SMAP underestimates soil moisture in vegetation-disturbed areas primarily as a result of biased surface temperature data
Introduction
Soil moisture (SM) provides water resources for global land evapotranspiration and has strong coupling with regional precipitation (Koster et al., 2004). SM can be modulated by vegetation type and biomass (Zhang et al., 2001; Kim and Wang, 2007), and vegetation correction is a critical step in the remote sensing of SM. Since the 1970s, satellite microwave remote sensing has initiated the age of global SM products (Njoku and Kong, 1977; Choudhury et al., 1979; Mo et al., 1982). The brightness temperature (TB) acquired from passive microwave radiometers consists of signal contributions from the soil surface and the overlying vegetation. To retrieve SM, vegetation effects are generally characterized by the vegetation optical depth (VOD or τ) and single scattering albedo (ω) parameters in τ-ω models (Schwank et al., 2018). As inferred from these models, an overestimation of vegetation effects might cause SM overestimation.
Launched in 2015, the L-band Soil Moisture Active Passive (SMAP) mission has provided a state-of-the-art global SM dataset. This dataset has seen extensive applications in global/regional evapotranspiration estimations (Purdy et al., 2018; Walker et al., 2019a), especially drought assessment (Liu et al., 2017; Mishra et al., 2017; Eswar et al., 2018; Sadri et al., 2018). The current SMAP baseline V-pol single channel algorithm (SCA-V) corrects for vegetation effects based on the 2000–2010 Moderate-resolution Imaging Spectroradiometer (MODIS) normalized difference vegetation index (NDVI) climatology (O'Neill et al., 2018). Vegetation water content (VWC) is estimated with a quadratic function of NDVI plus a vegetation type-dependent term for stem water content. VWC is then converted to VOD using a vegetation structure-dependent b-factor (Wigneron et al., 2017). Climatology-based parameterization has known impacts on SM retrieval, especially over areas with significant seasonal vegetation changes, e.g., croplands in the US Corn Belt (Walker et al., 2019b). Large errors are found at rapid vegetation growing stages due to interannual differences in vegetation health and phenology (Zwieback et al., 2018; Colliander et al., 2019; Singh et al., 2019). Both positive and negative NDVI anomalies can occur due to the earlier and later vegetation growing stages relative to the phenology implied in the 2000–2010 mean NDVI climatology.
Most of the current studies evaluate SMAP SM in terms of the correlation coefficient (R), bias, root mean square error (RMSE) and unbiased RMSE (ubRMSE) (Chan et al., 2016) and further investigate the dependence of these metrics on land use land cover (LULC), climate zone and other factors of interest (Ma et al., 2017; Ma et al., 2019; Zhang et al., 2019). In addition to time-invariant errors, recent studies also assess time-dependent errors as a result of the mismatch between the actual and climatological mean NDVI (Dong et al., 2018; Zwieback et al., 2018). To our best knowledge, however, few studies have evaluated the effects of interannual (compared to seasonal) vegetation changes on SMAP SM retrieval. Since the beginning of this century, drought has been claimed to cause vegetation disturbances (Pan et al., 2018; Wigneron et al., 2020). The affected land has lower NDVI values than the 2000–2010 climatology, indicating an overestimation of vegetation effects in the current SMAP SCA-V and therefore an overestimation of SMAP SM.
The primary objective of this study is to investigate the dependence of SMAP–reference SM differences on interannual NDVI changes in global vegetation-disturbed areas and the potential causal factors of such a dependency. To this end, the reference products included the European Space Agency (ESA) Climate Change Initiative (CCI) SM, the Global Land Data Assimilation System (GLDAS) SM, the Soil Moisture and Ocean Salinity (SMOS) L3 (SMOS-L3) SM and the SMOS-IC SM datasets. These datasets are independent of, are less dependent on, or are relatively insensitive to vegetation index climatology, differing from the SMAP SM dataset, which corrects for vegetation effects depending exclusively on the MODIS NDVI climatology. The hypothesis is that SMAP–reference SM differences can reveal the effects of vegetation biases on SMAP SM and the total SMAP–reference SM differences might increase with vegetation biases. The main results of this study are expected to have implications for SMAP SCA-V updates and global drought monitoring using the SMAP SM dataset.
