Unbiased Root Mean Square Error Research Articles

Evaluation of a regional crop model implementation for sub-national yield assessments in Kenya

CONTEXTCropping system models can be used to both assess regional food security and to monitor and predict agricultural drought. Agriculture in Kenya is extremely important to both the economy and food security of the country. OBJECTIVEThis study evaluated a regional implementation of a widely used crop model, the Decision Support System for Agrotechnology Transfer (DSSAT), within a coupled modeling framework, the Regional Hydrologic Extremes Assessment System (RHEAS), over Kenya. The goal of this study was to assess the ability of RHEAS to simulate the annual variability of maize yields at the county level and evaluate the uncertainty inherent in the model and inputs. METHODSThe RHEAS system implements a stochastic ensemble approach to account for field scale variabilities in crop management practices and underlying soil and weather conditions. Satellite-derived datasets were used to evaluate the land surface component of the system and seasonally disaggregated yield for 5 years was used to assess the performance of the cropping system model. RESULTS AND CONCLUSIONSThe median correlation between RHEAS and satellite-derived soil moisture and evapotranspiration estimates were 0.78, and 0.51, respectively, indicating that the model is able to capture the key drivers of the hydrological budget. Overall, RHEAS simulated yearly yield variations with a median correlation of 0.7 with reported yields, with the best performance in the short rains season. However, across both seasons, the RHEAS model was positively biased on the order of ∼1.6 MT/ha. The overall median unbiased RMSE was 0.66 MT/ha. The RHEAS system shows skill at simulating extreme departures in anomalies, and a majority of the time (62.5%) the reported yields fall within the interquartile range of the simulations. SIGNIFICANCEOne of the most important areas of improvement for the next generation of agricultural data and models is to better understand and communicate the inherent uncertainties. This is especially critical in data-limited regions. Here we present a modeling system and its implementation that begins to address these concerns. We demonstrate the ability to simulate broad trends in yields at the county level for sub-annual yields with skills that commensurate previous national/annual level studies.

Using Robust Regression to Retrieve Soil Moisture from CyGNSS Data

Accurate global soil moisture (SM) data are crucial for modeling land surface hydrological cycles and monitoring climate change. Spaceborne global navigation satellite system reflectometry (GNSS-R) has attracted extensive attention due to its unique advantages, such as faster revisit time, lower payload costs, and all-weather operation. GNSS signal reflected at L-band also has significant advantages for SM estimation. Usually, SM is estimated based on the sensitivity of GNSS-R reflectivity to SM, but the noise in observations can significantly impact SM estimation results. A new SM retrieval method based on robust regression is proposed to address this issue in this work, and the effects of roughness and vegetation on the effective reflectivity of the Cyclone Global Navigation Satellite System (CyGNSS) are reconsidered. Ancillary data are provided by the SM Active Passive (SMAP) mission. The retrieved results from the training sets and test sets agree well with the referenced SMAP SM data. The correlation coefficient R is 0.93, the root mean square error (RMSE) is 0.058 cm3cm−3, the unbiased RMSE (ubRMSE) is 0.042 cm3cm−3, and the mean absolute error (MAE) is 0.040 cm3cm−3 in the training sets. For the test, the correlation coefficient is 0.91, the RMSE is 0.067 cm3cm−3, the ubRMSE is 0.051 cm3cm−3, and the MAE is 0.044 cm3cm−3. The proposed method has been evaluated using in situ measurements from the SMAP/in situ core validation site; in situ measurements and retrieval results exhibit good consistency with the ubRMSE value below 0.35 cm3cm−3. Moreover, the SM retrieval results using robust regression methods show better performance than CyGNSS official SM products that use linear regression. In addition, the land cover types significantly affect the accuracy of SM retrieval, and the incoherent scattering in densely vegetated areas (tropical forests) usually leads to more errors.

Generation of global 1 km daily soil moisture product from 2000 to 2020 using ensemble learning

