Long-term changes in observed soil temperature over Poyang Lake Basin, China, during 1960–2016

A regional analysis of soil temperature (ST) is essential for improving our understanding of the soil thermal regime and its link with the atmosphere. This study attempts to assess trends in the Poyang Lake Basin (PLB) ST magnitude during 1960–2016 from station observations at multiple depths. The Mann–Kendall, Thiel-Sen, linear regression, and probability density statistics (PDF) are used for ST trend assessment with a significance level of 95%. The ST seasonal variability shows minimum values in winter (8℃) and maximum in the summer season (32 ℃). On an interannual scale, spring and winter seasons exhibited a significant increase in both land surface temperature (LST) (0.4℃, 0.4℃) and ST (0.3 °C, 0.15℃) magnitude than summer (LST − 0.1℃, ST 0.2℃) and autumn seasons (LST 0.3℃, ST 0.2℃). The northern basin exhibited a significant increase in LST, and ST magnitude, especially during the cold seasons (spring, winter) than the warm seasons. The maximum and minimum temperature trend and their diurnal difference infer an increase in the minimum temperature, especially during the summer, autumn, and winter seasons. The PDF further inferred that extreme cold events’ frequency decreased, and a significant increase in extreme warm events is obvious in the recent decade. The increasing trend in soil temperature magnitude is more in the northern basin than the high-altitude southern basin. Large-scale global warming and regional water and energy cycle changes can be the leading factors of such a warming trend.


Introduction
Global warming-induced climate change and extreme events are projected to modify the water and energy cycle (Renner et al. 2012).According to the Intergovernmental Panel's sixth assessment report on climate change (IPCC AR6), the global temperature was 1.09℃ higher in 2011-2020 than 1850-1900(IPCC 2022)).Compared with the global average temperature, the trend in temperature at high latitude and high altitude is greater (Sviličić et al. 2016;Thurgood et al. 2014;Yang et al. 2018;Zhu et al. 2018).Changes in temperature at the near surface correspond to changes in the land surface or skin temperature (LST), as well as in soil temperature (ST), which represents the subsurface temperature profile that supports biophysical processes (Couradeau et al. 2016).Changes in either of these two can impact land surface energy and water balance and are thus widely used to comprehensively assess global climate change (Knight et al. 2018).Both LST and ST are observed through in situ stations representing point scale information or satellites with relatively larger span and areaaveraged estimates (Hagan et al. 2019).Their applications further include its utility in developing the land surface hydrometeorological parameters retrievals (Parinussa et al. 2011).
The ST directly reflects the integrated changes in energy, mass, and water flux exchanges between the land and atmosphere (Hu et al. 2019;Zhang et al. 2001).Furthermore, ST mediates the phase change and water transport in the soil medium, influencing the physical, biological, and microbiological processes (Bond-Lamberty and Thomson 2010; Qin et al. 2020).ST influences soil microbial activity and higher nitrogen (N) holding capacity; on the other hand, a below-zero temperature can result in physical degradation of organic and inorganic media in the soil profile (Baptist et al. 2010;Edwards et al. 2007;Freppaz et al. 2008).Additionally, ST affects seed germination, transpiration, ideal sowing dates, irrigation intervals and treatments, and risk of infections in crops (Araghi et al. 2017;Calderón et al. 2014;Hasanuzzaman et al. 2013).In addition, instead of surface temperature, soil temperature from deeper layers has been demonstrated to have more information about the climate change impact.Short-term fluctuations are damped at depth, acting as a natural low pass filter on soil surface temperature time series (Knight et al. 2018).Hence, the study of ST and LST is of great importance.However, under the climate change context, most studies focus on the impact on air temperature while little attention is paid on the soil temperature.
As the largest freshwater lake in China, Poyang Lake is a relatively enclosed basin compared with other basins, analyzing its spatio-temporal variations and differences is of great significance to study the hydrological process and cycle within the basin.The Poyang Lake Basin also has the largest wetland system in China.It is one of the top ten ecological protection areas and one of the world's important ecological areas designated by the World Wide Fund for Nature.As a typical lake in the middle and lower reaches of the Yangtze River, Poyang Lake plays an important role in providing services and ecological functions such as flood storage, water source conservation, local climate regulation, biodiversity maintenance, and water supply for life, industry, and agriculture.With the intensification of climate change, as a typical basin in the subtropical monsoon climate region of the global climate system, the Poyang Lake Basin is reported to have a pronounced increase in air temperature since 1959, and the warming trend is basically consistent with that of the whole world.Reports based on meteorological observation data show that the air temperature had increased 0.65℃ during 1959-2008 and has an increasing rate of 0.15℃/10a.In the context of air temperature significant increasing trend in PLB, the temporal and spatial changes of soil temperature have still not been studied, which is disadvantageous to watershed climate and agricultural production.Facing the challenge of climate change, analyzing the impact of climate change on soil temperature and extremes in Poyang Lake Basin is of great significance for future flood control and drought resistance, water resource management, and ecosystem protection, and is also an urgent demand of global warming.
Studies based on soil temperature observations from in situ stations are reported to have a warming trend in many places such as Croatia (Sviličić et al. 2016) and Tibetan Plateau (Zhu et al. 2018).These studies also reported that the soil temperature from surface layer is more sensitive to the climate warming and shows the most significant warming trend during the long period observations.However, the study of PLB is still unexplored considering observational data scarcity.Some studies also found the warming trend was higher in soil temperature than in air (Knight et al. 2018;Shirvani et al. 2018).Where there is a lack of in situ data, some scholars used the ensemble of regional climate models and reanalysis data to analyze soil temperature trend in different regions and also found significant LST and ST warming in Swedish boreal forests (Jungqvist et al. 2014) and Tibetan Plateau China (Yang et al. 2020).However, due to soil properties, model forcing data, spatial scale mismatch, and parameterization scheme problems, soil temperature from these sources is also found to have deviations and uncertainties that could make the conclusions inconvincible (Yang et al. 2020).In addition, soil temperature retrieved from remote sensing satellite can only shed light on short-term soil temperature characteristics from shallower layers considering relative short time span.Hence, most studies on soil temperature for climatology research still rely on observational data.
In this study, ST observations from a dense meteorological network over the Poyang Lake Basin explore LST and ST trends during 1961-2016.Poyang Lake Basin is under continuous impact of climate change, where many studies have shown changes in surface water storage (Zhan et al. 2017;Zhou et al. 2016), temperature (Yue et al. 2015), and precipitation (Q.Zhang et al. 2014a, b), as well as increased occurrences of droughts (Zhang et al. 2017) and floods (Tian et al. 2011).The rest of the paper includes four sections; Section 2 describes the study area, data, and methods, Section 3 shows the results, Section 4 shows the discussion and conclusion.