Section snippets
MODIS NDVI
High-quality NDVI can be derived from MODIS red and near-infrared bands observations. The MODIS 16-day vegetation indices dataset was produced with a constraint view angle-maximum value compositing algorithm (Didan et al., 2015). In this study, the 0.05° × 0.05° Terra MODIS NDVI data (MOD13C1, C006) from 2000 to 2018 were collected from the Level-1 and Atmosphere Archive & Distribution System (LAADS) Distributed Active Archive Center (DAAC). For each year from 2000 to 2018, the annual maximum
Global distribution of vegetation-disturbed areas
Compared to the 2000–2010 climatology (Fig. A1), the relative NDVI changes from 2015 to 2018 are shown in Fig. 1. Large NDVI decreases are observed in tropical forests in Africa and India, likely caused by the 2015–2016 El Niño event (Wigneron et al., 2020). Drought might be a major cause of global NDVI decreases (Fig. 2). A concurrent NDVI decrease and meteorological drought are observed in 2015 in southern Africa and in 2018 in the southwest United States, central Australia, and north of the
How does SMAP SM behave under vegetation changes?
The CCI and GLDAS datasets show the underestimation of SMAP SM over both NDVI-increasing and decreasing areas (Fig. 4). For NDVI-increasing areas, vegetation effects are insufficiently corrected, and it is not difficult to determine the SMAP SM underestimation from radiative transfer (RT) equations. The part of radiation that should originate from vegetation is imposed to soil, causing drier SM retrievals. An NDVI decrease indicates an overcorrection of vegetation effects. The part of radiation
Conclusions
This study quantified the effects of vegetation biases on SMAP SCA-V SM (2015–2018) over global vegetation-disturbed areas determined by negative NDVI anomaly values relative to the 2000–2010 NDVI climatology. Supported by the ESA CCI, GLDAS, SMOS-L3 and SMOS-IC SM datasets, an NDVI decrease (LST increase) of 10% (1 K) might correspond to an underestimation of SMAP SM of ~0.007 (~0.008) m3·m−3. The results conflict with radiative transfer calculations, primarily as a result of the
Credit Author Statement
Xingwang Fan conceived and designed the study, supervised by Yuanbo Liu. Xingwang Fan carried out all data processing with support from Guojing Gan and Guiping Wu. Xingwang Fan draft the manuscript.
Declaration of Competing Interest
The authors declared that they have no conflicts of interest to this work.
Acknowledgments
This study was financially supported by the National Natural Science Foundation of China under grant No. 41701414 and the State Key Program of the National Natural Science Foundation of China under grant No. 41430855. We appreciate the supporting teams and providers of the ESA CCI, GLDAS, SMOS-L3, SMOS-IC and SMAP soil moisture datasets. The authors would like to thank the three reviewers and the associated editor Wigneron, J.-P. for their constructive comments.
References (51)
- et al.
Comparison of high-resolution airborne soil moisture retrievals to SMAP soil moisture during the SMAP validation experiment 2016 (SMAPVEX16)
Remote Sens. Environ.
(2019) - et al.
ESA CCI soil moisture for improved earth system understanding: state-of-the art and future directions
Remote Sens. Environ.
(2017) - et al.
Validation of Collection 6 MODIS land surface temperature product using in situ measurements
Remote Sens. Environ.
(2019) - et al.
Microwave retrievals of soil moisture and vegetation optical depth with improved resolution using a combined constrained inversion algorithm: application for SMAP satellite
Remote Sens. Environ.
(2020) - et al.
Drylands face potential threat under 2 °C global warming target. Nat
Clim. Change
(2017) - et al.
Intercomparison of land-surface parameterization schemes: sensitivity of surface energy and water fluxes to model parameters
J. Hydrol.
(2003) - et al.
Trend-preserving blending of passive and active microwave soil moisture retrievals
Remote Sens. Environ.
(2012) - et al.
