Rising soil temperature in China and its potential ecological impact

Global warming influences a series of ecological processes and ecosystems’ stability. Although comprehensive studies have been done to investigate responses of various ecosystem processes to rising air temperatures, less is known about changes in soil temperatures and their impact on below-ground processes, particularly in deep layers. Herein, we used 50 y of temperature data (1962–2011) from 360 sites in China to assess spatio-temporal changes in soil temperatures from the surface to a depth of 3.20 m. We determined, apparently for the first time, that soil surface temperature increased 31% more than air temperature, potentially leading to more carbon release to the atmosphere than predicted. Annual mean surface temperature increased by 2.07–4.04 and 0.66–2.21 °C in northern and southern China, respectively, with the greatest in winter. Warming occurred as deep as 3.20 m. The soil temperature rise was predicted to have increased soil respiration by up to 28%, reinforcing climate warming and extending the potential growing season by up to 20 d across China. However, use of only air temperature to estimate soil temperature changes would underestimate those impacts. In conclusion, these results highlighted the importance of soil warming and of using soil temperature to assess and predict soil processes.

more warming in certain soil layers with less warming in others (across regions) as compared to changes in air temperature (Table 1).
Over the entire country, on average soil warming was predicted to have increased annual soil respiration by 8.07% (range, 2.59 to 16.92%) in the top 0-0.20 m soil layer (the critical zone of biological activity where most carbon and nutrients are concentrated). Across seasons, the predicted increase in soil respiration was greatest in spring (12.87% on average) and second greatest in summer, except in Southwest China, East China and Southeast China, where the greatest increase occurred in winter (Fig. 4). Soil warming in winter in northern regions had limited effects on soil respiration, because winter temperature in those regions was usually below zero (despite warming), thereby limiting soil biological activity. Across regions, the greatest increase of soil respiration was predicted to have occurred in northern and/or western regions where climate is cooler (Fig. 4). We concluded that the predicted increase in soil respiration over the 50-y interval based on soil temperature increase was much higher than those based on air temperature increases in northwest cooler regions, particularly in spring and summer (up to 16.64% higher; Fig. 4 and Supplementary Fig. 1). We concluded that increased soil respiration enhancement by warming are likely to be underestimated if predictions are based on increases in air versus soil temperature.    Based on the change of daily mean surface soil temperature, the estimated date of temperature exceeding 0 °C in spring had advanced, whereas that of temperature decreasing to below 0 °C in autumn had delayed, thereby increasing the duration for surface temperatures exceeding 0 °C (Fig. 5). This duration roughly represents the potential growing season, during which many temperate plants can actively grow. Averaged across China, the potential growing season was extended by approximately 12 d over the 50 y. The greatest extension occurred in North Central China and Tibetan Plateau (20 d), whereas the smallest extension was in East China (approximately 6 d), with no change in Southeast China. The potential growing season extension shared a similar regional pattern with surface soil temperature changes. However, the extension of duration for daily minimum soil surface temperature above 0 °C, roughly representing the frost-free period, was much greater (Fig. 5). Averaged across China, the estimated annual frost-free period was extended by 20 d, mainly resulting from a substantial increase in East China (35 d). The estimated extension with soil temperature changes was 1 and 5 d longer for potential growing season and frost-free period, respectively, as compared to those estimated using air temperature changes ( Supplementary Fig. 2). This highlighted the importance of using soil temperatures in ecological studies where soil temperatures are more relevant (e.g., for processes involving soil or early crop growth).

