TL;DR
This paper introduces SWCTI, a new remote sensing index for estimating surface soil moisture, particularly in vegetated areas. By combining a water-sensitive index (SWCI) with a modified land surface temperature, SWCTI offers improved accuracy over existing methods like VSWI, showing strong correlations with ground-truth measurements.
What Problem Does This Paper Address?
Accurately monitoring soil moisture is crucial for various applications, including hydrological modeling, climate studies, and agricultural management. While remote sensing offers a scalable solution, existing indices often struggle with limitations. The Vegetation Supply Water Index (VSWI), a popular choice, uses the Normalized Difference Vegetation Index (NDVI) and Land Surface Temperature (LST). However, NDVI has a significant lag in reflecting water stress, and LST's influence on VSWI diminishes as temperatures rise, reducing its effectiveness. This paper aims to overcome these limitations by developing a more sensitive and accurate index for surface soil moisture estimation.
Methods
The core of this research is the development of a new index called the Surface Water Content Temperature Index (SWCTI). SWCTI is formulated as:
$SWCTI = \frac{SWCI}{LST - C}$
Here:
- SWCI (Surface Water Content Index) is the numerator. It's derived from MODIS Bands 6 and 7 reflectances, which are sensitive to water absorption features. SWCI is positively correlated with soil moisture, as increased soil moisture leads to higher vegetation water content and thus a larger difference between these bands.
- LST (Land Surface Temperature) is the base of the denominator. LST is sensitive to water stress; higher temperatures often indicate drier conditions due to reduced transpiration.
- C is a constant or adjustment factor. This modification to LST (LST - C) is intended to enhance the influence of temperature on the index, especially when LST values are high, addressing a key limitation of VSWI.
The paper also details a method for optimizing the value of 'C' by analyzing the coefficient of determination (R²) between SWCTI and in-situ soil moisture data across different years and soil depths. This optimization process involves plotting R² against various 'C' values to find the peak correlation.
Datasets
- MODIS Data: Terra MODIS eight-day land surface reflectance (MOD09A1) and eight-day average land surface temperature (MOD11A2) products were used. These were processed for the period of July to September from 2010 to 2014.
- In Situ Soil Moisture Data: Data from the multiscale Soil Moisture and Temperature Monitoring Network on the central Tibetan Plateau (CTP-SMTMN) were utilized. This network provided soil moisture measurements at four depths (0-5, 10, 20, and 40 cm) from 57 stations located in Naqu City, Tibetan Plateau.
- In Situ Meteorological Data: Daily precipitation data from 2010 to 2014 were acquired from the Tibetan Plateau Data Center to analyze rainfall events.
Key Insights and Ideas
- LST's Understated Role: The paper highlights that while LST is sensitive to water stress, its mathematical relationship in VSWI limits its contribution. The proposed SWCTI aims to rectify this by modifying LST in the denominator, giving it a more pronounced effect.
- SWCI as a Superior Numerator: SWCI, derived from water absorption bands, is presented as a more direct and sensitive indicator of water content in both soil and vegetation compared to NDVI, which can exhibit lags and saturation issues.
- The Importance of the 'C' Factor: The optimization of the constant 'C' is critical. It's not just an arbitrary number but a factor that significantly impacts the index's correlation with actual soil moisture, suggesting a nuanced relationship between temperature and soil moisture that needs careful calibration.
- Network Density Matters (to a point): The study found that while a denser network of ground stations improves the optimization of 'C', there's a point of diminishing returns. Around 37 stations were found to be sufficient for achieving stable and acceptable 'C' values.
Key Results
- Highest Correlation: SWCTI demonstrated the highest correlation with in-situ soil moisture among the tested remote sensing indices. The best correlation achieved was R = 0.801 (R² = 0.6409) for 0-5 cm soil moisture in sandy loam soil during August.
- Performance Improvement: Compared to VSWI, SWCTI showed significant improvements in the coefficient of determination (R²), with increases of up to 197% for 0-5 cm and 163% for 0-10 cm soil moisture.
- Soil Texture Sensitivity: SWCTI generally performed better with sandy loam soil than silty loam soil.
- Rainfall Responsiveness: SWCTI showed a clear sensitivity to rainfall events, generally increasing after rain and decreasing during dry periods, mirroring the behavior of soil moisture.
- Spatial Mapping: The generated 1-km resolution soil moisture maps showed that southeastern Naqu City is generally wetter, consistent with rainfall patterns, and captured temporal changes in moisture due to dry and rainy periods.
Limitations and Future Work
- In-Situ Data Limitations: The study acknowledges that point-scale in-situ soil moisture data cannot perfectly represent the true soil moisture within a MODIS pixel, introducing uncertainties in the calibration process.
- LST Influence: The strong influence of LST on SWCTI means that factors affecting LST (like clouds and rainfall) indirectly impact SWCTI's accuracy. Poor LST inversion quality can lead to reduced SWCTI accuracy.
- 'C' Factor Variability: While an optimal 'C' was determined for the study area, it's recognized that this value might vary with seasonal differences and soil textures.
- Future Work: The authors suggest future work should focus on optimizing 'C' and validating SWCTI in different geographical regions and diverse soil textures globally.
Summary updated