Study area and soil temperature data
Poyang Lake is the largest freshwater lake in China, located at the southern bank of the Yangtze River (between 28° 22N-29° 45N and 115° 47E-116° 45N) with a drainage area of around 1.62 × 10 5 km 2 (Fig. 1).The Poyang Lake Basin (PLB) is situated in a subtropical monsoon climate, with an elevation of 23 m below the mean sea level to > 2000 m (Guan et al. 2017).The annual mean precipitation during 1978-2013 is about 1650 mm, and the annual average air temperature is about 18 °C (Zhan et al. 2017).The precipitation season starts from late spring (April) until the end of summer (August).It brings in a huge flux of water from the Yangtze River and its tributaries, which increases the spatial coverage of the Poyang Lake (Tang et al. 2016;Tao et al. 2014).
The data used in this study is derived from the TianQing system of the China Meteorological Administration.The TianQing system is a comprehensive meteorological big data platform that integrates meteorological observations from stations across the country, satellite remote sensing data, and model simulation data.Through the TianQing system, we obtained soil temperature data for specific regions, including Jiangxi Province.These data are based on real-time observations from meteorological stations and simulated data generated using advanced meteorological models and data assimilation techniques.The TianQing system offers data download and query functions, enabling us to conveniently access the required soil temperature data.The soil temperature data from the TianQing system undergo rigorous quality control and calibration processes, ensuring their reliability and high quality.By utilizing data from the TianQing system, we can efficiently acquire a large amount of soil temperature data within a short period, facilitating our research objectives.The DEM and the ST observatories of the PLB are shown in Fig. 1; there are 72 stations monitoring soil temperature at eight depths of 0, 5, 10, 15, 20, 40, 80, and 160 cm, respectively.In the current study, 0 cm represents the surface temperature/skin temperature expressed as LST, whereas the rest of the layers are termed as soil temperature (ST).Owing to the data completeness, we have used daily soil temperature data spanning 1960-2016.The soil temperature data was checked for any missing values and inconsistencies using a standalone and relative statistical technique called standard normal homogeneity test (SNHT) (Toreti et al. 2011;Ullah et al. 2019a, b).The purpose of using the SNHT was to check for inconsistency in ST due to changes in instrumentations, retrieval techniques, and other non-climate factors on a relative scale.
The PLB's topography and land cover pattern are rather complex, so for observing the regional variation of soil temperature, two specific regions are selected.The first region is comprised of agricultural crop-land located at (28.1-28.8°N, 114.7-116.2°E).In contrast, the second region is in the southern mountain (25.4-26°N, 114.2-116.2°E) in a relatively dense vegetated area.The local elevation (Fig. 1) of the two regions is another aspect that characterizes the selected regions' diversity.