Discussion
Our results apparently provided the first nationwide images and direct evidence that soil had significantly warmed in China from 1962 to 2011, consistent with previous changes in air temperature [32][33][34][35] , but with a greater warming at soil surface and warming extended to a depth of 3.2 m 34 . We detected a continuous warming trend in soil from 1962 to 2011 in China, without an intermediate cooling period as reported 36 . The predicted increase in soil respiration due to soil warming can be considered an indicator for accelerated decomposition of soil organic matter, increased loss of soil carbon and turnover of soil nutrients. That the extended warming occurred at a depth of 3.2 m (limit of monitoring in this study) indicated a need to investigate its impact on soil processes in deep soil layers, as previous studies usually focused on shallow layers. Soil respiration change over time can be affected by many factors, e.g. vegetative cover change, changes in microbial community etc. There was not enough data to allow analysis of all these influences. However, our results illustrated potential impacts of soil temperature change alone on soil respiration. Hence increased soil respiration needs to be considered together with the changes in the carbon sink in above-and below-ground biomass in calculation of carbon balance in Chinese ecosystems 37 . Soil warming is expected to influence agronomic practices (e.g., sowing date), crop production, and other ecosystem processes such as plant phenology, vegetation shift, niche diversity and ecosystem stability. Such changes evaluated using air or soil temperatures estimated from air temperatures may need to be reviewed and revised, using detailed soil temperature data.
We made substantial efforts to verify the quality of the datasets we used. Consequently, we excluded stations where changes in measurement sites and instruments occurred, and where there were large data gaps. Ultimately only 360 stations were used and there different amounts of data for a given depth in a given region, resulting in some gaps (Fig. 3). Urbanisation has also influenced soil temperatures for the last few decades 32,38 . Although its impact is difficult to assess, that it was expected to affect both air and soil temperatures, it should not invalidate our conclusions.
We used a simple approach to estimate potential effects of warming on soil respiration, as one example for changes in soil biological processes. In the absence of soil temperature measurements, most ecological models use air temperature to simulate soil temperature with simple approaches consisting of a delay in time and damping effect with depth (e.g., APSIM 39 ). Consequently, estimated soil temperature from such models would have the same long-term temporal changes as those of the air temperature, which would underestimate long-term soil warming and the subsequent impact on ecological processes (e.g., respiration). Thus, our results highlighted the importance to consider differential warming in soil and air. Of course, air and soil temperature changes would have also influenced other ecological processes (e.g., vegetation growth and soil microbial community), leading to more complex influences on soil processes than what our simple approach would suggest. Nonetheless, our results provided substantial insights into potential impacts of soil warming on ecological processes.
Global warming has not only increased air and soil temperatures, but also lead to other changes, including precipitation distribution and vegetation shift. Concurrent changes in temperature and precipitation often cause more frequent drought in many areas 34 , affecting both plant and soil processes, and ecosystems' responses in general. Increasing winter soil temperatures created more favourable conditions for insects and diseases to survive the winter 8 . In addition, warming also led to faster melting of snow to reduce winter season snowpack and, hence, further affect water availability. All these processes can lead to more complex impacts on ecosystems, although that is beyond the scope of the current study, warranting further research.

Methods
Data source and collation. Temperature data were obtained from China Meteorological Administration (CMA) and provincial Meteorological Administration (http://data.cma.cn/data/detail/dataCode/SURF_CLI_ CHN_MUL_DAY_V3.0.html), covering 759 national standard meteorological stations across China. These data, which are not yet publically available, included daily maximum, minimum and mean soil temperatures, and monthly maximum, minimum and mean air temperatures from 1962 to 2011. This was apparently the first report of a long-term, nationwide analysis of soil temperature change. All temperature data were measured (CMA standard protocols) in the national CMA weather station network. Soil temperature was measured at eight depths, namely at the surface (0 m), in shallow soils: 0.05, 0.10, and 0.20 m, and in deep soils: 0.40, 0.80, 1.60 and 3.20 m. Surface soil temperature was measured with a glass mercury thermometer, with half buried in soil and the other half exposed in the air. Temperatures in shallow soil were measured using curved glass tube mercury thermometers with the sensing head buried in soil, whereas temperatures in deep soil were measured using glass straight mercury thermometers with the sensor buried at the corresponding depths. All observation sites were bare soil. Temperatures were recorded, on average, four times daily (02:00, 08:00, 14:00 and 20:00) to calculate daily mean soil temperature at eight layers. Furthermore, daily maximum and minimum surface temperatures were recorded by maximum and minimum thermometers.
