Northeast China (NEC) has become one of China's most obvious examples of climate change because of its rising warming rate of 0.35 °C/10 years. As the indicator of climate change, the dynamic of surface soil moisture (SSM) has not been assessed yet. We investigated the spatiotemporal dynamics of SSM in NEC using a 32-year SSM product and found the following. (1) SSM displayed the characteristics of being dry in the west and wet in the east and decreased with time. (2) The seasonal difference was found for the temporal dynamics of SSM: it increased in summer and decreased in spring and autumn. (3) For all four regions studied, the temporal dynamics of SSM were similar to those of the whole of NEC, but with different rates of SSM change. Moreover, SSM in regions B and D had a lower spatial variance than the other two regions because of the stable spatial pattern of cropland. (4) The change rates for SSM were consistent with that observed for the warming rates, which indicated that SSM levels derived from remote sensing data will correlate with climate change. In summary, a wetter summer and a drier spring and autumn were observed in NEC over the past 30 years.

climate change, microwave remote sensing, spatiotemporal dynamics, surface soil moisture

Surface soil moisture (SSM) is one of the key variables in hydrology and meteorology. It controls the infiltration of rainfall through the soil and the surface evaporation of moisture from the soil surface. It also affects water and energy exchange between the land surface and the atmosphere, which in turn affects the global weather system (de Rosnay et al. 2009; Jung et al. 2010; Zheng & Zhao 2010; Liu et al. 2013).

In China, past investigations of the spatial and temporal dynamics of soil moisture were based on meteorological station data (Huang & Ding 2007; Zuo & Zhang 2008). The relationships between soil moisture and other climate change variables (e.g. air temperature and rainfall) have also been studied (Chahine 1992); a positive relationship was shown between soil moisture and rainfall, while a negative relationship was shown between soil moisture and air temperature. This indicates that soil moisture is an effective indicator in climate change research.

In Northeast China (NEC), soil at different depths contains varying degrees of moisture, especially in the spring and autumn (Zuo & Zhang 2008). In the summer, soil moisture shows clear vertical distribution characteristics, with drier conditions in the topsoil and wetter conditions further down (Huang & Ding 2007). In addition to the meteorological station data, reanalysis products of SSM such as ERA-Interim (Zeng et al. 2015), ERA40 and NCEP (National Centers for Environmental Prediction) (Zuo & Zhang 2008) have also been used to detect changes in soil moisture.

Soil moisture measurements from the meteorological stations are spatially discrete, and one meteorological station can represent a spatial range of tens to hundreds of square kilometers. Because of the heterogeneity of micro topography and vegetation, the spatial range represented by SSM measurements is not constant between stations. SSM data collected by a given meteorological station represent a spatially averaged value with a depth of around 0–10 cm. Similarly, soil surface albedo, which affects the surface energy balance, is more sensitive to SSM at a shallower depth (Zhang et al. 2011). SSM retrieved from remote sensing data can respond to the changes of SSM in the first 5 cm of topsoil, and this depth varies with the electromagnetic wavelength used by the satellite sensor (e.g. advanced microwave scanning radiometer – earth observing system (AMSR-E) or microwave imaging radiometer with aperture synthesis) and with the amount of SSM itself. A deeper detection depth can be achieved by using longer wavelengths on drier soil.

Uncertainty must be evaluated before remotely sensed SSM data can be used in further analysis. Li et al. (2013) compared two reanalyzed soil moisture datasets (European Center for Medium-range Weather Forecasting (ECMWF) and NCEP) and one remotely sensed soil moisture dataset (AMSR-E). The results showed that SSM from AMSR-E, ECMWF and NCEP had a consistent spatial distribution, and that AMSR-E had less SSM than ECMWF and NCEP. Compared with in situ SSM data from the meteorological station, AMSR-E was drier. The relative soil moisture data from the meteorological stations were converted into volumetric soil moisture based on field capacity and soil density to produce the same units as used in the AMSR-E SSM product. Despite this, the sensed depth of AMSR-E was approximately 0–3 cm, shallower than the sampling depth of 0–10 cm at the meteorological station. Huang & Ding (2007) observed that the top soil moisture was drier than the deep soil; it is reasonable that SSM from AMSR-E (with 0–3 cm sensed depth) was drier than the meteorological SSM (with 0–10 cm sensed depth). Liu et al. (2013) found that the spatial distribution of SSM in the Tibetan Plateau was consistent with the spatial distribution of rainfall by comparing the AMSR-E SSM and meteorological rainfall data from 2003 to 2010. All these studies show that SSM derived from microwave remote sensing data has the potential to support research into the spatial and temporal variations of SSM. Furthermore, remotely sensed SSM data have a relatively fine spatial resolution compared to meteorological data and reanalysis data, which have been used to evaluate the temporal and spatial changes in soil moisture, both regionally and globally (Liu et al. 2013; Rötzer et al. 2015). Remotely sensed SSM is likely to become the main method for evaluating the change in soil water storage at the land surface because of its spatial and temporal frequency.

