Different conditioning factors play specific roles in estimating high-resolution surface soil moisture databases. Nine (9) conditioning factors were selected carefully for the estimation of surface soil moisture, namely land use land cover, soil texture, normalized difference vegetation index, land surface temperature, topographic wetness index, rainfall, elevation, slope, and distance from rivers were selected. The spatial distribution pattern of these parameters was mapped and statistical databases were built with their sub-classes (Fig. 2 and Table 1). The land use and land cover (LULC) map was prepared by spectral classification of the Landsat satellite image. The classification produces a LULC database, which presents a total of nine (9) major classes. They are dense Forest, low dense forest, shrub land, outcrop/barren land, mountain grassland, urban and built-up, inland water, river water, and agriculture. The low dense forest is the dominant class, which covers an area of 35.38% of the study area. The low dense forest is mostly dominated in the eastern, western, and some pockets of the southern region of the study area (Fig. 2a). The shrubland is the second largest land cover class (26%) dominated in the middle portion and some pockets of the eastern and northern parts of the study area. A total FR index of 7.44 was contributed by LULC for the FR model. The highest frequency ratio (FR) value (1.41) was calculated for shrubland (Table 1) based on the FR equation (Eq. 8), which indicates a higher correlation to the soil moisture content (Zhang et al. 2011; Kidron and Gutschick 2013). A soil texture map was produced based on the soil classification scheme by the United States Department of Agriculture (USDA). A total of eight (8) textural classes are found in the study area. They are silty clay, sandy loam, sandy clay loam, silty clay loam, silty loam, sandy clay, sand, and loamy sand (Fig. 2b). Two other classes peat and lake were included in the map as additional sub-classes. Sandy clay loam is the largest soil texture class, which covers an area of 31.24% of the study area. Sandy clay loam is found in the middle portion of the study area where the river is flowing and its floodplain area. The peat class (1.13%) is found in some pockets of the eastern part of the study site. The maximum total FR index of 9.33 was contributed by the soil texture parameter into the FR model compared to other parameters. The maximum frequency ratio (FR) value of 1.57 was calculated for the Peat soil class (Table 1), which describes a strong correlation with the surface soil moisture (Petrone et al. 2004; Takada et al. 2009).
Normalized difference vegetation index (NDVI) was calculated through the band ratio of the subtraction and addition of the near-infrared and red bands (Bhandari et al., 2012), which have a dynamic response to soil moisture variation (Ahmed et al. 2017). The output NDVI value is ranged from − 0.39 to 0.67 (Fig. 2c). Furthermore, the study area was categorised into five (5) different zones based on the calculated NDVI value, like (i) less than 0.1 (20.24%), (ii) 0.1–0.15 (13.19%), (iii) 1.15–0.30 (18.62%), (iv) 0.30–0.45 (31.83%), and (v) more than 0.45 (16.11%). FR value of 1.41 is calculated for the 2nd category (0.1–0.15) and a total FR index of 5.20 is contributed by the soil texture parameter into the FR model (Table 1). Generally, the land surface temperature (LST) has a negative relationship with surface soil moisture except in high-latitude regions (Ghahremanloo et al. 2019; Jiang et al. 2023). The modeled LST of this area is varied from 9⁰ to 47⁰ centigrade (C). The spatial database on calculated LST was grouped into five (5) classes, namely (i) less than 20⁰ C (3.03%), (ii) 20⁰ C − 25⁰ C (15.36%), (iii) 25⁰ C − 30⁰ C (61.44%), (iv) 30⁰ C − 35⁰ C (17.03%), and (v) more than 35⁰ C (3.14%). The highest temperature is observed in the township area in the eastern part and some pockets of middle and northwest parts of the study area, whereas the lowest temperature is found in the higher altitude sections in the western portion of the study area (Fig. 2d).
