Non-linear responses of net ecosystem productivity to gradient warming in a paddy field in Northeast China

Study site

Our experiment was conducted from May to October 2018 in a paddy field (123°33′E, 41°49′N) at Shenyang Agricultural University. The experimental field is in a semi-arid and semi-humid monsoon climate with a mean annual temperature of 13.1 °C and a mean annual precipitation of 610 mm. North-japonica 2 seedlings were transplanted into the growing field on May 31, 2018, with a row space of 30 cm ×13 cm and 3 plants per pot. Fertilizer was applied to the field using a base fertilizer (96 kg N, 69 kg P₂O₅ and 45 kg K₂O), a greening fertilizer (34.5 kg N), and a tiller fertilizer (34.5 kg N) every hectare. The applied water volume and weeding techniques were similar among treatments and were related to local practices.

Experimental design

We used a far infrared open warming system (Dong et al., 2011; Nijs et al., 1997) consisting of a heating component, a power element, and control and monitoring parts. The heating component had a far infrared heated lamp, an iron bracket, and a white stainless steel reflector mounted 60–70 cm above the canopy that was manually lifted according to the rice height. Power was provided by an AC 220V power station. A computer served as the control part while the thermocouple thermometer monitored the real time air temperature around the canopy and below the lampshade.

The warming magnitude was set to increase approximately 1.5 °C, in accordance with the Paris Agreement to control the global mean temperature within that measurement. Different warming magnitudes were achieved using far infrared lamps with different powers, given that 1000 W warming per hour increases the temperature approximately 1 °C. The 5 warming gradients were: control (CK), W1 (500 W), W2 (1,000 W), W3 (1,500 W), and W4 (3,000 W). Each treatment was replicated three times in an area 2 m ×2 m randomly distributed in the field. The intervals among the replicates and treatments were 3 m and 4 m, respectively.

Carbon flux measurements

From June to October 2018, NEP and ER were measured by the static chamber-infrared gas analyser method every 20 days on sunny days between 8 a.m. and 11 a.m. (Niu et al., 2013). Five measurements were taken during the growing season.

A square PVC frame 0.5 m ×0.5 m was inserted into the soil of each plot at a depth of approximately 15 cm. A transparent chamber (0.5 m ×0.5 m ×1.2 m) was buckled to the frame and two small electric fans were installed on either side. The buckled chamber was attached to an infrared gas analyser (CIRAS-3 Portable Photosynthesis System, PP Systems, USA) to measure the CO₂ concentration. The CO₂ concentration in the chamber was recorded at 2-second intervals for a 90-second period after the transparent chamber was run for 30 s, to avoid capturing the interference of human activities. NEP was calculated based on 45 continuously observed CO₂ concentrations (ρ_c) in the transparent chamber.

The chamber was lifted into the direction of the wind and covered with a black cloth once the NEP observations were completed. The above steps were repeated to measure the CO₂ concentration in the chamber. The carbon exchange between the ecosystem and the atmosphere was considered to be the ER since the chamber was shielded from light.

Data analysis

We used 45 continuous CO₂ concentration (ρ_c) observations to estimate NEP and ER by calculating the ρ_c change rate (dρ_c /dt) from a linear fitting:

NEP $=\frac{\mathrm{V}}{\mathrm{A}}\frac{d{\rho }_c}{dt}$ E$\mathrm{R}=\frac{V}{A}\frac{d{\rho }_c}{dt}$

where NEP is the net ecosystem productivity, ER is the ecosystem respiration, V is the chamber volume, A is the chamber covered area, ρ_c is the CO₂ mass concentration, and t is time.

Gross primary productivity (GPP) was calculated as the algebraic sum of NEP and ER.

The repeated-measures analysis of variance (ANOVA) was used to assess the effects of the warming treatments on NEP, ER, and GPP. The intrinsic effect test and the inter-subject effect test were conducted to investigate the influences of measuring dates and warming treatments on NEP, ER, and GPP, respectively. Linear regression was employed to examine the responses of NEP and its components to temperature, and relationships between NEP and its components, whose performance was validated by both p value and R². All statistical analyses were conducted with SPSS software (SPSS 22.0 for Windows, SPSS Inc., Chicago, IL, USA).

Effects of warming on NEP

The measuring dates, warming treatments, and their interactions were all found to significantly affect NEP (Table 1, p < 0.01).

NEP showed a convex parabolic curve with a significant difference between measuring dates (Fig. 1A) during the growing season. From late June to early August, NEP monotonically increased and reached its peak (20.47–25.32 µmol m⁻² s⁻¹) in early August. However, NEP decreased from early August and reached its lowest value (6.19–10.70 µmol m⁻² s⁻¹) in the middle of September.

NEP significantly differed among treatments (Fig. 1A, p < 0.05). All warming treatments had a significantly higher NEP than the control (CK). In addition, different warming magnitudes resulted in significant differences in NEP. The highest NEP appeared at 1500 W warming (W3), corresponding to a temperature increase of 1.37 °C, with an NEP increase of 6.01 µmol m⁻² s⁻¹. Lower warming magnitudes both had an NEP increase, though the NEP increasing magnitudes were lower than that at 1500 W warming (W3). 500 W warming (W1) and 1000 W warming (W2) only led to an NEP increase of 2.58 and 3.89 µmol m⁻² s⁻¹, respectively. The minimum increase appeared at 3000 W warming (W4) with an NEP increase of 1.16 µmol m⁻² s⁻¹.