Our objective is to use SSM values derived from remote sensing data to analyze the spatial and temporal variations of SSM in NEC, an area sensitive to climate change, to assess how SSM has changed in NEC over the past 30 years; and to evaluate the change characteristics of SSM. We make a brief conclusion about SSM dynamics in NEC, and discuss their potential effects on ecological and environmental change.

The study area

NEC (38°24′00″–53°20′24″N, 115°03′00″–135°01′12″E) is located at the middle to high latitudes of the northern hemisphere. It contains three provinces (Heilongjiang, Jilin and Liaoning) and four cities in Inner Mongolia (Hailaer, Chifeng, Xilinguole, Tongliao), and covers an area of 7.873 × 105 km2, occupying 8.2% of China's total land area (Figure 1). The altitude of the region declines from the north to the south, and its heartland is the NEC Plain. Various types of land cover are unevenly distributed in the study area. Cropland and forest are the main land cover types in NEC. Heilongjiang Province has the richest forest resources of all three provinces, and is also one of China's water-rich provinces. Jilin Province is one of China's six major forestry areas. Liaoning Province has rich water resources. Most grassland in NEC is distributed in eastern Inner Mongolia. Cropland area occupies approximately 30–40% of NEC, which has a typical temperate continental monsoon climate, with a cold, arid winter and a warm, rainy summer. The mean annual temperature range is 3–7 °C.
Figure 1
Figure 1. The location of NEC and the four study regions.
View largeDownload slide

The location of NEC and the four study regions.

Figure 1
Figure 1. The location of NEC and the four study regions.
View largeDownload slide

The location of NEC and the four study regions.

Close modal

To analyze the temporal dynamics of soil moisture for the main vegetation types in NEC, four regions were selected (Figure 1): Region A (grassland), Region B (cropland), Region C (forest) and Region D (cropland). Regions A, B and C were located at similar latitudes, and were distributed from west to east. They belonged to subarid, subhumid and humid climate zones, respectively, covering all the climate zones present in NEC; together, they accounted for the three major vegetation types and the three major climate zones in NEC. Region D was cropland and was located at a similar longitude to Region B but at lower latitude. The comparison of changes in SSM between Regions B and D was intended to demonstrate the difference in soil moisture dynamics for the two latitude zones. The proportion of each major land use type was about 70% or higher; the percentage of each land cover type for each region is listed in Table 1.

Table 1

The percentage of land cover type for each of the four regions

Land coverRegion A (%)Region B (%)Region C (%)Region D (%)
Forest 0.0 0.3 69.0 1.2 
Shrub 0.1 0.9 11.8 1.6 
Grass 82.2 7.0 14.5 5.7 
Wetland 11.2 5.0 1.4 2.0 
Crop 3.9 81.6 2.6 79.4 
Build-up 0.2 4.7 0.6 8.1 
Water 2.4 0.5 0.0 2.2 
Land coverRegion A (%)Region B (%)Region C (%)Region D (%)
Forest 0.0 0.3 69.0 1.2 
Shrub 0.1 0.9 11.8 1.6 
Grass 82.2 7.0 14.5 5.7 
Wetland 11.2 5.0 1.4 2.0 
Crop 3.9 81.6 2.6 79.4 
Build-up 0.2 4.7 0.6 8.1 
Water 2.4 0.5 0.0 2.2 
View Large

Materials

The essential climate variable-soil moisture product

The essential climate variable-soil moisture (ECV SM) product is the first purely multi-decadal satellite-based soil moisture product that spans over 32 years on a daily basis and at a spatial resolution of 0.25 °. This product synergistically combines the soil moisture data from four passive (scanning multi-channel microwave radiometer (SMMR), special sensor microwave/image (SSM/I), TRMM microwave imager (TMI) and AMSR-E) and two active (European Remote Sensing Active Microwave Instrument (ERS AMI) and advanced scatterometer (ASCAT)) coarse-resolution microwave sensors into a global dataset spanning 1978–2010 (Liu et al. 2012; Dorigo et al. 2015). The soil moisture was estimated using the Land Parameter Retrieval Model (Owe et al. 2008) for passive sensors, and the change detection algorithm (Wagner et al. 1999) for active sensors. The first version of ECV SM was released in June 2012 and covered the 32-year period from 1978 to 2010. For more details on the data harmonization procedure, readers are referred to Liu et al. (2011, 2012) and Wagner et al. (2012).

