Evaluation of GLDAS soil moisture product over Kermanshah province, Iran

Evaluation of temporal and spatial variation of soil moisture may lead to the understanding of eco-hydrological processes (Zuo et al. 2019). In watershed management practices, it is of great importance to estimate the soil moisture, especially when studying the soil moisture distribution under vegetation, micro-topography, and precipitation events impacts (Das & Paul 2015). The soil moisture can be estimated using remote sensing, physical modelling, and in situ measurements or using a synthetic approach inclusive of such methods (Zeng et al. 2016; Zhuang et al. 2020). Remote sensing satellites like SMAP (Soil Moisture Active–Passive) have enabled global-scale observations of soil moisture data (Yuan & Quiring 2017). The accuracy of satellite-based data differs both in terms of the data type and the geographical location (Fang et al. 2016).

Weather stations are usually non-uniformly distributed over the basins and that is a challenge for accurately studying the spatial variation of parameters (Amini et al. 2019a, 2019b). Therefore, researchers in recent decades have used remote sensing and network data to evaluate parameters like soil moisture (Das et al. 2008). Using the soil moisture data recorded in 200 stations, Albergel et al. (2012) evaluated the extracted data from satellites with ground-based data and models. The results indicated that the satellite-driven soil moisture data simulated by the European Centre for Medium-Range Weather Forecasts' global model (ECMWF) have acceptable accuracy.

One of the drawbacks of in situ observation data is that the recording of this data in many areas is done by ground stations with a long-distance and non-uniform distribution. Satellites are a commonly used device for monitoring soil moisture. However, their single satellite period is limited, which may not be sufficient for climate change research (Wagner et al. 2012; Zhuang et al. 2020). Remote sensing can detect large-scale coverage, but it is highly reliant on the algorithm inversion method (Alizamir et al. 2021). Moreover, it cannot measure moisture in deep soil (Zhang et al. 2005). Therefore, the research is challenging due to the limitations of current measurements and monitoring techniques. For many applications in the literature, the Global Land Data Assimilation System (GLDAS) soil moisture is considered the reference source of soil moisture. GLDAS data at several spatial scales have been utilized to authorize soil moisture satellite-derived besides in situ measurements of soil moisture. That is because it evaluates soil moisture using some models of land surface through model-derived data, satellite observations, and data assimilation in situ (Wagner et al. 2003; 2012; Li et al. 2015; Kim & Choi 2015).

Chen et al. (2013) evaluated the moisture of soil precision data found by the Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) as well as the simulated soil moisture through GLDAS versus the observed data for the centre region of Tibet. According to the results, no output of AMSR-E has a good enough estimation of soil moisture in cold seasons of the year. It was found that the data obtained by the GLDAS model regarding the simulated soil moisture underestimated the soil moisture at the surface (0–5 cm). However, for a layer of 20–40 cm from the ground surface, the amount of moisture was evaluated with appropriate accuracy. Velde et al. (2014) presented the soil moisture map by using a microwave sensor for the Tibetan Plateau and validated the results to estimate soil moisture using GLDAS data which pointed out the sensor's capability. By the usage of an assimilation algorithm, Xiao et al. (2016) completed the soil moisture data of the satellite spatially and temporally during a 1-year period. The results illustrated that the data from the assimilation algorithm can effectively restore the time series of soil moisture both spatially and temporally in agreement with the simulated data given by the GLDAS model. Yuan & Quiring (2017) evaluated the data of soil moisture simulated and Coupled Model Intercomparison models, CMIP5 for the United States using the station and satellite data. They showed that the CMIP5 models overestimated the usefulness of the near-surface soil moisture (0–10 cm) and the soil column moisture (0–100 cm) in the western and underestimated them in the eastern parts of the US. Using GLDAS data, Park et al. (2017) suggested a soil moisture downscaling model. A time series pattern was visible in the 1 km moisture in the soil downscaled similar to that of GLDAS and weather stations.

Drought is usually viewed as one of the costliest natural hazards over the globe, causing damaging impacts on society and eco-environment (Shen et al. 2022). Liu et al. (2019) compared the two datasets GLDAS surface soil moisture and multi-satellite retrieved data for global drought analysis. According to the results, both datasets exemplify analogous patterns of duration, frequency, and severity of the drought. Drought indices including the Standardized Precipitation Index (SPI) have been widely used in the analysis and monitoring of drought events (Um et al. 2018). A methodology using the SPI was proposed by Aksoy et al. (2021) to develop critical drought intensity-duration-frequency (IDF) curves. Sun et al. (2022) proposed a Modified Drought Severity Index (MDSI) with a local optimization method. They evaluated the drought monitoring performance of MDSI across China and detected high correlations between MDSI and soil moisture, SPI at a 3-month scale, and Actual Drought-Affected Areas (ADA).