Furthermore, different warming treatments resulted in diverse NEP increase rates (Figs. 2A–2E) during the whole growing season. CK had the lowest rate increase (0.92µmol m⁻² s⁻¹ per 1 °C temperature increase, Fig. 2A) while W2 had the highest rate increase (1.48 µmol m⁻² s⁻¹ per 1 °C temperature increase, Fig. 2C). Other warming treatments had a similar increasing rate around 1.3µmol m⁻² s⁻¹ per 1 °C temperature increase (Figs. 2B, 2D, 2E).

Therefore, NEP responded non-linearly to the gradient warming (Fig. 1B). As temperatures increased, the NEP revealed a convex parabolic curve with its peak value appearing between 1500 W and 2000 W warming, while its peak increasing rate occurred around 1000 W warming.

The increasing temperature led to a linear increase of NEP (Fig. 2F) with the combined effects of measurement dates and warming treatments. A 1 °C temperature increase resulted in an NEP increase of 1.16 µmol m⁻² s⁻¹ with an R² of 0.60.

Effects of warming on ER

The measuring dates, warming treatments, and their interactions also significantly impacted ER (Table 1, p < 0.01).

ER showed a convex parabolic curve over time during the growing season and differed significantly over the dates of measurement (Fig. 3A). From late June to early August, ER showed an upward trend and reached its peak (11.05–17.82 µmol m⁻² s⁻¹) in early August. ER started to decrease after early August and reached its lowest value at the end of the growing season.

ER differed significantly among treatments (Fig. 3A, p < 0.05). All warming treatments had a higher ER than the control (CK), which had the lowest ER of 1.63–3.38 µmol m⁻² s⁻¹. The highest ER appeared at 1,500 W warming (W3), which corresponded with a temperature increase of 1.37 °C, with an increase of 4.04 µmol m⁻² s⁻¹. 500 warming (W1) and 1,000 W warming (W2) resulted in an ER increase of 1.69 µmol m⁻² s⁻¹ and 2.75 µmol m⁻² s⁻¹, respectively, which were both lower than that of W3. 3000 W warming led to the least ER increase of 0.91 µmol m⁻² s⁻¹.

Different warming treatments induced diverse ER increase rates during the growing season (Figs. 4A–4E). CK and W3 had the lowest (0.64 µmol m⁻² s⁻¹ per 1 °C temperature increase, Fig. 4A) and highest (1.38 µmol m⁻² s⁻¹ per 1 °C temperature increase, Fig. 4D) rate of increase, respectively. Other warming treatments had a similar rate of increase of approximately 1.0µmol m⁻² s⁻¹ per 1 °C temperature increase (Figs. 4B, 4C, 4E).

ER responded non-linearly to gradient warming (Fig. 3B). As temperatures increased, ER showed a convex parabolic curve with its peak value appearing between 1,500 W and 2,000 W warming, whereas the highest rate increase occurred at 1,500 W warming.

With the combined effects of measuring dates and the warming treatments, ER increased linearly with the increasing temperature (Fig. 4F). A 1 °C temperature increase resulted in an ER increase of 0.92 µmol m⁻² s⁻¹ with an R² of 0.63.

Effects of warming on GPP

The measuring dates, warming treatments, and their interactions also significantly affected GPP (Table 1, p < 0.01).

Measuring dates significantly affected GPP, resulting in a convex parabolic curve of GPP during the growing season (Fig. 5A). From late June to early August, GPP monotonically increased and peaked (31.52–43.14 µmol m⁻² s⁻¹) in early August. GPP began to decrease after early August and reached its lowest values (7.82–14.06 µmol m⁻² s⁻¹) at the end of September.

GPP also significantly differed among treatments (Fig. 5A, p < 0.05). All warming treatments took a significantly higher GPP than the control (CK). In addition, GPP differed significantly among warming treatments. The highest GPP corresponded to a GPP increase of 10.05 µmol m⁻² s⁻¹ at 1,500 W warming (W3), which was associated with a temperature increase of 1.37 °C. 500 W warming (W1) and 1,000 W warming (W2) led to a GPP increase of 4.27 µmol m⁻² s⁻¹, and 6.63 µmol m⁻² s⁻¹, respectively. 3,000 W warming (W4) only increased GPP 2.07 µmol m⁻² s⁻¹.

Furthermore, different warming treatments resulted in diverse GPP increase rates (Figs. 6A–6E) during the growing season. CK had the lowest rate of increase (1.56µmol m⁻² s⁻¹ per 1 °C temperature increase, Fig. 6A), whereas W2 and W3 had the highest rate of increase around 2.6 µmol m⁻² s⁻¹ per 1 °C temperature increase (Figs. 6C–6D). Other warming treatments had a similar rate of increase around 2.0 µmol m⁻² s⁻¹ per 1 °C temperature increase (Figs. 6B, 6E).

GPP responded non-linearly to the gradient warming (Fig. 5B). As temperatures increased, GPP showed a convex parabolic curve, with its peak value occurring between 1,500 W and 2,000 W warming, whereas the highest rate of increase occurred at 1,500 W warming.

GPP linearly increased with the increasing temperature when the combined effects of measuring dates and warming treatments were considered (Fig. 6F). A 1 °C temperature increase resulted in a GPP increase of 2.08 µmol m⁻² s⁻¹ with an R² of 0.64.

Relationships between NEP, GPP, and ER

NEP showed a significantly positive relationship with GPP during the measurement period. This was also true for the relationship between NEP and ER (Fig. 7, p <0.01). A 1 µmol m⁻² s⁻¹ GPP increase led to an NEP increase of 0.57 µmol m⁻² s⁻¹, with an R² of 0.97. However, a 1 µmol m⁻² s⁻¹ ER increase resulted in an NEP increase of 1.19 µmol m⁻² s⁻¹, with an R² of 0.84. Therefore, GPP showed a closer relationship with NEP than ER.