The topographic wetness index (TWI) quantifies the topography-based soil moisture variation (Raduła et al. 2018; Kopecký et al. 2021). The TWI has a positive relationship with surface soil moisture (Maduako et al. 2017). The calculated TWI ranged from 0.47 to 42.31 with an average value of 8.38 (Fig. 2e). The maximum range of TWI is spread over the middle part of the watershed area where the topographic slope is very gentle (Qin et al., 2011). The spatial database of TWI was further reclassified into five (5) categories, namely (i) less than 5.0 (4.43%), (ii) 5.0–7.5 (38.46%), (iii) 7.5–10.0 (38.10%), (iv) 10.0–12.5 (9.09%), and (v) more than 12,.5 (9.91%). LST and TWI contribute a total FR index of 4.76 and 4.87 to the FR model respectively (Table 1). Rainfall is the primary source of soil moisture (Li et al. 2016) in the study area. The relationship between rainfall and surface soil moisture is highly linear (Sehler et al. 2019). The mean annual rainfall in the study area ranged from 1350 mm to 3850 mm. The study area was classified into five (5) different classes based on rainfall intensity and the class statistics for each class was generated, specifically (i) less than 1500 mm (0.85%), (ii) 1500 mm to 2000 mm (14.55%), (iii) 72000 mm to 2500 mm (57.11%), (iv) 2500 mm to 3000 mm (10.22%), and (v) more than 3000 mm (17.27%) (Fig. 2f and Table 1).
Elevation and slope have greater impact on surface soil moisture (Moeslund et al. 2013). As water flows downhill under the influence of gravity, the higher elevation areas are characterized by lower soil moisture and lower elevation areas are dominated by higher moisture conditions (Qiu et al. 2001). The elevation of the study area is varied from 0 to 1789.41 m. Based on the altitude variation, the study area was categorized into five (5) different groups, namely (i) less than 200 m (63.84%), (ii) 200 m to 400 m (21.39%), (iii) 400 m to 600 m (6.54%), (iv) 600 m to 800 m (3.92%), and (v) more than 800 m (4.82%) (Fig. 2g). The time stability of soil moisture is lowest in the gentle slope area compared to the moderate and steep slopes (Cai et al., 2019). Based on the variation of the slope, the study area was divided into five (5) categories. They are (i) less than 2⁰ (53.59%), (ii) 2⁰ to 5⁰ (14.55%), (iii) 5⁰ to 10⁰ (7.87%), (iv) 10⁰ to 20⁰ (13.43%), and (v) more than 20⁰ (10.55%). A higher slope (More than 20⁰) is found in mountain areas in the northeast, southeast, and northwest parts of the study area (Fig. 2h). Both gentle slope (less than 2⁰) and lower altitude (less than 200 m) are located in the middle part of the basin area. Elevation and slope parameters shared a total FR index of 3.27 and 4.16 in the FR model respectively (Table 1). In general, soil moisture varies by distance from the river (Horvath, 2002). Soil situated near the river is characterised by higher moisture than soils located at a distance from the river (Kumar et al., 2016). Five different buffer areas were generated through proximity analysis from the river to determine the effect of river on the soil moisture, such as (i) less than 200 m (24.62%), (ii) 200 m to 400 m (14.04%), (iii) 400 m to 600 m (10.85%), (iv) 600 m to 800 m (8.40%), and (v) more than 800 m (42.09%) (Table 1).
Table 1
Conditioning factors used for estimation of soil moisture (SM) through FR model
Value | Class name or Description | Histogram | % of Histogram | Potential SM points | % of Potential SM points | Frequency ratio (FR) | Total FR index |
Land use and Land cover |
1 | Dense Forest | 47374 | 2.36 | 2 | 0.885 | 0.37 | 7.44 |
2 | Low dense forest | 710143 | 35.38 | 68 | 30.088 | 0.85 |
3 | Shrub land | 521832 | 26.00 | 83 | 36.726 | 1.41 |
4 | Outcrop/barren lands | 43242 | 2.15 | 2 | 0.885 | 0.41 |
5 | Mountain grassland | 421354 | 20.99 | 35 | 15.487 | 0.74 |
6 | Urban and built-up | 15443 | 0.77 | 2 | 0.885 | 1.15 |
7 | Inland water | 4765 | 0.24 | 0 | 0.000 | 0.00 |
8 | River water | 140449 | 7.00 | 19 | 8.407 | 1.20 |