Theil-Sen estimator
To estimate the ST magnitude's long-term trend, we have used an unbiased nonparametric statistical metric called the Theil-Sen (TS) estimator (Sen 1968).The TS estimator is insensitive to outliers if any exist in the data (Ahani et al. 2012;Dinpashoh et al. 2011).For a time series of N data points, the slope can be estimated as: In Eq. ( 1), x i and x j are the temperature values at time intervals i and j.Many studies have used this approach to produce robust estimates of the trends (Bhatti et al. 2020;Wang et al. 2018).

Mann-Kendall test
The nonparametric Mann-Kendall (MK) test (Mann 1945;Kendall 1975) analyzes ST magnitude trends' significance.The MK is a rank-based statistical method, and it can identify the possibility of statistically significant trends at multiple probability intervals.For the current study, a significance level of 5% is selected as a threshold to classify the trend as significantly positive or negative.The test statistics S is defined as follows: In Eq. ( 2), S is defined as the test statistics, and x 1 , x 2, … ..x n denotes the number of data points at time j .For any n ≥ 10, S is supposed to be normally distributed with a mean value of zero, as shown in Eq. ( 3).Its variance can be estimated with Eq. ( 4): In Eq. ( 4), n denotes the number of records, m shows the number of tied groups, and t i shows the number of data points.The standardized test statistics Z can be calculated through Eq. ( 5) as: The positive and negative values of Z indicate the presence of an increasing and decreasing trend, respectively.The null hypothesis is that no significant ST magnitude trend can be accepted if Z is not significant.When |Z|> 1.96, the original hypothesis is rejected, and the trend of the time series is not significant at the 5% confidence bound and vice versa. (2)

Extreme threshold definition
To assess the trends and identify the changes in extreme temperature events, we use the percentile approach (Ullah et al. 2019a, b;Zhang et al. 2008), with a predefined threshold as an indicator of the extreme event.The ST was arranged in ascending order, and then the 95th (5th) percentile was used as extreme temperature values.Days with soil temperature higher (lower) than the specified threshold were defined as an extremely high (low) soil temperature event.

Soil temperature profile
Figure 2a shows the percentage of valid ST observations at multiple soil depths; the percentage of valid ST observations is > 95% at 0, 5, 10, 15, and 20 cm depths, whereas the percentage of valid observations at 40, 80, and 160 cm depths is less than 25% respectively.For the current study, the depths considered are 0, 5, 10, 15, and 20 cm, respectively.Figure 2b shows the annual mean soil temperature; the deviation in mean LST (0 cm) with an average of 20.2 °C is relatively higher than 5, 10, 15, and 20 cm depths.At 5 and 10 cm depths, the climatological mean ST is 19.5 °C among all layers.Overall, the ST appears to increase exponentially with increasing depth and reaches 20.6 °C at 160 cm.In the spring and summer seasons (Fig. 2b1 and 2), the ST magnitude decreases with depth, while an increase with depth is obvious in the autumn and winter season (Fig. 2b3 and 4).
Changes in surface meteorological fields such as air temperature, relative humidity, rainfall, wind speed, and solar radiation may affect the vertical soil temperature distribution.The seasonal variation in ST is relatively uniform at 5 to 10 cm and 15 to 20 cm; thus, we select the land surface temperature (LST) at 0 cm and soil temperature (ST) (10 and 20 cm) for further analysis in the rest of the study.

