the plastic mulch, PVC collars were buried into the soil under the mulch, with Scotch Tape sealing the interspaces. Detailed measurement methods are further described in Berger et al. (2013). The soil temperature and soil moisture adjacent to each PVC collar at a depth of 5 cm were monitored using the auxiliary sensors of the LI-8100A concurrently with the soil CO2 flux measurements. The amount of drip irrigation was determined using water meters installed on the branch pipes of the drip irrigation system. The precipitation was measured using a tipping bucket rain gauge (model TE525MM, Campbell Scientific Inc., Logan, UT, USA), which was mounted 0.7 m above the ground.

The soil respiration from different areas at a particular time of a day was calculated as the average of three replicates. The daily mean Rs was calculated as the average Rs measured at various times in a day. The Rs in the mulched ridges was calculated based on the area ratio of Rs measured through the planting holes and the plastic mulch: Rr-m=Rh-m⋅Ah-m+Rp-m⋅Ap-m, where Rh-m and Rp-m represent the soil respiration from the planting hole and plastic mulch, respectively, and constitute the soil respiration in the ridge (Rr-m). The term A represents the area ratio of the different parts, and Ah-m and Ap-m are 0.3 and 0.7, respectively, in our field.

The seasonal accumulative Rs in the ridges and furrows was calculated by summing the Rs values over the measurement period (Yu et al., 2016; Berger et al., 2013). The soil respiration in plastic mulched and non-mulched fields was calculated based on the area ratio of Rs through ridges and furrows: Rm=Rr-m⋅Ar-m+Rf-m⋅Af-mRnm=Rr-nm⋅Ar-m+Rf-nm⋅Af-m, where Rm and Rnm represent soil respiration in mulched and non-mulched fields, respectively, and Ar-m and Af-m are the area ratios of the ridge and furrow, respectively, which are the same for mulched and non-mulched fields and are 0.73 and 0.27, respectively, in our field.

We used t tests to test the significance of differences between the Rs values from the furrows and ridges of mulched and non-mulched fields.

The relationships between Rs and soil temperature and moisture were analysed using regression analysis in SPSS (Statistical Package for the Social Sciences) software. The van't Hoff equation was used to represent the relationship of Rs with soil temperature (van't Hoff, 1898): Rs=AebT, where Rs represents the soil respiration, T represents the soil temperature, and A is the intercept of soil respiration when the soil temperature is 0 ∘C (i.e. reference soil respiration). Moreover, b represents the temperature sensitivity of soil respiration. The Q10 value, which describes the change in soil respiration over a 10∘ increase in soil temperature, is calculated as Q10=e10b. Considering that low and high values of soil water content both limit soil respiration, we adopted a quadratic equation to simulate the effect of soil moisture on soil respiration according to Davidson et al. (1998): Rs=aSWC2+bSWC+c, where SWC is the soil water content, and a, b, and c are regressed parameters.

Figure 2 shows the dynamics of albedo, soil moisture, soil temperature, and cotton leaf area index (LAI), which suggest that these environmental factors and crop growth conditions are modified by PFM and other cultivation practices. Other than two snowfall events that occurred in January 2015 and January 2016 and elevated the albedo beyond 0.4, the albedo was altered by cultivation, as shown in Fig. 2b. In early March, it was increased by the spring irrigation conducted 1 month before sowing. Then, it was decreased by ploughing several days before mulching on approximately 20 April. After plastic mulching in April, the surface albedo showed a sudden rise and then slowly decreased with crop canopy development. In general, the albedo reached its minimum value along with the highest value of LAI during the bud stage in August and then increased very slowly with leaf fall.

The spatial distributions of soil moisture and soil temperature were both affected by plastic mulching. As shown in Fig. 2a, the soil moisture in ridges was mostly higher than that in furrows from the effect of frequent drip irrigation. Figure 2c shows that the soil temperature in the mulched ridge was higher than that in the open furrow. However, in the later growth stages, the soil temperature in the furrow became similar to or even exceeded that in the ridge due to canopy development.