9 | Agriculture | 102468 | 5.11 | 15 | 6.637 | 1.30 |
Soil Texture |
1 | Silty clay | 91671 | 4.57 | 2 | 0.88 | 0.19 | 9.33 |
2 | Sandy loam | 176212 | 8.78 | 19 | 8.41 | 0.96 |
3 | Sandy clay loam | 627064 | 31.24 | 67 | 29.65 | 0.95 |
4 | Silty clay loam | 406699 | 20.26 | 43 | 19.03 | 0.94 |
5 | Peat | 22672 | 1.13 | 4 | 1.77 | 1.57 |
6 | Silty loam | 22603 | 1.13 | 3 | 1.33 | 1.18 |
7 | Sandy clay | 14248 | 0.71 | 0 | 0.00 | 0.00 |
8 | Sand | 555087 | 27.66 | 78 | 34.51 | 1.25 |
9 | Lake | 6567 | 0.33 | 1 | 0.44 | 1.35 |
10 | Loamy sand | 84247 | 4.20 | 9 | 3.98 | 0.95 |
Normalized Differential Vegetation Index (NDVI) |
1 | Less than 0.1 | 406280 | 20.24 | 47 | 20.80 | 1.03 | 5.20 |
2 | 0.1–0.15 | 264815 | 13.19 | 42 | 18.58 | 1.41 |
3 | 1.15–0.30 | 373813 | 18.62 | 38 | 16.81 | 0.90 |
4 | 0.30–0.45 | 638820 | 31.83 | 63 | 27.88 | 0.88 |
5 | More than 0.45 | 323342 | 16.11 | 36 | 15.93 | 0.99 |
Land Surface Temperature (LST) degree C |
1 | Less than 20 | 60822 | 3.03 | 3 | 1.33 | 0.44 | 4.76 |
2 | 20–25 | 308266 | 15.36 | 20 | 8.85 | 0.58 |
3 | 25–30 | 1233168 | 61.44 | 149 | 65.93 | 1.07 |
4 | 30–35 | 341812 | 17.03 | 43 | 19.03 | 1.12 |
5 | More than 35 | 63002 | 3.14 | 11 | 4.87 | 1.55 |
Topographic Wetness Index (TWI) |
1 | Less than 5.0 | 88958 | 4.43 | 6 | 2.65 | 0.60 | 4.87 |
2 | 5.0–7.5 | 771982 | 38.46 | 76 | 33.63 | 0.87 |
3 | 7.5–10.0 | 764626 | 38.10 | 95 | 42.04 | 1.10 |
4 | 10.0–12.5 | 182526 | 9.09 | 26 | 11.50 | 1.27 |
5 | More than 12.5 | 198978 | 9.91 | 23 | 10.18 | 1.03 |
Rainfall in mm |
1 | Less than 1500 | 16990 | 0.85 | 2 | 0.88 | 1.05 | 4.65 |
2 | 1500–2000 | 292094 | 14.55 | 25 | 11.06 | 0.76 |
3 | 2000–2500 | 1146286 | 57.11 | 148 | 65.49 | 1.15 |
4 | 2500–3000 | 205126 | 10.22 | 22 | 9.73 | 0.95 |
5 | More than 3000 | 346574 | 17.27 | 29 | 12.83 | 0.74 |
Elevation in m |
1 | Less than 200 | 1281891 | 63.87 | 162 | 71.68 | 1.12 | 3.27 |
2 | 200–400 | 429324 | 21.39 | 50 | 22.12 | 1.03 |
3 | 400–600 | 131360 | 6.54 | 10 | 4.42 | 0.68 |
4 | 600–800 | 78617 | 3.92 | 3 | 1.33 | 0.34 |
5 | More than 800 | 85878 | 4.28 | 1 | 0.44 | 0.10 |
Slope in degree |
1 | Less than 2 | 1075600 | 53.59 | 144 | 63.72 | 1.19 | 4.16 |
2 | 2–5 | 292084 | 14.55 | 39 | 17.26 | 1.19 |
3 | 5–10 | 157971 | 7.87 | 11 | 4.87 | 0.62 |
4 | 10–20 | 269603 | 13.43 | 20 | 8.85 | 0.66 |
5 | More than 20 | 211812 | 10.55 | 12 | 5.31 | 0.50 |
Distance from river in m |
1 | Less than 200 | 494112 | 24.62 | 59 | 26.11 | 1.06 | 5.06 |
2 | 200–400 | 281756 | 14.04 | 40 | 17.70 | 1.26 |
3 | 400–600 | 217794 | 10.85 | 21 | 9.29 | 0.86 |
4 | 600–800 | 168666 | 8.40 | 18 | 7.96 | 0.95 |
5 | More than 800 | 844742 | 42.09 | 89 | 39.38 | 0.94 |
The rating was assigned to each sub-class of all the nine conditioning parameters based on the FR value presented in Table 1. The FR value is ranged from 0 to 1.57. FR value of more than 1 indicates a strong correlation to the soil moisture event, whereas less than 1 refers to a weak correlation (Sarkar and Mondal 2020). The total FR index was calculated for each parameter after aggregation of all FR values sponsored by individual sub-classes. Finally, the overlay analysis was performed after integrating all the parameters with their FR characteristics based on Eq. 9, and the high-resolution (30 m spatial resolution) soil moisture index database was developed. The soil moisture index value is ranged from 4.57 to 11.61 with an average index value of 9.01 (Fig. 3a). A higher soil moisture index value indicates wet soil and a lower index indicates dry soil. The resulting soil moisture index database was further reclassified into five (5) soil groups based on surface soil moisture index. They are (i) very low moisture (less than 6.0), (ii) low moisture (6.0 to 7.0), (iii) moderate moisture (7.0 to 8.0), (iv) high moisture (8.0 to 9.0), and (v) very high moisture (More than 9.0) (Fig. 2b). The result identifies that almost 56.89% of the land areas are classified as