Performance of SMAP, AMSR-E and LAI for weekly agricultural drought forecasting over continental United States
J. Hydrol.
(2017) - et al.
Inter-comparison of satellite-retrieved and global land data assimilation system-simulated soil moisture datasets for global drought analysis
Remote Sens. Environ.
(2019) - et al.
Satellite surface soil moisture from SMAP, SMOS, AMSR2 and ESA CCI: a comprehensive assessment using global ground-based observations
Remote Sens. Environ.
(2019)
Drought monitoring with soil moisture active passive (SMAP) measurements
J. Hydrol.
Increasing global vegetation browning hidden in overall vegetation greening: insights from time-varying trends
Remote Sens. Environ.
Global-scale surface roughness effects at L-band as estimated from SMOS observations
Remote Sens. Environ.
SMAP soil moisture improves global evapotranspiration
Remote Sens. Environ.
New refinements and validation of the collection-6 MODIS land-surface temperature/emissivity product
Remote Sens. Environ.
Modelling the passive microwave signature from land surfaces: a review of recent results and application to the L-band SMOS & SMAP soil moisture retrieval algorithms
Remote Sens. Environ.
A comprehensive validation of the SMAP enhanced Level-3 soil moisture product using ground measurements over varied climates and landscapes
Remote Sens. Environ.
The global SMOS Level 3 daily soil moisture and brightness temperature maps
Earth Syst. Sci. Data
NASA/GSFC/HSL, GLDAS Noah Land Surface Model L4 3 hourly 0.25 x 0.25 degree V2.1
Assessment of the SMAP passive soil moisture product
IEEE T. Geosci. Remote
Global land precipitation: a 50-yr monthly analysis based on gauge observations
J. Hydrometeorol.
Effect of surface roughness on the microwave emission from soils
J. Geophys. Res.-Oceans
MODIS Vegetation Index user’s Guide (MOD13 Series)
The error structure of the SMAP single and dual channel soil moisture retrievals
Geophys. Res. Lett.
SMAP soil moisture change as an indicator of drought conditions
Remote Sens.
Cited by (25)
Integration of microwave satellite soil moisture products in the contextual surface temperature-vegetation index models for spatially continuous evapotranspiration estimation
2023, ISPRS Journal of Photogrammetry and Remote SensingContinued decline of global soil moisture content, with obvious soil stratification and regional difference
2023, Science of the Total EnvironmentImproved estimation of vegetation water content and its impact on L-band soil moisture retrieval over cropland
2023, Journal of HydrologyCitation Excerpt :Finally, 10-day VOD climatology is produced from MODIS NDVI climatology composites based on the relationship established in the previous two steps. Since this (static) NDVI climatology cannot capture reported phenological shifts associated with global warming or the impact of inter-annual vegetation variations due to drought and fluvial conditions (Colliander et al., 2019; Fan et al., 2020; Gao et al., 2020a; Shen et al., 2022), the direct use of NDVI-derived VOD climatology in SM retrieval algorithms can lead to reduced SM accuracy (Judge et al., 2021). This raises two questions: (1) can dynamic VIs be used to obtain VWC and VOD estimates that significantly improve SM retrievals relative to the current SMAP approach of utilizing an NDVI climatology to estimate VWC and VOD?
Error and uncertainty characterization of soil moisture and VOD retrievals obtained from L-band SMAP radiometer
2022, Remote Sensing of EnvironmentCitation Excerpt :This signifies that the τ-ω model could be adequate to represent the multiple scattering effects to accurately retrieve SM using the land cover-specific model parameter ranges (Table 2). The remaining uncertainty in the SM retrievals could be attributed to other factors such as surface and canopy temperature (TSor TC) (Fan et al., 2020; Ma et al., 2021), dielectric models (Konings et al., 2011), roughness models (Konings et al., 2011; Peng et al., 2017), land cover heterogeneity (Ma et al., 2019), inversion techniques (Chaubell et al., 2020; Gao et al., 2020b, 2020c; Karthikeyan et al., 2020), and internal noise in the brightness temperatures. In the context of VOD, the ω parameter primarily drives the overall and site-specific uncertainty contributions at all reference sites.