Abstract. Motivated by the lack of long-term global soil moisture products with both high spatial and temporal resolutions, a global 1 km daily spatiotemporally continuous soil moisture product (GLASS SM) was generated from 2000 to 2020 using an ensemble learning model (eXtreme Gradient Boosting – XGBoost). The model was developed by integrating multiple datasets, including albedo, land surface temperature, and leaf area index products from the Global Land Surface Satellite (GLASS) product suite, as well as the European reanalysis (ERA5-Land) soil moisture product, in situ soil moisture dataset from the International Soil Moisture Network (ISMN), and auxiliary datasets (Multi-Error-Removed Improved-Terrain (MERIT) DEM and Global gridded soil information (SoilGrids)). Given the relatively large-scale differences between point-scale in situ measurements and other datasets, the triple collocation (TC) method was adopted to select the representative soil moisture stations and their measurements for creating the training samples. To fully evaluate the model performance, three validation strategies were explored: random, site independent, and year independent. Results showed that although the XGBoost model achieved the highest accuracy on the random test samples, it was clearly a result of model overfitting. Meanwhile, training the model with representative stations selected by the TC method could considerably improve its performance for site- or year-independent test samples. The overall validation accuracy of the model trained using representative stations on the site-independent test samples, which was least likely to be overfitted, was a correlation coefficient (R) of 0.715 and root mean square error (RMSE) of 0.079 m3 m−3. Moreover, compared to the model developed without station filtering, the validation accuracies of the model trained with representative stations improved significantly for most stations, with the median R and unbiased RMSE (ubRMSE) of the model for each station increasing from 0.64 to 0.74 and decreasing from 0.055 to 0.052 m3 m−3, respectively. Further validation of the GLASS SM product across four independent soil moisture networks revealed its ability to capture the temporal dynamics of measured soil moisture (R=0.69–0.89; ubRMSE = 0.033–0.048 m3 m−3). Lastly, the intercomparison between the GLASS SM product and two global microwave soil moisture datasets – the 1 km Soil Moisture Active Passive/Sentinel-1 L2 Radiometer/Radar soil moisture product and the European Space Agency Climate Change Initiative combined soil moisture product at 0.25∘ – indicated that the derived product maintained a more complete spatial coverage and exhibited high spatiotemporal consistency with those two soil moisture products. The annual average GLASS SM dataset from 2000 to 2020 can be freely downloaded from https://doi.org/10.5281/zenodo.7172664 (Zhang et al., 2022a), and the complete product at daily scale is available at http://glass.umd.edu/soil_moisture/ (last access: 12 May 2023).

Evaluation of GLDAS soil moisture seasonality in arid climates

ABSTRACT We evaluated the Global Land Data Assimilation System surface soil moisture product (GLDAS v. 2.1) against in situ soil moisture networks in arid climates in Australia and the United States, using common statistical metrics and seasonality metrics. Our results showed that GLDAS performed well (root mean square error (RMSE) = 0.100 m3/m3; unbiased RMSE (ubRMSE) = 0.060 m3/m3; correlation coefficient (R) = 0.555 on average) but systematically overestimated the soil moisture values (Bias = 0.067 m3/m3). The performance was better in Australian Oznet and the U.S. Climate Reference Network (USCRN), compared to the US Soil Climate Analysis Network (SCAN) network. In terms of seasonality, GLDAS soil moisture seasons were biased to start earlier; on average, drying and wetting transitions started 28 and 16 days earlier than in situ data, respectively. The end dates of GLDAS seasonal transitions showed good agreement with in situ data; the errors in transition timings were limited to within a week. This tendency is stronger in the US networks compared to the Australian network.

Assessing and comparing the subseasonal variations of summer soil moisture of satellite products over eastern China

AbstractUsing the 43‐station soil moisture (SM) at a depth of 0–10 cm over eastern China, the near‐surface (about 0–5 cm) SM products from satellite missions including the level‐2 product of Chinese FengYun 3C (FY3C), level‐2 neural network product of the European Soil Moisture and Ocean Salinity (SMOS) and level‐4 product of the US Soil Moisture Active Passive (SMAP) were evaluated and compared during summers of 2015–2018. Thus, features of those products can be identified for their further application of climate study over eastern China. Large diversity is found among the satellite and field SM in terms of spatial distribution. This disagreement may cause the poor performance of the relationship between spatial coefficients of variation (CV) and mean SM. Compared with the field SM, the FY3C SM has smaller bias than the other satellite products. FY3C also performs better than SMOS in terms of the root‐mean‐square error (RMSE) and unbiased RMSE (ubRMSE) but has the smallest correlation coefficient (R). The SMAP product is generally the best among the three products in terms of RMSE, ubRMSE and R. However, a good performance in those metrics does not guarantee the same results on various time scales. On subseasonal time scales, the R in FY3C is the smallest among the three products, and the SMAP product has the largest R, but its amplitudes of the subseasonal variations are much smaller than the field observations. This indicates that when the SMAP products are applied for the analysis on subseasonal SM–atmosphere interactions, the effects of SM may be underestimated. On 10–30 days and above 60 days, dry period tends to have large spatial CV but this phenomenon is weak in all the satellite products. On the other hand, dry area tends to have large temporal CV on the four time scales, and this relationship is the strongest in SMAP but the weakest in FY3C. Therefore, there are large uncertainties of variability among different satellite SM products on subseasonal time scales over eastern China. Besides seasonal and overall performance, more attention is called to the variations on different time scales.