Soil temperature climatology
Figure 3 shows the mean magnitude of LST (0 cm) and ST (10 cm and 20 cm) for spring (a1-3), summer (b1-3), autumn (c1-3), and winter (d1-3) seasons.For the spring season, LST exhibits a dipole pattern with a higher magnitude (> 21 °C) in the south than north (< 15 °C), where LST is consistently higher than ST.During the summer season, LST shows a maximum temperature of > 31 °C in the entire basin, decreasing north and south.The magnitude of ST is relatively lower than LST, with an average of 29 °C.During the autumn season, both LST and ST have the same magnitude across all depths (21-23 °C), and a decrease in magnitude is visible towards the north.For the winter season, both LST and ST have shown higher values at the southern basin (11 °C), with a consistent decrease in the northern basin (8 °C).
From Fig. 3, it can be inferred that both LST and ST are minimum during winter and maximum during the summer season.The seasonal mean LST and ST components in the north (Fig. 4a) and southern basin (Fig. 4b) closely followed the hydrometeorological cycle as reported in previous studies (Lai et al. 2014;Tian and Yang 2017;Tian et al. 2016;Zhan et al. 2017).

Trend in interannual soil temperature
Figure 5 shows the interannual Theil-Sen (TS) values of LST and ST for (a1-3) yearly, (b1-3) spring, (c1-3) summer, (d1-3) autumn, and (e1-3) winter seasons.The TS test statistic is used to calculate increase (decrease) in temperature magnitude on a decadal scale, whereas MK is used to test the trend's significance; the stations with no significant MK trend are shown as a white circle.The annual mean LST exhibited a significant increase of > 0.3 °C day −1 in the northern basin and > 0.2 °C day −1 in the southern basin.For ST, the 10 and 20 cm depths trend was > 0.2 °C day −1 , with multiple stations having no significant increase.For the spring season, LST and ST (10 cm) have exhibited a significant trend of 0.4 °C day −1 , whereas ST (20 cm) increased 0.3 °C day −1 relatively smaller increase from the surface layers.During the summer season, the LST trend is either not significant or limited to fewer stations with an increase (0.3 °C day −1 ) and a decrease (− 0.1 °C day −1 ) in others.For ST, relative to the LST, a higher number of stations have an increasing trend (0.2 °C day −1 ).During the autumn season, a significant trend persists in LST (0.3℃ day −1 ) and ST (0.1 to 0.2℃ day −1 ) across the entire basin.For the winter season, the trend exhibited is relatively higher for LST (0.4℃ day −1 ), whereas for ST, the magnitude of trend (0.2℃ day −1 ) is relatively smaller than LST.
Figure 6 further shows the linear trend of the LST and ST at the northern (region I) and southern basin (region II), respectively.The regression coefficient inferred a steady increase (0.017) in annual mean soil temperature.For the spring season, in the northern/southern basin, both LST and ST have increased with a regression coefficient of 0.03/0.016.For the summer season, a slight decrease/increase in soil temperature magnitude with a regression coefficient of − 0.003/0.003 is obvious in the northern/southern region, inferring different dynamics for the varying degree of the temperature trend.
For the autumn season, an increasing trend over both the northern/southern basins exists with regression coefficients of 0.019/0.018,respectively.For the winter season, the best fit line indicates an increasing trend with regression coefficients of 0.021/0.020each, respectively.
From Figs. 5 and 6, it can be concluded that in the PLB, both LST and ST has experienced a significant increase during the study period.During the summer season, the increase is either smaller or non-significant for LST and ST across the entire basin.The northern basin (region I) appeared to be warming relatively faster (Table 1) than the southern basin (region II).The intensity of the increase in temperature magnitude is relatively higher for LST; furthermore, cold seasons (winter and spring) appeared to be warming faster than the warm season (summer).The non-significant changes in summer could be due to consistent monsoon precipitation affecting soil thermal regime on a continuous scale.The findings are in agreement with Jungqvist