The ECV SM product has been validated using in situ soil moisture data at a depth of 0–5 cm. Dorigo et al. (2015) used ground-based soil moisture observations at 596 sites from 28 historical and active monitoring networks worldwide to assess random errors of the EVC SM product based on the triple collocation technique. They found that the average Spearman correlation coefficient between ECV SM and all in situ observations was 0.46, and that the unbiased root-mean-square difference was approximately 0.04 cm3/cm3. Similar validation results have also been achieved in the Tibetan Plateau (Zeng et al. 2015). Furthermore, Liu et al. (2012) evaluated the uncertainties of ECV SM over various land cover types (e.g. range land, cropland, wooded savanna, grassland) using in situ soil moisture data from the International Soil Moisture Network, and found that the average Spearman correlation coefficient between ECV SM and all in situ observations was greater than 0.6. These results indicate that the ECV SM can be used to analyze the dynamics of SSM both regionally and globally. Moreover, the ECV SM dataset overcomes the inconsistencies among various active and passive soil moisture products, and retains the dynamic change in soil moisture of the various data sources, including seasonal variation and interannual variability. Therefore, the ECV SM dataset can be used to analyze the temporal dynamic of SSM over a longer time span.

Due to the effect of low temperatures in NEC, the soil may be frozen from November to April. Therefore, the months with unfrozen soil (from May to October) were selected to analyze the dynamics of SSM in NEC. In addition, only ECV SM products for the period of 1979–2010 were used because the SSM dataset provided by the ECV SM product in 1978 started in November.

Meteorological data

NEC has 101 meteorological stations, and their observational dataset is shared by the China Meteorological Data Sharing Service System (CMDSSS; http://cdc.nmic.cn/home.do). The monthly rainfall and air temperature we used was downloaded from CMDSSS. For the yearly and seasonal dynamics of SSM, rainfall was summarized as total annual rainfall, and the air temperature was summarized as the annual average air temperature.

Methods

Temporal dynamic analysis

To examine the interannual dynamics of SSM in NEC, the spatially averaged SSM was computed from:
formula
1
where is the spatially averaged SSM, N is the number of SSM/I pixels and is the SSM at location n and time t. The average annual SSM was calculated from:
formula
2
where is the average annual SSM and T is the total number of days in a year that SSM was retrieved successfully.
Thus, the percent of soil moisture difference (PSMD) could be derived according to:
formula
3
where PSMD represents the relative change of SSM in NEC. A positive value of PSMD denoted an SSM level that was wetter than the average annual SSM, and a negative value denoted one that was drier.

Spatial variability and its time-variant and time-invariant contributors

The relationship between the spatial mean and the spatial variance of SSM is a commonly used tool for the investigation of spatial variability (Brocca et al. 2007; Famiglietti et al. 2008). In this study, we examined this relationship separately for the four regions, each of which was in a different climate class. The spatial mean SSM was determined using Equation (1). The corresponding spatial variance was calculated according to:
formula
4
To investigate factors influencing the spatial distribution of SSM, we followed the approach of Mittelbach & Seneviratne (2012), and decomposed the into its time-variant and time-invariant contributors, thereby determining the degree to which factors linked to soil moisture dynamics and to temporally stable features contributed to spatial variance. To achieve this, every observation was split into its temporal mean ⁠, which is given by:
formula
5
and its temporal anomaly, A(n,t), so that can be described by:
formula
6
Accordingly, can be expressed as:
formula
7
Substituting Equations (6) and (7) into Equation (4) results in:
formula
8
which can be rebuilt to:
formula
9
This simplifies to:
formula
10
where the variance of the temporal mean is the temporal invariant part of the equation, and the temporal variant part consists of the sum of the covariance between the temporal mean and anomalies and the variance of the anomalies ⁠. This allows us to consider the relative contributions of temporal variant and invariant components to the spatial variance, along with their temporal evolution in the four study areas.

The entire NEC area

The dynamics of spatial pattern

Figure 2 shows the monthly average SSM in July in NEC for four different years and gives a general overview of the spatial distribution of soil moisture. The spatial patterns appeared to be similar for all years. Lower values occurred in the grassland-dominated west of NEC, and higher values occurred in the east and north, where the main land cover is cropland and forest. This spatial distribution was consistent with the spatial arrangement of climate zones in NEC: the subarid, subhumid and humid zones are distributed from west to east (Zhang et al. 2010). Although the range of SSM values in Figures 2(a)–2(d) differed from each other, the values were low relative to the other three years; this was reflected in an overall mean SSM of 0.15 cm3/cm3 (Figure 2(a)). Figures 2(b)–2(d) had similar average SSM levels, approximately 0.22 cm3/cm3, and only showed a small decreasing trend with time. Overall, although SSM in NEC fluctuated, the spatial pattern of low values in the west and high values in the east and north did not change with time over the past 30 years.
Figure 2
Figure 2. Spatial patterns of the monthly SSM for four specific times: (a) July 1980, (b) July 1990, (c) July 2000 and (d) July 2010. A ‘Null’ value indicates no data, a pixel where SSM was not retrieved successfully or a pixel where the land cover type was water.
View largeDownload slide

Spatial patterns of the monthly SSM for four specific times: (a) July 1980, (b) July 1990, (c) July 2000 and (d) July 2010. A ‘Null’ value indicates no data, a pixel where SSM was not retrieved successfully or a pixel where the land cover type was water.