GLDAS assessment is usually limited to observation-rich regions of the world. However, only a few assessment activities have been conducted in the Middle East (Bi et al. 2016), particularly in a semiarid area such as northwest Iran which encountered several environmental perils (Amini 2020). Knowledge of spatio-temporal variation of soil moisture particularly in the no-station areas is conducive to recognition and prediction of the hazards. To use these databases properly, a consideration of their accuracy and efficiency is required. The aim of this research was to investigate the possibility of using the GLDAS data in determining soil moisture over Kermanshah province, Iran. Also, an investigation of soil moisture was conducted and its relationship with drought was explored. For this purpose, measured data related to GLDAS data at different soil depths during the period 2000–2014 were used. Statistical indicators were used to evaluate GLDAS data, and the relationship between soil moisture and drought was obtained.

The GLDAS model was developed to generate optimal fields of land surface states. It does so by combining satellite-based and ground data products based on observation and employing data adaptation algorithms and modelling of land surface to keep track of the changes (Rodell et al. 2004). The model was cooperatively devised by the scientists of the National Centre for Environmental Prediction (NCEP), the National Oceanic and Atmospheric Administration (NOAA), the Goddard Space Flight Centre, and the National Aeronautics and Space Administration (NASA) to generate various products. It provides users with a collection of land surface water and energy fluxes including moisture data of the soil in four depths of 0–10, 10–40, 40–100, and 100–200 cm (Cai et al. 2017). To generate optimal fields of land surface states and fluxes, the GLDAS integrates ground-based data and satellite sets to parameterize, force, and constrain a set of offline land surface models. The Goddard Earth Sciences Data and Information Services Centre website hosts and distributes the GLDAS data. The GLDAS model generates and integrates several global and observation-based data at multi-resolutions. The overall GLDAS modelling strategy is offline land surface models, to integrate multiple, large amounts of observational data and be activated globally by the Land Information System. The resolution of GLDAS products is 3 h. In addition, through the average temporal of 3-h products, the monthly products are produced (Bi et al. 2016).

In Equation (6), P_i, ⁠, and stand for the ith year precipitation, the average precipitation, and the standard deviation, respectively.

At meteorological stations, soil moisture is recorded at various depths of 5, 10, 20, 30, 50, 70, 100, 150, and 200 cm and at weekly, daily, and hourly time scales. The moisture in the soil was measured using Time Domain Reflectometry (TDR) equipment. For conductivity of electrical and soil water content measuring, the TDR is a well-known and accurate method (Skierucha et al. 2012). In this research, the measured data corresponding to the GLDAS data was used. The data of moisture in the soil pertaining to the synoptic stations of the study area and the networked data related to the GLDAS model with a spatial resolution of 0.25° × 0.25° for depths 0–10, 10–40, 40–100, and 100–200 cm from the ground during the 2000–2014 period were received and processed. Given this resolution, every single point represents the distribution of moisture in the soil throughout the grid with a 25 × 25 km dimension. The analysis of the GLDAS model implied that 42 grids of the model cover the study area. The points which represent each grid of the GLDAS network are designated in Figure 1.

To evaluate the estimations from the GLDAS model versus soil moisture collected from stations, the former was processed for the nearest point to the Sararoud station during the period of 2006–2012 (84 months). The Sararoud station was with the longest and most complete observed data in the area of study. The observed and predicted moisture data in the soil are shown in Table 1. As regards monthly scale and among the cold months of the year, January enjoyed the most surface soil moisture and August experienced the least. The average soil moisture in the Sararoud station occurs at the surface layer of 14.56 kg/m².

Table 2 contains the results concerning data accuracy regarding the GLDAS model against the data of the Sararoud station under various statistical indicators.

The coefficient of determination and agreement index in Table 2 shows relative conformity between the values of the measured moisture of soil and the GLDAS data. The bias index shows an average moisture difference of 3.63 kg/m² between the observed and GLDAS data. According to the output of the statistical indicators presented in Table 2, the efficiency of the GLDAS model in the estimation of monthly soil moisture is judged ‘good’ to ‘normal’. These studies show that the model efficiency may vary in different areas. Almost the same labelling for the efficiency of the GLDAS model was due to Velde et al. (2014), Xiao et al. (2016), and Park et al. (2017). Based on the statistical analysis, the GLDAS data could be used as soil moisture representative for the study area.