Interannual maximum and minimum LST trend
Figure 7 shows the trend in maximum and minimum temperature and their diurnal differences on an interannual scale.This maximum (T max ) represents the maximum temperature, which is usually recorded during daytime and vice versa for the minimum (T min ) temperature.In contrast, their respective difference shows the diurnal difference (T DTR ) for the entire basin.The TS and MK test statistics show the increase in magnitude and statistical confidence in estimating the trend.The stations with no significant trend are shown as white circles, respectively.
Figure 7a1 shows that the annual mean of T max has an increasing trend (0.3℃ day −1 ) in the northern basin, whereas in the southern basin, a decrease (− 0.02℃ day −1 ) is obvious.Furthermore, T min (Fig. 7a2) has shown a uniform increasing trend (0.03℃ day −1 ) over the whole basin; however, the MK statistics suggest that this increasing trend is not significant at multiple stations.The T DTR (Fig. 7a3) has a significant decreasing trend (− 0.3℃ day −1 ), inferring warming of the nighttime temperature.For the spring (Fig. 7b1-3) season, T max exhibited a significant trend (> 0.4℃ day −1 ) except the southeastern basin with a decreasing trend (< − 0.02℃ day −1 ) obvious for few stations.For T min , a significant trend (> 0.3℃ day −1 ) exists, which appeared to be relatively consistent in the northern basin than the rest of the basin.The T DTR for the spring season shows an increasing trend in the northern parts (> 0.2℃ day −1 ), while at the southern parts, a decreasing trend (< − 0.3 ℃ day −1 ) is persistent.The T DTR difference between the northern and southern basins could be attributed to regional variation in altitude, land cover, and agricultural practices.
During the summer season (Fig. 7c1-3), T max significantly decreased (> − 0.4 ℃ day −1 ), except few stations in the central basin with an increasing trend (0.4℃ day −1 ).The T min has a significant increase (0.3℃ day −1 ) in the entire basin, whereas the T DTR during the summer shows a decreasing trend (− 0.4℃ day −1 ).For the autumn season, both T max (Fig. 7d1) and T min have a significant increasing trend (> 0.4℃ day −1 ) in the entire basin.The T DTR (in Fig. 7d3) exhibited a increasing trend (− 0.3℃ day −1 ) in the western basin, whereas a decreasing trend (−0.3℃ day −1 ) in the eastern parts of the basin.During the winter season, T max (Fig. 7e1) has a significant increasing trend (0.2℃ day −1 ), while fewer stations have a slight decreasing trend (< = − 0.2 ℃ day −1 ).For T min , a significant increase (0.5℃ day −1 ) in the northern basin is relatively higher than in the southern basin (> 0.3℃ day −1 ).The T DTR exhibits a significant decrease (− 0.4℃ day −1 ) in the northern basin, while the southern basin has a relatively higher decreasing trend (> − 0.4℃ day −1 ).
From Fig. 7, an apparent asymmetrical pattern between T max and T min in the context of global warming exists.One obvious asymmetrical feature is that the increase in T min magnitude is greater than T max (except spring), inferring a higher night temperature trend.The T DTR inferred an increase in the minimum temperature during summer, autumn, and winter seasons, whereas during winter, the maximum temperature has a more significant increase than the minimum.Possible drivers of such increase could be the increased local droughts, extreme temperature, decreased evapotranspiration, and seasonal fluctuation in the water balance (Xu et al. 2014;Liu et al. 2011;Tao et al. 2014;Ye et al. 2013).
Figure 8 shows the probability density function (PDF) of the LST and ST during 1960-2016 using the 5 th percentile (− 5.98) and 95 th percentile (6.37) as extremes.For a detailed assessment of extreme events' frequency and intensity, both LST and ST are grouped into three phases (1960-1979, 1980-1999, and 2000-2016), respectively.In Fig. 8a1, the extremely low LST frequency illustrates a slight increase from the first and second phases, whereas there is a remarkable decrease in the third phase.For the extreme events, there is a relatively cold period from 1980 to 1999, and a considerable increase after 2000.Meanwhile, extreme low LST events' frequency decreases during this period, while extreme high LST events increased.Figure 8 b and c describe the ST (10 and 20 cm layers) PDFs; we see that the third phase's soil temperature values are higher than those in the previous two phases.
Compared to the LST, nevertheless, the PDFs of the 10 cm and 20 cm soil temperature become narrower, which is because the LST is more susceptible to the influence of wind, radiation, vegetation cover, and other factors than the temperature in deeper soil layers.Simultaneously, the extreme high and low ST events' annual frequency and intensity also indicate variation compared to the LST.Interestingly, at 10 cm and 20 cm depth, compared to the first phase, the second phase PDF exhibits lower ST values, which reveals that the second phase is a relatively cold period.The annual frequency of the extremely low ST in the 10 cm and 20 cm soil is higher than that of the LST.In summary, LST has been significantly increased after 2000, and the annual frequency of extreme high LST events also remarkably increased, almost twice as much as that in the second phase.For minimum LST events, the frequency significantly decreases with an almost invariable annual intensity.We note a relatively cold period over 1980-1999 in the 10 cm and 20 cm depths, while there is no such period for the LST.These findings suggest that the absolute values of regional-scale soil temperature experience a significant change.The annual frequency of the extreme high LST further shows a constant increase in recent years.These findings agree with the results of previous studies of Sviličić et al. (2016), who also reported an increase in soil temperature with the rising frequency and intensity of extreme events.This increase in temperature magnitude and extreme events can be associated with global-scale climate change, global warming, and enhanced greenhouse gas emissions (Nossal et al. 2016;Voigt et al. 2017).Considering such LST and ST changes may help overcome the crops' risk on multiple sowing and growth stages, stresses, crop rotation, and sowing dates locally.A detailed study focusing on the local influence of such an increase in LST and ST on soil biophysical and microbial activities may further help the agronomic aspects of agriculture.