Figure 2
Figure 2. Spatial patterns of the monthly SSM for four specific times: (a) July 1980, (b) July 1990, (c) July 2000 and (d) July 2010. A ‘Null’ value indicates no data, a pixel where SSM was not retrieved successfully or a pixel where the land cover type was water.
View largeDownload slide

Spatial patterns of the monthly SSM for four specific times: (a) July 1980, (b) July 1990, (c) July 2000 and (d) July 2010. A ‘Null’ value indicates no data, a pixel where SSM was not retrieved successfully or a pixel where the land cover type was water.

Close modal

Temporal stability: annual dynamics

Throughout the study area, the averaged SSM computed from the ECV SM product was approximately 0.194 cm3/cm3 for the past 30 years. Soil became wetter from 1979 to 1988, and drier from 1989 to 2010. Overall, SSM displayed a decreasing trend from 1979 to 2010 (Figure 3). This result may be explained partly by the changes in rainfall and air temperature, because SSM is also influenced by groundwater levels and artificial irrigation. However, the effect on SSM is not discussed here, because we focused on the change of SSM and its relationship with air temperature and rainfall. Meteorological station data showed an increasing trend for rainfall from 1979 to 2010, along with a decreasing trend for air temperature over this same period. Therefore, a positive relationship exists between SSM and rainfall (R = 0.64) and a negative one exists between SSM and air temperature (R = –0.35), as shown in Table 2. This is consistent with previous results found from in situ SSM measurements taken from meteorological stations (Huang & Ding 2007; Zuo & Zhang 2008). This also indicates that remote sensing-based SSM products are similar to in situ SSM data from meteorological stations in their ability to represent the dynamics of soil moisture.
Table 2

The correlation coefficients (R) of SSM and air temperature (TA), and of SSM and annual rainfall (P). NE represents all of NEC; ‘ + ’ and ‘–’ denote positive and negative correlation, respectively

RRegion ARegion BRegion CRegion DNE
SSM & TA –0.19 –0.36* –0.49* –0.35* –0.35* 
SSM & P +0.59** +0.48** +0.16 +0.54* +0.63** 
RRegion ARegion BRegion CRegion DNE
SSM & TA –0.19 –0.36* –0.49* –0.35* –0.35* 
SSM & P +0.59** +0.48** +0.16 +0.54* +0.63** 

*Difference is significant at the 0.05 level.

**Difference is significant at the 0.01 level.

View Large
Figure 3
Figure 3. Temporal dynamics of (a) the spatially averaged annual SSM, (b) annual total rainfall (P) and (c) averaged air temperature (TA) for the entire study area.
View largeDownload slide

Temporal dynamics of (a) the spatially averaged annual SSM, (b) annual total rainfall (P) and (c) averaged air temperature (TA) for the entire study area.

Figure 3
Figure 3. Temporal dynamics of (a) the spatially averaged annual SSM, (b) annual total rainfall (P) and (c) averaged air temperature (TA) for the entire study area.
View largeDownload slide

Temporal dynamics of (a) the spatially averaged annual SSM, (b) annual total rainfall (P) and (c) averaged air temperature (TA) for the entire study area.

Close modal

Seasonal change of SSM

With the exception of interannual variability, the seasonal dynamics of SSM were also important for vegetation growth and agricultural production. Here, we defined spring as May, summer as June, July and August, and autumn as September and October, because SSM data from only May to October were used. Through analyzing the temporal dynamics of SSM, different change trends and rates were found for the three seasons: SSM showed a decreasing trend in spring and autumn, with a higher rate of change in autumn, and an increasing trend in summer with a rate of change between those of spring and autumn (Figure 4). In summary, SSM decreased approximately 30 and 3.8% in the autumn and spring, respectively, between 1979 and 2010, and increased 10% in the summer. When considering the whole year (spring, summer and autumn), SSM decreased overall, with a change rate of approximately 23% over the past 30 years.
Figure 4
Figure 4. The interannual dynamic of percentage of soil moisture difference (PSMD) for different seasons in NEC from 1979 to 2010 for (a) spring (May), (b) summer (June, July and August) and (c) autumn (September and October).
View largeDownload slide

The interannual dynamic of percentage of soil moisture difference (PSMD) for different seasons in NEC from 1979 to 2010 for (a) spring (May), (b) summer (June, July and August) and (c) autumn (September and October).