The average values of soil moisture concerning 42 points of the GLDAS model during the statistical period in different depths are shown in Figure 5(a). As seen in Figure 5(a), moisture increases, along with the growth of depth, which agrees with the reported results by Cai et al. (2017). The average moisture of soil in the second, third, and fourth layers has been multiplied proportionately to the first layer as 3.2, 6, and 12.5 times, respectively. Incidentally, the average moisture of the fourth layer has increased 2.1 times relative to the third layer.

The average moisture in the study area for different depths on the annual scale during the 2000–2014 period is shown in Figure 5(b). Figure 5(b) gives the averages in depths 0–10, 10–40, 40–100, and 100–200 cm during the statistical period as 18.92, 61.27, 112.72, and 236.17 kg/m², respectively. The variations of annual average surface soil moisture are usually slight and are amplified in parallel with the growth of depth. On the annual scale, the values of moisture in the soil at depths 0–10 cm have not undergone palpable variations; however, from 2007 onward, a decreasing trend has commenced in the surface layer and has been lengthened to 2010. The least value of soil moisture in 2010 is reported to be 16.92 kg/m², and the utmost value obtained in 2007 was to be 20.16 kg/m². As for the depths 10–40 cm, a rising trend of moisture content in soil in the period 2000–2004 was observed, and a reverse decreasing drift occurred in the years 2004–2009 before a second almost increment trend from 2009 on. With regard to the depths 40–100 cm, one discerns an increasing trend up to the year 2004, and following that, a decreasing sense is observed towards 2009. The same trend holds for the depths 100–200 cm. In the lowest soil depths, the greatest moisture content was recorded in 2004 and the least was evidenced to happen in 2009. According to Amini et al. (2019a, 2019b), the most severe wet and drought in Kermanshah province corresponds, respectively, to the years 2004 and 2009, which is inversely proportional to the soil moisture levels at these periods.

Figure 5(c) shows the average soil moisture in different depths on the monthly scale during the statistical period.

Figure 5(c) shows that the maximum average value of moisture in the soil at the depths 0–10 and 10–40 cm obtained in February was, respectively, 29.38 and 84.18 kg/m². As to the depths 40–100 cm, the maximum value of 145 kg/m² was attained in March, and in depths 100–200 cm it was recorded as 249 kg/m² in April. Conversely, the minimum value of average moisture in soil recorded in Kermanshah province for the depth 0–10 cm was obtained in August as 10 kg/m², for 10–40 cm in September was 44.43 kg/m², as for the 40–100 cm depths the minimum was recorded to be 96.61 kg/m² in October, and at depths 100–200 cm it was announced as 227 kg/m² in November. Therefore, as the depth increases, the alterations of soil moisture in any layer relative to the other happen with a lag of one month, much in agreement with the results of Cai et al. (2017). At shallow depths, i.e., in 0–40 cm of the ground, the increasing and decreasing trends keep pace with each other so that in winter it escalates with high values, and then approaching the warm period the decrement trend carries on to the end of summer.

The monthly variations of soil moisture in depths 10–100 cm are great and in deep layers they are low owing to reduced evaporation in the lower layers. Akin to annual variations, the layers of 40–100 cm and 100–200 cm differ drastically in moisture values on the monthly scale.

Figure 6(a)–6(d) depicts the mapping of average soil moisture in the study area during 2000–2014 resulting from GLDAS data in four different depths. Figure 6(a) illustrates that in the layer of soil 0–10 cm, the utmost value of moisture in the soil is recorded for Kermanshah City, Ravansar, and Sarpol Zahab. The lowest value of 15 kg/m² was associated with Somar, the research site south-west. By and large, the southern parts of the province enjoy less soil moisture than the north. Unlike the soil surface layer, the maximum and minimum average moisture for the 10–40 cm layer was recorded, respectively, as 81 kg/m² in the western city of Sarpol Zahab and 51 kg/m² in Dalahoo and Kangavar, centrally positioned and the province's eastern region. The central areas at 10–40 cm depth, though not relishing the highest soil moisture, remain among those regions with the most moisture in the ranking of areas (Figure 6(b)). The greatest average soil moisture value of 157 kg/m² in 40–100 cm corresponds to the west, and the least was reported in the eastern portion of the area of study as reaching 86 kg/m² (Figure 6(c)). The soil moisture in this layer takes on similar conditions as in the 10–40 cm layer except that the areas of moisture loss in the 40–100 cm layer enjoy more spatial expansion, particularly in the eastern part of the study area. The supreme moisture in soil average for the depth of 100–200 cm was reported to happen in Kermanshah city, Ravansar, and Sarpol Zahab as 287 kg/m², and the minimum turned out to be 202 kg/m² in Kangavar (Figure 6(d)). The minimum soil moisture cores were observed, for the depths 10–40 cm and 40–100 cm, in Kangavar and Sonqor, in the eastern part of the province. Scrutinizing the drought in Kermanshah province, Amini et al. (2019a, 2019b) concluded that the maximum occurrence of rainfall is in the north, and the minimum is in the southern parts. In the present research, the maxima and minima of the surface layer moisture are reported to happen in these areas as well. It seems that the climatic conditions have the most influence on the surface layer of the soil which is in the acceptable range with the results of Cai et al. (2017).