Discussion and conclusion
The current study assessed the Poyang Lake Basin (PLB) soil temperature trend at multiple depths from 1960 to 2016.The soil temperature profile was classified as land surface temperature (LST) at 0 cm and soil temperature (ST) at 10 cm and 20 cm depths.Several innovative findings can be driven from this study: A significant increasing trend for LST and ST is observed throughout the PLB, and this warming is consistent with concomitant atmospheric warming there.Our findings agree with the previous studies that have reported a consistently increasing trend in soil temperature in different places such as Canada, Northeast Iran, and Croatia (Qian et al. 2011;Sviličić et al. 2016;Yeşilirmak 2014;Araghi et al. 2017;Kousari et al. 2013).In addition, warming is more pronounced in the surface layer.This is in accordance with other studies.In response to atmospheric changes in temperature, it was observed that trends in soil temperatures were delayed at deeper layers.Qian et al. (2011) explained this phenomenon as a result of the temperature changes being transmitted with depth through the soil profile.The warming trend is pronounced in cold seasons (0.22 to 0.32 °C/decades for LST) in this study.In the surrounding region, there are no similar studies on soil temperature trends or vulnerability due to extreme soil temperatures to which these results can be compared with.In more recent studies in the world (Qian et al. 2011;Yeşilirmak 2014), soil temperature in warm seasons is pronounced (summer, 0.25 °C/decades) compared with other seasons for western South-East Europe and Canada (Qian et al. 2011;Sviličić et al. 2016;Yeşilirmak 2014).This can be related to the different regional climates there.For the winter season in western South-East Europe and Canada, snow cover probably reduced the thermal insulation of snow on the underlying soil during winter months.In particular, the trends of maximum and minimum temperatures and diurnal ranges, as well as the regularities of extreme events, are also revealed in this study.While most studies focus on the mean soil temperature, the maximum and minimum temperature and extreme events are more influenced by global warming.In addition, all above findings in this study are analyzed with more recent and long spanning time from 1960 to 2016.More recent studies in the world (Qian et al. 2011;Yeşilirmak 2014) are dealing with soil temperature trend analysis in older periods 1958-2008 and 1970-2006, respectively.Since these observed periods are not reference climate or recent standard periods, the results of this study are thus significant in the climatological sense.
In PLB, variation of LST is influenced by many water circulation factors such as wetland area, precipitation, and evapotranspiration.The extent of the PLB's wetland area has shown dramatic variation at the seasonal scale, including an increase during April and October, attributed to a significant decrease in precipitation, evapotranspiration (ET), and upstream reservoir operation (Liu et al. 2021;Mei et al. 2016).Changes in precipitation and air temperature can affect soil moisture and the soil thermal regime; reduced precipitation and increased air temperature can enhance soil thermal processes and soil temperature (Yeşilirmak 2014;Qian et al. 2011).Both precipitation and air temperature in the Poyang Lake Basin have decreased/increased (Tian et al. 2016), whereas a substantial decrease in ET due to enhanced solar radiation during spring, summer, and autumn seasons might have influenced the trend in LST and ST.