Figure 4
Figure 4. The interannual dynamic of percentage of soil moisture difference (PSMD) for different seasons in NEC from 1979 to 2010 for (a) spring (May), (b) summer (June, July and August) and (c) autumn (September and October).
View largeDownload slide

The interannual dynamic of percentage of soil moisture difference (PSMD) for different seasons in NEC from 1979 to 2010 for (a) spring (May), (b) summer (June, July and August) and (c) autumn (September and October).

Close modal

The selected regions

The temporal change

To quantitatively analyze SSM changes over time, the rate of change of SSM was defined by the slope of the line fit between PSMD and time. The annual SSM decreased slightly between 1979 and 2010 for all four regions (Figure 5). The averaged SSM levels were approximately 0.15, 0.24, 0.21 and 0.24 cm3/cm3 for Regions A, B, C and D, respectively. SSM levels decreased annually at rates of –0.31, –0.05, –0.58 and –0.34% for Regions A, B, C and D, respectively. This result indicates that SSM decreased approximately 9.8, 1.6, 18.7 and 10.7% overall for Regions A, B, C and Region D, respectively, over the past 30 years. Sorted by the rate of change of SSM, we had: crop < grass < forest.
Figure 5
Figure 5. Temporal dynamics of PSMD for the four regions from 1979 to 2010 for (a) Region A, (b) Region B, (c) Region C and (d) Region D.
View largeDownload slide

Temporal dynamics of PSMD for the four regions from 1979 to 2010 for (a) Region A, (b) Region B, (c) Region C and (d) Region D.

Figure 5
Figure 5. Temporal dynamics of PSMD for the four regions from 1979 to 2010 for (a) Region A, (b) Region B, (c) Region C and (d) Region D.
View largeDownload slide

Temporal dynamics of PSMD for the four regions from 1979 to 2010 for (a) Region A, (b) Region B, (c) Region C and (d) Region D.

Close modal
As mentioned above, the change of SSM relates to rainfall and air temperature. The temporal changes of both are given in Figures 6 and 7, respectively. All four regions showed an increasing trend for air temperature. The correlation coefficient between air temperature (TA) and SSM is given in Table 2 for each region. A negative correlation was found between TA and SSM. The highest correlations were found for Region C, followed by Region B and Region D, and low correlation was found for Region A. Conversely, a positive correlation was found between SSM and annual rainfall for all regions (Figure 7). Compared with the other three regions, Region C displayed the lowest R value. This may be explained by the influence of forest on rainfall: trees and groundcover reduce water infiltration into the soil, which reduces the correlation between soil moisture and rainfall. Annual rainfall in NEC and in all four regions fell over time, while air temperature rose. These variations caused the following results: rising air temperature increases surface evapotranspiration, and thus increases the soil moisture loss. Without including the change in groundwater or human factors, we predict that SSM will continue to decrease based on the conservation of energy and matter, and these changes will be consistent with the results derived from SSM remote sensing products.
Figure 6
Figure 6. Temporal dynamics of air temperature (TA) for all four regions (1979–2010). (a) Region A, (b) Region B, (c) Region C and (d) Region D.
View largeDownload slide

Temporal dynamics of air temperature (TA) for all four regions (1979–2010). (a) Region A, (b) Region B, (c) Region C and (d) Region D.

Figure 6
Figure 6. Temporal dynamics of air temperature (TA) for all four regions (1979–2010). (a) Region A, (b) Region B, (c) Region C and (d) Region D.
View largeDownload slide

Temporal dynamics of air temperature (TA) for all four regions (1979–2010). (a) Region A, (b) Region B, (c) Region C and (d) Region D.

Close modal
Figure 7
Figure 7. Temporal dynamics of annual rainfall (P) for each region (1979–2010). (a) Region A, (b) Region B, (c) Region C and (d) Region D.
View largeDownload slide

Temporal dynamics of annual rainfall (P) for each region (1979–2010). (a) Region A, (b) Region B, (c) Region C and (d) Region D.

Figure 7
Figure 7. Temporal dynamics of annual rainfall (P) for each region (1979–2010). (a) Region A, (b) Region B, (c) Region C and (d) Region D.
View largeDownload slide

Temporal dynamics of annual rainfall (P) for each region (1979–2010). (a) Region A, (b) Region B, (c) Region C and (d) Region D.