Plastic film mulching can also affect plant phenology. As shown in Fig. 2d, the LAI began to increase with seed germination, reached its maximum value during the bud stage in August, and then decreased with leaf fall. The LAI in the mulched field was higher than that in the non-mulched field during the comparative experiment year of 2016, particularly in the vigorous growing stages.

As shown in Fig. 3, the magnitude and amplitude of Rs were rather different in different years. For example, the soil respiration fluxes in the non-mulched ridges were 1–6, 4–7, and 3–11 µmol m-2 s-1, respectively, in the 3 years. The seasonal variation in Rs was generally mostly affected by soil temperature dynamics (this correlation will be further addressed in Sect. 3.4), although some anomalies occurred. For example, on DOY 180 in 2016, the Rs rates in the non-mulched ridge and planting hole reached peak values, while those in the furrows in both the mulched and non-mulched fields were fairly low. On the following DOY 192, however, the situation was reversed, and on DOY 235, all Rs fluxes experienced an abnormal declining and then rising cycle. These anomalies may be related to the SWC dynamics caused by irrigation and precipitation, which will be further explained in Sects. 3.5 and 3.6.

Rs showed significant spatial variability at the field scale. As shown in Fig. 3, the results in 2015 and 2016 indicated a consistently higher soil CO2 emission rate from the ridge than from the furrow in the non-mulched field. In the mulched field, as indicated by Fig. 3c, the Rs from the plastic film was very low, while the rate from the planting hole was higher than that from the furrow most of the time. For the Rs from the furrow, its rate in the mulched field generally exceeded that in the non-mulched field in 2015 and 2016 except on DOY 222 in 2016, which occurred just after a 12.8 mm rainfall event, as shown in Fig. 8.

Figure 4 depicts the seasonal accumulative Rs and precipitation over the three experimental years. It should be noted that the Rs from the mulched ridge is the area-weighted summation of the terms from the plastic mulch and planting holes. A prominent feature of the figure is that the Rs fluxes over the ridge and furrow in the mulched field are consistently larger than the corresponding values in the non-mulched field. However, this magnitude relation was not significant in the furrow in 2015 and 2016 or in the ridge in 2016 at a significance level of 0.05 (Table 1). Overall, the seasonal average Rs was 444.69 g C m-2 in the mulched field and 359.9 g C m-2 in the non-mulched field during the growing season over the 3 years. The accumulative Rs in the mulched field was indeed significantly larger than that in the non-mulched field in the years of 2014 and 2015. However, for the year of 2016, with a substantial precipitation amount of 130 mm, the positive deviation of the mulched field Rs was not significant.

Additionally, the difference in the furrow Rs between the mulched and non-mulched field was smaller than the difference in the ridge over all the 3 years, and the magnitude of such differences decreased from 2014 to 2016. It should be noted that the seasonal precipitation presented an increasing trend from 2014 to 2016. This indicates that more precipitation tends to eliminate the Rs differences between mulched and non-mulched fields.

AllRs fluxes in the different locations in the mulched and non-mulched fields showed increasing trends with temperature (Fig. 5), which were fitted using the exponential equation described in Sect. 2.3. However, their correlations are very weak and vary with location and time. The furrow possessed a higher R2 than the ridge because of the relatively stable soil moisture in the furrow. Additionally, the Q10 values in the furrows were much lower than those in the ridges.