very high moisture, 26.10% as high moisture, 10.18% as moderate, 4.71% as low moisture, and 2.12% as very low moisture (Table 2). The high-moisture soil and very high moisture soil classes (relatively wet soil classes) are identified in the middle section of the study area, where the Markham River and the surrounding floodplain area are located. These areas are enriched with higher moisture content because of higher topographic wetness index, lower elevation, and flat slopes (Fig. 2). Shrubland, sandy clay loam, lower NDVI, moderate surface temperature, and proximity to the river are the other dominant characteristics that caused higher surface soil moisture in the lower part of the Markham Valley and surrounding areas.
Table 2
Spatial distribution of soil moisture zones in the study area
Class no. | Soil moisture class | FR index range | Histogram | % Area |
1 | Very low | Less than 6 | 42461 | 2.12 |
2 | Low | 6–7 | 94519 | 4.71 |
3 | Moderate | 7–8 | 204339 | 10.18 |
4 | High | 8–9 | 523870 | 26.10 |
5 | Very high | More than 9 | 1141881 | 56.89 |
The purpose of the study was to estimate surface high-resolution soil moisture in the final catchment of the Markham River basin through the FR statistical approach. Although the reference point datasets were generated from 9 km SMAP level-4 data through fishnet analysis, the FR model generates a high-resolution spatial surface soil moisture database at a spatial resolution of 30 meters. A statistical spatial interpolation process can predict the soil moisture at unknown locations using known soil moisture information (Srivastava et al. 2019), but the critical or location variation can’t be incorporated into the prediction. So, the FR model is an alternative statistical method was selected for this study (Snepvangers et al. 2003). The topographic slope, elevation, topographic wetness index (TWI), and land surface temperature (LST) were measured statistically using linear trend line analysis (R-squared) to determine how well they fit the regression model (Shaw et al. 2023). The coefficient of determination (R-squared) values for topographic slope, elevation, TWI, and LST were determined to be 0.669, 0.4593, 0.2823, and 0.2694, respectively. The linear regression analysis identifies topographic slope has the maximum impact on soil moisture determination, followed by topographic height, TWI, and LST. Accuracy and success rate are essential to validate model-based estimation of soil moisture (Delgoda et al. 2016; Chi et al. 2019). The high-resolution estimated soil moisture index database was validated through prediction accuracy and success rate. The success rate of the FR model was calculated using several successful input reference points (207) over the total number of input reference points (226) with known soil moisture conditions, which were used as training for the soil moisture estimation. The success rate was computed as 91.59% (Table 3). On the other hand, the prediction accuracy was calculated using 57 reference points, which were not used for the FR modeling. The prediction accuracy of the estimation was calculated as 93.98% (Table 3), which is very good evidence to validate the FR model in the estimation of high-resolution surface soil moisture. The choice of the FR model over the MCDA is the best option for estimating surface soil moisture because the FR model estimates with better efficiency compared to any GIS-based MCDA model (Khosravi 2016; Wang and Li 2017).
Table 3
Prediction accuracy and succession rate for the Estimation of soil moisture
Soil moisture class | FR index range | Validation [20% flood points] | Accuracy (high and very high class) | Prediction Accuracy %) | Training [80% flood points] | Success ((high and very high class) | Success Rate (%) |
Very low | Less than 6 | 0 | 53 | 93.98% | 0 | 207 | 91.59% |
Low | 6–7 | 1 | 2 |
Moderate | 7–8 | 3 | 17 |
High | 8–9 | 13 | 38 |
Very high | More than 9 | 40 | 169 |
Total | 57 | | 226 | |