Studies have shown the possible role of latent and sensible heat fluxes, leading to soil temperature fluctuation and heat transfer throughout the soil profile (Hu and Feng 2003).There exists a correlation between soil temperature and solar radiation (Yeşilirmak 2014), which may vary in each region/ season due to variations in precipitation, land cover, and topographic distribution (Tabari et al. 2011).Recent studies conducted over China suggest that a decrease in solar Fig. 7 The spatial distribution of the magnitude for the LST changes per decade and the MK test with relevance to their long-term mean values for T max (first column), T min (second column), and T DTR (third column) annually and for the four seasons during 1960-2016 ◂ radiation has altered this relationship, affecting soil temperatures over the Poyang Lake Basin and adjacent regions (Liu et al. 2015;Liao et al. 2015).An increase in soil temperature magnitude over the PLB can be linked to global warming and climate change, which is part of a larger picture, as suggested by Tian et al. (2016) andIPCC (2022).A relatively increased LST and ST during the winter and spring seasons could be linked with a decrease in the cold front events and wind speed (Lin et al. 2013;Jiang et al. 2010;Xu et al. 2006).This has possibly reduced the southward expansion of cold wind fronts during the winter and spring seasons, leading to increased soil temperature.Tian and Yang (2017) reported that wind speed variability increases near-surface temperature and precipitation over major river basins in China.Their study further suggests that the air temperature has shown relatively higher trends during cold seasons.In summer, the increase was relatively low, possibly due to monsoon precipitation affecting the soil thermal regime with ample water.
Although this study provides a comprehensive description of LST and ST variability based on relatively dense station distribution and reliable observation data, further work is necessary to investigate patterns of gridded ST data such as remote sensing and reanalysis data, as this remains unclear in PLB.A comprehensive analysis based on point-based (in situ) data and gridded data could provide a more precise picture of ST variation in PLB.Moreover, data uncertainty issue should also be considered in this and future study.For in situ station data that has a spanning 1961-2016, some of the stations suffered from location transferring; data from stations were affected by the surrounding environment and the nature of underlying surface.For remote sensing data, the accuracy of LST products can be influenced by aerosol contamination, cloud cover, surface elevation, etc. (Shen et al. 2020).When concerning reanalysis data, uncertainties also exist which depend on model construction and parameterization scheme.The findings of this study are restricted to LST and ST characterization.A thorough investigation on the dynamics and underlying mechanisms of change in soil temperature regimes is needed.As the warming of LST and ST is mutual influence by climate warming and human activities, previous studies have revealed that land cover change such as marshland loss, shrub encroachment (Shen et al. 2022), and urban expansion (Baqa et al. 2022) can have greater effect on LST than air temperature, because LST is more sensitive to land surface characteristics and surface energy budget.Therefore, to fully grasp the impacts of climate warming and human activities, future study should thoroughly investigate the dynamics and underlying mechanisms of change in soil temperature regimes (Shen et al. 2020).

Fig. 1
Fig. 1 Topography and study regions of the Poyang Lake basin (PLB); the black dots represent the density of the meteorological stations, and the black squares denote the two study regions each in the north and the south parts of the basin

Fig. 4
Fig. 4 Seasonal variations of the soil temperature for the period 1960-2016 at the (a) northern and (b) southern Poyang Lake Basin for selected depths et al. (2014),Qian et al. (2011),Yeşilirmak (2014), andSviličić et al. (2016), who reported a similar trend in soil temperature in a variety of climates.Recent studies suggested a positive relationship between soil temperature and near-surface meteorological fields, including surface temperature, evaporation, wind, and precipitation(Tabari et al. 2011;Zhai and Tao 2017;Yeşilirmak 2014).Thus, changes observed in these variables(Lu et al. 2009;Tian and Yang 2017;Zhai and Tao 2017) could be associated with increasing or decreasing soil temperature trends.

Fig. 8
Fig. 8 The PDF of the LST and ST, the intensity, and frequency of extreme high and low soil temperature events of (a) LST, (b) ST 10cm , and (c) ST 20cm in different phases from 1960 to 2016