The year 2014 was chosen to investigate the response of Rs to irrigation because of the very low number of precipitation events occurring in this year, and the results are shown in Fig. 6. It is clear that the soil moisture in the non-mulched ridge was always lower than that in the furrow in the mulched field except for some days immediately following irrigation. Relatively higher soil moisture favours soil respiration, and the Rs in the furrow in the mulched field was consequently always higher than that in the non-mulched ridge. Another dominant feature shown in Fig. 6 is the quick response of soil moisture and Rs to irrigation. The soil moisture experienced a quick increase after irrigation, while the Rs underwent a decline, which indicates that too much water in soil may restrain its respiration. Due to the configuration of the drip tape and plastic mulch, the soil moisture and respiration in the ridges of the mulched and non-mulched fields experienced similar but more drastic variations than those in the furrow.

To investigate the response of Rs to irrigation in more detail, the Rs dynamics across an irrigation cycle were explored. As Rs measurements were conducted on random dates between two irrigation events, data from different days after irrigation were collected to analyse the Rs variation. The effect of irrigation is presented by plotting Rs versus the number of days after irrigation during an irrigation cycle of approximately 6 days. The results in Fig. 7a show again that the Rs rate in the non-mulched ridge was extremely low immediately after irrigation and then slowly recovered, while irrigation had almost no influence on soil respiration in the furrow of the mulched field. The Rs rates from both the furrow and ridge reached maximum values on the fourth day after irrigation and then began to decrease over the soil drying process. The relation between Rs and soil moisture can be expressed in the form of a binomial equation, as shown in Fig. 7b, which indicates that Rs is very low in dry soil and increases with soil moisture. However, Rs shows a declining trend when soil moisture exceeds a certain threshold. The threshold is approximately 0.25 in the furrow of the mulched field and approximately 0.2 in the non-mulched ridge. The above thresholds are approximately 60 and 50 % of the water-filled pore space, respectively.

The year 2016 was chosen to investigate the response of soil respiration to precipitation because a significant amount of rainfall occurred in this year. As shown in Fig. 8, Rs exhibited similar responses to irrigation in the planting hole and plastic mulch and non-mulched ridges, while it presented similar responses to precipitation in the furrows of mulched and non-mulched fields. In particular, three large rainfall events, with amounts of 12.8, 36.8, and 48 mm, occurred on DOY 222, 192, and 235 in 2016, respectively. As we can see from Fig. 8, the light event (12.8 mm, DOY 222) had little effect on the soil moisture or Rs, the moderate event (36.8 mm, DOY 192) restrained Rs in the non-mulched ridge and planting hole but increased Rs in the furrows of the mulched and non-mulched fields, and the heavy event (48 mm, DOY 235) restrained Rs in all parts of the mulched and non-mulched fields.

Using the heavy event as an example, the effect of precipitation on Rs during a wetting–drying cycle was more closely investigated before and after the event. As shown in Fig. 9, the Rs rates at all locations were restrained by substantially high soil water content. Then, the Rs recovered slowly with a decline in SWC and remained steady after 3 days.

Our experiment indicates that the planting hole emitted more CO2 than the furrow, with the largest values of 8.0 and 6.6 µmol m-2 s-1, respectively, during the observation period (Fig. 3). In addition, the plastic mulch itself can also emit CO2 at a rate of 3.6 µmol m-2 s-1. Considering that plastic mulch occupies most of the ridge area, it is also an important pathway of CO2 emissions in mulched fields. In fact, the soil CO2 emission rate of the plastic mulch depends on film features, including its thickness, texture, and colour. For example, according to Berger et al. (2013), thick black PE mulch releases extraordinarily low N2O emissions, while high amounts of N2O can be emitted from polyethylene film that is only 0.02 mm thick (Nishimura et al., 2012). Q. Liu et al. (2016) also reported that transparent plastic film emits more CO2 than black plastic mulch. Local farmers in our study area often use clear polyvinyl chloride (PVC) film with a thickness of only 0.008 mm because of its low price. This film has a relatively high diffusion capacity for CO2, as indicated by our results. Additionally, thin and low-density plastic film is easily damaged, resulting in plastic film residue, which can affect crop germination, water absorption, nutrition, and yield. Plastic film residue can also inhibit soil microbial activity, which reduces soil fertility, causing substantive costs to the environment and farmers (J. Wang et al., 2016; Adhikari et al., 2016). High-density plastic film is therefore recommended for the purpose of reducing soil CO2 emissions and plastic film residues despite its higher price. In general, the planting hole, furrow, and plastic mulch are primary pathways that are responsible for CO2 emissions in a mulched field. A comprehensive measurement scheme including different locations is therefore necessary to assess Rs in a mulched field. Our results can potentially be used to correct the reported CO2 emissions measured only at the furrow in a mulched field (Qian-Bing et al., 2012; Q. Liu et al., 2016).