Close modal
Although falling trends in SSM were found for all regions (similar to the results for the whole of NEC), there were also seasonal differences in SSM changes at the regional level. Figure 8 shows the change rate of PSMD derived from yearly SSM data. An increase in the change rate of PSMD in summer was observed for Regions B, C and D. Over the past 30 years, the SSM for Regions B, C and D increased by approximately 15, 20 and 5%, respectively, while the SSM for Region A decreased by 12%. For spring and autumn, SSM levels fell for all four regions and the rate of decrease of PSMD in autumn was higher than that in spring. The decrease rate of SSM was 1–2 times greater in autumn than in spring for Regions A and C, and about 15 times greater over the same seasons for Regions B and D. This result shows that there is a clear drying trend in autumn, which correlates to cropland being harvested and surface soil being left to dry out. This increases the likelihood of dust-related weather in NEC in autumn.
Figure 8
Figure 8. The rate of change (or slope) of PSMD from 1979 to 2010 for the four regions over three seasons.
View largeDownload slide

The rate of change (or slope) of PSMD from 1979 to 2010 for the four regions over three seasons.

Figure 8
Figure 8. The rate of change (or slope) of PSMD from 1979 to 2010 for the four regions over three seasons.
View largeDownload slide

The rate of change (or slope) of PSMD from 1979 to 2010 for the four regions over three seasons.

Close modal

The highest change rate among the four regions was observed in Region C. Coincidentally, this result was consistent with the magnitude of air temperature increase: air temperature in NEC increased more quickly in the northeast than in the southwest (Northeast Regional Climate Change Assessment Report 2012). Figure 8 shows the rate of change of SSM per season against time for each region. Regions A, B and C were distributed from west to east, and their rate of change for each season also displayed a rising trend from west to east. Regions B and D were distributed from north to south, and the rate of change per season in Region B was larger than in Region D. These results show that the rate of change in NEC also increases from southwest to northeast. Thus, a change of SSM in NEC can be used to represent the change of air temperature.

The spatial dynamics

Figure 9 shows the relationship between spatial mean and spatial variance for all four regions over three seasons. This relationship was mainly controlled by physical processes such as precipitation and evapotranspiration over large scales (Li & Rodell 2013). In this study, this relationship displayed obvious differences among regions within a given climate zone. For Region A, the main land cover was grass and there was no obvious seasonal change in SSM; the scatter plot of spatial mean and spatial variance for the three seasons studied displayed a similar spatial distribution (Figure 9). Both Region B and Region D were cropland. Clearly, the spatial mean of SSM in spring was lower than in summer and in autumn, and the spatial variance was higher in summer and autumn due to wetter soil. This result was also applicable to Region A. However, Region C, which was mostly forest, displayed a high spatial mean of SSM in spring, which may be caused by snow melt water in the forest. In summary, the spatial variance of SSM was high for Region A compared with the other three regions, and this result may be explained by the strong heterogeneity of Region A. Two points with high spatial variance for Region D in Figure 9 may result from strong rainfall or irrigation.
Figure 9
Figure 9. The relationship between spatial mean and spatial variance of SSM for the four regions over three seasons.(a) Region A, (b) Region B, (c) Region C and (d) Region D.
View largeDownload slide

The relationship between spatial mean and spatial variance of SSM for the four regions over three seasons.(a) Region A, (b) Region B, (c) Region C and (d) Region D.

Figure 9
Figure 9. The relationship between spatial mean and spatial variance of SSM for the four regions over three seasons.(a) Region A, (b) Region B, (c) Region C and (d) Region D.
View largeDownload slide

The relationship between spatial mean and spatial variance of SSM for the four regions over three seasons.(a) Region A, (b) Region B, (c) Region C and (d) Region D.

Close modal

The contribution to spatial variance

The spatial variance can be split into two parts: a temporal invariant part, ⁠, and a temporal variant part, consisting of the sum of and ⁠. Figure 10 shows the temporal dynamics of the two parts. For all four regions, the seasonal change in both the invariant and the variant parts was observed. In Figure 10(c), the temporal invariant part became negative, with values as low as –10%. This was induced by the covariance part of the temporal variant component. Negative values indicated a change of relationship between the temporal mean and the anomaly (Mittelbach & Seneviratne 2012; Rötzer et al. 2015). Overall, the temporal variant part of was higher than the temporal invariant part for Regions B and D, accounting for 74 and 67% of ⁠, respectively. Conversely, the temporal variant part was lower than the temporal invariant part for Regions A and C, accounting for 44 and 34% of ⁠, respectively. Less change was likely to occur for the spatial pattern of cropland cover in Regions B and D. Assuming unchanged land cover, the spatial variance of SSM was influenced by natural factors such as terrain and rainfall. Terrain is also a variable that does not change quickly, and rainfall is of approximately uniform distribution throughout the selected regions. Therefore, a lower temporal invariant part was reasonable for cropland. As for Region A, a relatively high temporal invariant part may result from grassland cover, as grass is affected by irregular grazing and mowing. Similarly, the forest (Region C) is affected by forest fires and the illegal felling of trees, and, relative to other types of land cover, the SSM in forest has a small temporal change due to the shading effect of the trees.
Figure 10
Figure 10. Time series of percentages of single contributors to for (a) Region A, (b) Region B, (c) Region C and (d) Region D. In (c), data are missing for 2000–2004.
View largeDownload slide

Time series of percentages of single contributors to for (a) Region A, (b) Region B, (c) Region C and (d) Region D. In (c), data are missing for 2000–2004.