We eliminated stations that experienced changes in site of measurements, instruments, and severe interruption of continuous soil temperature records (see below). Most weather stations in China began operation in the 1950 s. Because of instrument malfunctions in the earliest years, temperature data before 1962 contained more gaps, especially for soil temperature. Therefore, we used only data from 1962 to 2011. For soil temperature, detailed data quality check involved exclusion of months, with more than one third missing data and exclusion of data outliers. For any month, two thirds of daily records must have been available. This criterion reduced the number of stations (with total 12 mo valid data) to 360. Therefore, daily soil temperature records from 360 stations were used to calculate monthly average soil temperatures. For any month and station, if the calculated average temperature departed from the 50-y average (1962-2011) for that month by three times of the standard deviation from the mean, the value was treated as an outlier, and replaced by the average at the three nearest stations for that month and year. Then average monthly soil temperatures ( = … A i , 1, 2, , 50 i ) at each site were used. In contrast, monthly air temperatures at the 360 stations were used directly.
Soil temperature change. Trends for changes in soil temperatures were analysed using the monthly averages across 50 y (for each month, station, region and whole China) together with linear regression and Student's t-test for significance 40 . The rate of change per year was derived as the slope of the linear regression line (k), and the total change in 50 y was calculated as k × 50.
For each season, i.e. winter (Dec-Feb), spring (Mar-May), summer (Jun-Aug), autumn (Sep-Nov), seasonal average were calculated from those of the three months first, and then seasonal absolute changes were obtained using the same regression process for monthly trend and annual trend.
In order to assess regional patterns of temperature changes, we divided China into eight climatic regions, i.e. Northwest China, North Central China, Northern Plains, Northeast China, Tibetan Plateau, Southwest China, East China, and Southeast China 41 . The eight regions generally reflect climatic and vegetation conditions in China, coinciding roughly with the country's socioeconomic macro-regions. We averaged changes at individual stations to calculate regional and notational mean changes. Potential impact of soil temperature change on soil respiration. We used a simple but commonly adopted approach to estimate potential impacts of soil temperature change on soil respiration 24,25 . Soil respiration at a typical temperature can be expressed as: T T 0 1 0 ( ) /10 0 Where R and R 0 are soil respiration at soil temperatures T and T 0 , respectively, and Q 10 is the factor by which respiration is multiplied when temperature increases by 10 °C. We assumed that Q 10 was equal to 2, a commonly accepted value 24 .
The relative percentage change of soil respiration (P) after soil temperature change of ∆T (i.e., T − T o ) can be calculated as: We calculated the P with both air (P air ) and soil (P soil ) temperature changes using ∆T between 1962 and 2011 estimated based on the absolute change. For this purpose, soil temperature at 0.10 m was used, representing the average temperature in the 0-0.20 m soil layer. We further estimated the percentage difference of P soil relative to P air , i.e., soil air air Because both P air and P soil were calculated based on actual changes in air and soil temperatures, this calculation enabled us to assess potential bias of soil respiration response predictions based on changes in air temperatures. In many ecological models, soil temperatures were simulated from air temperatures; therefore, changes in the simulated soil temperature would be the same or very similar to those in air temperature. Consequently, such models would underestimate impacts of warming on soil respiration.
Change in the duration of potential growing season and frost-free period. The duration in days from the time when daily mean surface soil temperature raised to above 0 °C to the time when it decreased below 0 °C was used herein to represent the potential growing season for crops 42,43 . We determined this duration for each year and station using the monthly average of daily mean surface soil temperature. The two adjacent months with soil temperatures below and above 0 °C were identified first. Then, linear interpolation of the two monthly temperature values was used to estimate the dates (we called '0 °C date') when the temperature raised and dropped to 0 °C. The use of monthly average of daily temperatures eliminated the need to calculate the moving averages of daily temperatures for determination of the dates when temperatures steadily remained above or below 0 °C. The duration in days between the two '0 °C date' was then calculated for each year and each station. The trend of change in durations was analysed using the same method as applied to soil temperatures. Many eco-biological processes may be activated when the minimum surface soil temperature rises above 0 °C 29,31 ; therefore, we defined the frost-free period using the change in minimum surface soil temperature, and the same method was used to define potential growing period. We also calculated these two durations using air temperatures for comparison purposes.