Our experiment also showed higher soil CO2 emission rates from furrows and ridges in the mulched field compared to the corresponding locations in the non-mulched field. Therefore, PFM can indeed promote soil respiration in our study area. This is principally due to improved soil temperature, soil moisture, and crop growth as a result of plastic mulching (see Fig. 2). Improved crop growth conditions result in the production of more root biomass and litter fall, which will promote root respiration and litter fall decomposition. Moreover, improved soil temperature and soil moisture can promote the activities of roots and microorganisms to increase the mineralization of soil organic carbon, for example, by stimulating the decomposition of buried crop straw (Y. P. Wang et al., 2016). This result was partly confirmed by Yu et al. (2016), who reported that furrow Rs in the mulched field is greater than that in the non-mulched field. However, they also reported that the Rs rates from mulched and non-mulched ridges are similar, which differs from our results. Furthermore, some other studies found contrasting results (i.e. PFM decreases Rs) in the northern Xinjiang Uygur Autonomous Region of China (Li et al., 2011), the Loess Plateau of China (Xiang et al., 2014), southwest China (Lei, 2016), and central Japan (Okuda et al., 2007). Additionally, Berger et al. (2013) found that PFM significantly decreases N2O emissions in South Korea. Therefore, the effects of plastic mulch on Rs differ in different areas. Our work reveals that the difference in Rs between mulched and non-mulched fields depends on the precipitation amount. This could be the reason leading to the contradictory results, which will be discussed in more detail in the following section.

Our results indicate that the substantially high SWC occurring right after irrigation and precipitation restrained Rs, and this effect decreased as the soil moisture returned to the normal level (Figs. 7a and 9). In contrast, in natural ecosystems, precipitation always immediately increases Rs, similar to water addition after a long drought in a tallgrass prairie ecosystem in Oklahoma, USA (Liu et al., 2002), and 12 mm of precipitation in an oak/grass savanna ecosystem in California (Xu and Baldocchi, 2004). This is due to the so-called soil degassing effect, which is the non-steady-state CO2 efflux at the soil surface occurring mostly during rainfall or irrigation after long periods of drought (Luo and Zhou, 2006). In agricultural systems, however, frequent irrigation occurs to satisfy crop water requirements and maintains favourable soil moisture. This further renders higher Rs than in natural ecosystems, particularly in arid areas. Our results further indicate that SWC that is either too low or too high can restrain Rs, which can be expressed by a quadratic equation (Fig. 7b). The quadratic (parabolic) relationship between SWC and Rs has also been detected in maize fields, a tallgrass prairie, and oak/grass savannah ecosystems (Yinkun et al., 2013; Xu et al., 2004; Liu et al., 2002; Mielnick and Dugas, 2000). This is because that lower water content affects the diffusion of soluble substrates, while higher water content affects the diffusion and availability of oxygen (Davidson et al., 2006; Linn and Doran, 1984). Our results confirm findings by Wang et al. (2010), who reported that irrigation stimulates Rs but that too much water reduces it, especially shortly after irrigation (Wang et al., 2010). In addition to our quadratic functional relation between SWC and Rs, the effect of SWC on Rs has also been described by linear, logarithmic, or parabolic functions in different ecosystems around the world (Davidson et al., 2000). For example, in a mountain oasis in Oman, soil respiration has been described to be linearly correlated with the SWC (Wichern et al., 2004). To be noted, the range of SWC in Wichern et al. (2004) is from 0.14 to 0.25, which is lower than the soil moisture threshold necessary to restrain Rs found in our study; therefore, the authors did not find a parabolic correlation like we did. More theoretical efforts should be made to reconcile these different experimental results and obtain a general relationship between SWC and Rs.