Zoning maps were supplied for various months. For instance, Figure 6(e)–6(h) portrays the soil moisture zoning for January. The maximum and minimum zones of soil moisture in this month in depth 0–10 cm were 33.18 and 22.63 kg/m², respectively. As for depths 10–40 cm, they are obtained as 89.08 and 66.71 kg/m², respectively, and for 40–100 cm the extremum values were 166.11 and 83.11 kg/m². Finally, for 100–200 cm, the highest and lowest extents of soil moisture were 286.09 and 195.72 kg/m², respectively. According to the zoning maps for all months, it is inferred that spatially for depths 0–10 cm the northern parts experience the maximum moisture of soil and the southern regions, especially south western areas, have the minimum moisture in nearly all months of the year. The maximum moisture cores for depths 10–40 cm disperse across the province so that Kermanshah city and Ravansar in the centre and north, and Sarpol Zahab in the west have the peak soil moisture. At these depths, the least moisture value belongs to Kangavar and Dalahoo. With regard to the depths of 40–100 cm, it could be mentioned that the dispersion of minimum moisture cores happened in the eastern and central portions of the research area, while the highest moisture is observed in Sarpol Zahab in the west. As to depths 100–200 cm, the dispersal of maximum moisture cores is typically in Kermanshah city and Ravansar in the centre, and in Sarpol Zahab in the western part of the province. Also, the lowest amount of moisture of soil arises in the east, south, and south-west of the province. In this latter depth, due to taking distance from the ground surface followed by the evaporation reduction, the soil moisture enjoys the least changes during the months of the year with almost similar situations of minimum and maximum cores throughout the year. By investigating the mapping of drought severity in Kermanshah province during the period under study and in the month of January, Amini et al. (2019a, 2019b) took the result that in January the most extreme drought concerns the south and south-west, and the least is associated with the north. This is in compliance with the soil moisture state in these areas. The studies conducted on the subject of the climatic conditions of the study area in the various statistical period show a positive correlation with the results of the current research (Moradi et al. 2016; Rajabi 2016; Bazrafshan 2017; Emadodin et al. 2019).

Table 3 shows some statistical properties for annual soil moisture and SPI at the depths of 100–200 cm from the ground.

According to Figure 7, the minimum and maximum values of SPI are attained, respectively, in 2004 and 2009. This result shows that 2004 has been passed with the highest rainfall during the wet period and that the most severe drought has taken place in 2009. The greatest and least normalized soil moisture corresponded, respectively, to the very same years of 2004 and 2009. Hence, one deduces that the soil moisture values arising from GLDAS have a reverse relationship with the severity of drought and are in direct relation with SPI values, which agrees with the reported results by Sun et al. (2022). Moradi et al. (2016), Rajabi (2016), Bazrafshan (2017), and Emadodin et al. (2019) have shown, using the SPI values, that the area of Kermanshah has experienced extreme drought in the recent decade. This is in accordance with the annual soil moisture values given in the present research. Spennemann et al. (2015) showed that the moisture of soil values obtained from the GLDAS model is in great correlation with those of SPI, in agreement with our results in the current work. Resorting to the moisture of soil in the monitoring of drought is an alternative adopted by several authors (Mo 2008; Houborg et al. 2012; Quan et al. 2012; Pozzi et al. 2013; Sedri et al. 2019).

In this research, we studied the spatial and temporal variations of soil moisture in Kermanshah province the GLDAS model. Statistical indicators were employed to evaluate the accuracy of moisture data given by GLDAS. Using these data, the monthly and annual soil moisture changes in depths 0–10, 10–40, 40–100, and 100–200 cm from the ground were inspected. The results taken via this investigation are as follows:

• The values of statistical indicators guaranteed that GLDAS data can be used with an acceptable approximation to estimate moisture in areas with limited or no data.

• The drought index, SPI, and soil moisture value were proven to be directly interrelated and hence, soil moisture had a reverse relation with drought severity.

• From data for various depths, it was found that the more severe the drought, the greater the effect on soil moisture in the deeper layers.

• The methodology used in this study can be adapted to monitor drought, fill the missing data of temporal and spatial series of soil moisture and overcome the problem of lack of soil moisture data in areas without synoptic stations.

This research received no external funding.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.