The effect of soil salinity on the use of the universal triangle method to estimate saline soil moisture from Landsat data: application to the SMAPEx-2 and SMAPEx-3 campaigns

ABSTRACT

In this article, a method based on UTM called salinity-based soil moisture content (S_SMC) is developed. Since the soil moisture depends on the soil salinity (SS) in semi-arid regions, the S_SMC method employs the SS as an effective and augmented variable in conventional UTM to estimate SMC in these areas. In calibration step, initially, a linear regression model between the land surface temperature (LST), the normalized difference vegetation index (NDVI), and the SS is applied using in situ measurements to assess the influence of the SS in SMC estimation. Then, a non-linearity model is conducted through insertion of more terms in the linear equation and an optimal model of S_SMC is yielded. Moreover, the SS is obtained using a linear model from two selected salinity indices derived from Landsat images and in situ measurements. In estimation step, the LST, NDVI, and the SS are obtained using Landsat data. The S_SMC method is evaluated in the Soil Moisture Active Passive Experiment (SMAPEx)-2 and SMAPEx-3 campaigns in wet and dry conditions, respectively, over two scenes of Landsat images. The results demonstrated that the S_SMC method is appropriate in non-irrigated areas. In these areas, the S_SMC method improves R² (coefficient of determination) from 22% to 65% in SMAPEx-2 and from 24% to 50% in SMAPEx-3. Moreover, the results have shown that the SMC can be estimated at satellite level with a root mean square error of 0.06 and 0.02 (m³ m⁻³) in wet and dry condition, respectively. Therefore, the SS is a key parameter to adjust conventional UTM to improve the SMC estimation by the S_SMC method.

Acknowledgements

The authors are thankful to Prof. Jeffrey Walker, Group Director of the Soil Moisture Active Passive Experiment Experiments (SMAPEx) from the Australian Research Council for providing in situ data during the study. We also would like to thank the providers of the Landsat imagery held in the USGS archives and reprocessing data sets (landsat.usgs.gov) and thank Dr E. Ghambari from the University of Melbourne, Parkville, Vic, Australia, for the topography and atmospheric correction of the Landsat imagery data.

Disclosure statement

No potential conflict of interest was reported by the authors.