Our results indicate that the correlations between Rs and temperature and the temperature sensitivity (i.e. Q10) are rather low in our PFM field equipped with drip irrigation (Table 2). The obtained R2 values of 0.18–0.44 are much lower than the reported values from natural ecosystems, such as in a tall grass prairie in central Oklahoma, USA, with an R2 of 0.77–0.97 (Luo et al., 2001), and in the Harvard Forest in central Massachusetts, USA, with an R2 of 0.8 (Davidson et al., 1998). The obtained Q10 values of 1.25–1.65 (Table 2, except for in the planting hole) are below the median of 2.4 reported in a literature review of global soil respiration (Raich and Schlesinger, 1992). Additionally, they are much smaller than the Q10 of 3.8 found in rain-fed maize cropland on the Loess Plateau of China (Xiang et al., 2012). In contrast, relatively higher correlations between Rs and SWC indicate that the SWC may be the main factor affecting Rs in a PFM field under drip irrigation. Lower Q10 values indicate that the sensitivity of Rs to temperature has been weakened by higher variation in the soil moisture induced by irrigation and precipitation.

Our results clearly reveal the confounding influence of PFM and precipitation on soil respiration. The hydrological responses to precipitation in the field were changed by the impermeable plastic mulch, which is the reason that the effect of precipitation on Rs differed in the mulched and non-mulched fields. For example, the Rs rate in the non-mulched ridge was higher than that in the furrow of mulched fields and planting hole during 2016, in which high precipitation occurred. However, this result contrasted with the results from 2014 and 2015, during which less rainfall occurred. Additionally, although the soil respiration rate in the mulched field was always higher than that in the non-mulched field during all 3 years, the significance of this magnitude relation decreased with increasing precipitation. Therefore, we can speculate that the magnitude at which the mulch accelerates soil respiration should be related to the amount of precipitation.

To verify the above assertion, a meta-analysis was carried out. The relationship between the amount of annual precipitation (P) and the differences in the annual Rs (denoted as df, i.e. Rs in the non-mulched field minus that in the mulched field) was analysed (Fig. 10). The relevant studies include studies conducted in an arid area (P=45.7 mm) in southern Xinjiang (Yu et al., 2016), a semiarid area (P=160 mm) in northern Xinjiang (Li et al., 2011), a semi-humid area (P=566.8 mm) on the Loess Plateau of China (Xiang et al., 2014), a subtropical monsoon area (P=1105 mm) in southwest China (Lei, 2016), and a temperate monsoon climate area (P=1954 mm) in Japan (Okuda et al., 2007). The df was found to have a linear relationship with the amount of precipitation. Under the condition of 200 mm of annual precipitation, the Rs rates in the mulched and non-mulched fields are roughly identical. When the annual precipitation was greater than 200 mm, the Rs was lower in the mulched field than in the non-mulched field. This is the reason why the results of some studies contrasted with our results showing that PFM decreases Rs.

Based on the relationships between precipitation and soil respiration in the PFM fields obtained above, plastic film mulching is recommended for application in areas with precipitation greater than 200 mm, i.e. semi-arid and humid areas, to decrease soil CO2 emissions and increase soil carbon sequestration. Decreasing soil CO2 emissions indicates increasing soil organic carbon and maintenance of soil fertility to obtain a stable yield. Our results are consistent with those of Zhang et al. (2018), who concluded that PFM where precipitation is greater than 230 mm can result in a stable crop yield on the Loess Plateau.