Figure 10
Figure 10. Time series of percentages of single contributors to for (a) Region A, (b) Region B, (c) Region C and (d) Region D. In (c), data are missing for 2000–2004.
View largeDownload slide

Time series of percentages of single contributors to for (a) Region A, (b) Region B, (c) Region C and (d) Region D. In (c), data are missing for 2000–2004.

Close modal

The temporal and spatial patterns and the dynamics of SSM in NEC were investigated over the past 30 years using a remote sensing soil moisture product, ECV SM. Our three main conclusions are as follows.

First, SSM in the west of NEC was lower than in the east. The grassland-dominated west of NEC has a semiarid climate, and the east, dominated by forest and cropland, has a humid climate. The spatial pattern of SSM was consistent with the climate zone. For the whole study area, SSM displayed a decreasing trend over time, and we found a positive correlation with the change in rainfall and a negative correlation with the change of air temperature, as indicated by the data from local meteorological stations. Although a decreasing trend of SSM was found for the whole study area over the past 30 years, a clear seasonal change was observed in summer when SSM tended to rise about 10%. A stronger decline of 30% was observed in autumn, compared with a decline of 3.8% in spring. Overall, yearly averaged SSM declined approximately 20%.

Second, four regions were selected to analyze the differences in temporal and spatial dynamics of SSM under different types of land cover and in different climate zones. Similarly to the whole of NEC, the yearly averaged SSM of these four regions displayed a decreasing trend in general, though an increasing trend was found in summer, except in Region A. Although a similar trend of SSM change was found for all four regions, the rates of change differed. Comparing the change rate of each season, the following relationships of the change rate occur: grass < crop < forest. However, for the change rate of yearly averaged SSM, the above relationship changes to: crop< grass < forest. The reason is that SSM in grassland has a decreasing trend in summer rather than an increasing trend. Moreover, the spatial variance of SSM in Region A was higher than that of the other three regions. It included a relatively high temporal invariant part and may have resulted from the change in vegetation cover affected by the irregular grazing and mowing. The spatial variance for Regions B and D was relatively small, and it was mainly controlled by the temporal variant part. This may be because these two regions were both cropland and their spatial patterns seldom changed. Region C had a small spatial variance.

Third, the decreasing rate of SSM in NEC was consistent with the rising rate of air temperature. The decreasing rate of SSM increased from southwest to northeast, and the rising rate of air temperature also increased in the same manner. This shows that SSM values derived from remote sensing data have the potential for use in regional climate change research.

In this paper, the ECV SM product was confirmed to be feasible for analyzing the temporal-spatial change characteristics of SSM at regional scale by evaluating its spatial consistency with climate zone. The ECV SM product revealed that SSM in NEC had a drier trend over the past 30 years, with a different change rate for grassland, cropland and forest. Especially for cropland, SSM had significant seasonal change characteristics, such as an increase in summer and a decrease in spring, and this would make the situation ‘spring drought, summer waterlogging’ more serious in NEC. To a certain extent, this change of SSM would change seeding time and affect the grain production.

This paper was supported by the National Natural Science Foundation of China (No. 41301369, 41501408), Major Program of the National Natural Science Foundation of China (91125001). Thanks to the Climate Change Initiative (CCI) of the European Space Agency (ESA) for providing the ECV SM product and the National Meteorological Information Center (NMIC) for providing the meteorological data in NEC.

The authors declare no conflict of interest.

Brocca
L.
Morbidelli
R.
Melone
F.
Moramarco
T.
2007
Soil moisture spatial variability in experimental areas of central Italy
.
J. Hydrol.
333
,
356
–
373
.
Google Scholar
Crossref
Search ADS
 
de Rosnay
P.
Drusch
M.
Boone
A.
Balsamo
G.
Decharme
B.
Harris
P.
Kerr
Y.
Pellarin
T.
Polcher
J.
Wigneron
J. P.
2009
AMMA Land surface model intercomparison experiment coupled to the community microwave emission model: ALMIP-MEM
.
J. Geophys. Res.
114
,
D05108
.
Google Scholar
Crossref
Search ADS
 
Dorigo
W. A.
Gruber
A.
De Jeu
R. A. M.
Wagner
W.
Stacke
T.
Loew
A.
Albergel
C.
Brocca
L.
Chung
D.
Parinussa
R. M.
Kidd
R.
2015
Evaluation of the ESA CCI soil moisture product using ground-based observations
.
Remote Sens. Environ.
162
,
380
–
395
.
Google Scholar
Crossref
Search ADS
 
Famiglietti
J. S.
Ryu
D.
Berg
A. A.
Rodell
M.
Jackson
T. J.
2008
Field observations of soil moisture variability across scales
.
Water Resour. Res.
44
,
W01423
.
Google Scholar
 
Huang
Y.
Ding
Y.
2007
Temporal and spatial characteristics of summer soil moisture vertical distribution in northeast China
.
Sci. Meteorol. Sinica
27
(
3
),
259
–
265
(
in Chinese
).
Google Scholar
 
Jung
M.
Reichstein
M.
Ciais
P.
Seneviratne
S. I.
Sheffield
J.
Cescatti
A.
Goulden
M. L.
Bonan
G.
Chen
J.
de Jeu
R.
Dolman
A. J.
Eugster
W.
Gerten
D.
Gianelle
D.
Gobron
N.
Heinke
J.
Kimball
J.
Law
B. E.
Montagnani
L.
Mu
Q.
Mueller
B.
Oleson
K.
Papale
D.
Richardson
A. D.
Roupsard
O.
Running
S.
Tomelleri
E.
Viovy
N.
Weber
U.
Williams
C.
Wood
E.
Zaehle
S.
Zhang
K.
2010
Recent decline in the global land evapotranspiration trend due to limited moisture supply
.
Nature
467
,
951
–
954
.
Google Scholar
Crossref
Search ADS
PubMed
 
Li
A.
Lu
Q.
Yang
X.
Zhang
J.
2013
AMSR-E soil moisture compared with ECMWF and NCEP soil moisture
.
Remote Sens. Technol. Appl.
28
(
4
),
666
–
673
(
in Chinese
).
Google Scholar
 
Liu
Y. Y.
Parinussa
R. M.
Dorigo
W. A.
de Jeu
R. A. M.
Wagner
W.
Van Dijk
A. I. J. M.
McCabe
M. F.
Evans
J. P.
2011
Developing an improved soil moisture dataset by blending passive and active microwave satellite-based retrievals
.
Hydrol. Earth Syst. Sci.
15
(
2
),
425
–
436
.
Google Scholar
Crossref
Search ADS
 
Liu
Y. Y.
Dorigo
W. A.
Parinussa
R. M.
De Jeu
R. A. M.
Wagner
W.
McCabe
M. F.
Evans
J. P. V.
van Dijk
A. I. J. M.
2012
Trend-preserving blending of passive and active microwave soil moisture retrievals
.
Remote Sens. Environ.
123
,
280
–
297
.
Google Scholar
Crossref
Search ADS
 
Owe
M.
de Jeu
R. A. M.
Holmes
T. R. H.
2008
Multisensor historical climatology of satellite-derived global land surface moisture
.
J. Geophys. Res. Earth Surf.
113
,
F01002
.
Google Scholar
Crossref
Search ADS
 
Rötzer
K.
Montzka
C.
Vereecken
H.
2015
Spatio-temporal variability of global soil moisture products
.
J. Hydrol.
522
,
187
–
202
.
Google Scholar
Crossref
Search ADS
 
The Writing Committee
2012
Northeast Regional Climate Change Assessment Report
.
China Meteorological Press
,
Beijing
.
Wagner
W.
Lemoine
G.
Rott
H.
1999
A method for estimating soil moisture from ERS scatterometer and soil data
.
Remote Sens. Environ.
70
,
191
–
207
.
Google Scholar
Crossref
Search ADS
 
Wagner
W.
Dorigo
W.
de Jeu
R.
Fernandez
D.
Benveniste
J.
Haas
E.
Ertl
M.
2012
Fusion of Active and Passive Microwave Observations to Create an Essential Climate Variable Data Record on Soil Moisture
.
XXII ISPRS Congress
,
Melbourne
,
Australia
, pp.
315
–
321
.
Google Scholar
 
Zhang
J.
Qian
W.
Ding
T.
2010
Characteristics and trends of rainfall events in northeast China from May to September during 1956–2008
.
Meteorol. Month.
36
(
8
),
1
–
7
(
in Chinese
).
Google Scholar
 
Zhang
Y.
Wang
X.
Pan
Y.
Wang
Z.
Hu
R.
2011
The dependence of surface albedo on soil moisture in an arid desert area
.
J. Desert Res.
31
(
5
),
1141
–
1148
(
in Chinese
).
Google Scholar
 
Zuo
Z.
Zhang
R.
2008
Temporal and spatial features of the soil moisture in boreal spring in eastern China
.
Sci. China D Earth Sci.
38
(
11
),
1428
–
1437
.
Google Scholar
 
© IWA Publishing 2016