Time series soil moisture retrieval from SAR data: Multi-temporal constraints and a global validation

Time series algorithms for soil moisture retrieval from synthetic aperture radar (SAR) data have steadily increased in popularity over the past decade due to the feasibility of decoupling the effect of other surface variables, and the increasing availability of dense time series SAR data. While soil moisture inversion from time series data can utilize more independent observations, the value of further constraints on the inversion process are widely acknowledged. However, how to constrain a time series retrieval for global soil moisture mapping is still unresolved. In this study, three kinds of time series constraints were further developed and evaluated, including the use of 1) temporal behavior of soil moisture and soil moisture bounds; 2) temporal behavior of vegetation or time-invariant vegetation; and 3) time series ensemble skill. The effect of these constraints was investigated using 4 years (2016–2019) C-band Sentinel-1 data collected over 547 worldwide stations from 17 networks available on the international soil moisture network (ISMN) and intensive ground samples collected during the Fifth Soil Moisture Active and Passive Experiment (SMAPEx-5). While the effect of these temporal retrieval skills varies in time and space, the global validation yielded four general suggestions: 1) the assumption of time-invariant vegetation contributed negatively even for a short retrieval period of ≤12 days; 2) reliable soil moisture bounds of each retrieval period can substantially improve the retrieval statistics at the cost of an underestimated soil moisture range; 3) the temporal constraints of soil moisture and vegetation need to be used together with the soil moisture bounds for reliable estimation; 4) the use of an ensemble retrieval could partly remove the retrieval uncertainties at the expense of underestimating soil moisture variation. The use of these constraints resulted in a competitive correlation coefficient (R: 0.64), root mean square error (RMSE: 0.072 m3/m3) and unbiased RMSE (ubRMSE: 0.052 m3/m3) at a spatial grid of 100 m, with similar performance achieved across a retrieval window up to 132 days.

Soil Moisture Retrieval From Sentinel-1 and Sentinel-2 Data Using Ensemble Learning Over Vegetated Fields

Soil moisture is a valuable basic data in climate, hydrological models, and agricultural applications. The rapid development of remote sensing technology can be used to monitor changes in soil moisture at multiple spatial and temporal scales. In this article, we unfolded a soil moisture retrieval method using ensemble learning combined with the Water Cloud Model(WCM) by Sentinel-1 and Sentinel-2 with multi-source datasets. First, using the WCM, the influence of vegetation cover on the backscattering coefficient was removed, where we use three vegetation index (Enhanced Vegetation Index (EVI), Normalized Difference Vegetation Index (NDVI) and Normalized Difference Water Index (NDWI)) for analysis and comparison. Then, combined with other multi-source datasets, a soil moisture retrieval model was established based on the ensemble learning algorithm. Here we choose two familiar ensemble learning algorithms for analysis and comparison, using pearson correlation significance analysis, which are the Random Forest (RF) and the Adaptive Boosting (AdaBoost). The results revealed that the RF model performed is slightly superior to the AdaBoost model. The optimal performance mean absolute error (MAE), root mean square error (RMSE)and the unbiased RMSE (ubRMSE)of RF model are 2.289 vol%, 2.934 vol%, 2.934 vol% respectively, which are slightly better than AdaBoost model. EVI is suitable for WCM model to remove vegetation scattering effect. It shows that it is attainable to utilize the ensemble learning method to inversion of SM using radar data. The proposed framework maximizes the potential of WCM, RF model, and multi-source datasets in deriving spatiotemporally continuous SM estimates, which should be valuable for SM inversion development.

Gap Filling of the ESA CCI Soil Moisture Data Using a Spatiotemporal Attention-Based Residual Deep Network

As an essential climate variable, soil moisture (SM) exerts an indispensable influence on numerous disciplines. However, various degrees of data gaps exist in current microwave SM products. Therefore, this article proposed a spatiotemporal attention-based residual deep network (STARN) to reconstruct gaps of the daily SM data from the Climate Change Initiative program of the European Space Agency (ESA CCI) over the Qinghai–Tibet Plateau (QTP) during unfrozen seasons (May to September) from 2001 to 2021. The developed model is an end-to-end residual network embedded with three attention modules to comprehensively consider the potential relationship between SM and surface variables. Evaluation results revealed that the proposed model could well reconstruct SM gaps with an overall median <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i> and unbiased RMSE (ubRMSE) values of 0.52 and 0.054 m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> /m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> , while the overall median <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i> and ubRMSE values for the ESA CCI SM were 0.41 and 0.058 m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> /m <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> . Besides, comparison with five baseline methods (e.g., the artificial neural network, convolutional neural network, extreme gradient boosting, long-short term memory, and DCT-PLS model) indicated that the STARN model had certain advantages over the five baseline models with higher correlation and more reasonable distribution patterns. The <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">R</i> /ubRMSE values for the five models were 0.38/0.057, 0.34/0.058, 0.40/0.058, 0.41/0.056, and 0.41/0.058, respectively. The pretraining using the ERA5-Land SM data could further improve the accuracy of generated seamless SM data since the ERA5-Land and ESA CCI SM complemented each other to a certain extent on the QTP. In summary, by leveraging the spatiotemporal information and attention modules, the STARN model showed great potentials in SM gap filling.

Evaluation of Several Satellite-Based Soil Moisture Products in the Continental US.

Satellite-based soil moisture products are suitable for large-scale regional monitoring due to the accessibility. Five soil moisture products including SMAP, ESA CCI, and AMSR2 (ascending, descending, and average) were selected in the continental United States (US) from 2016 to 2021. To evaluate the performance of the products and assess their applicability, ISMN (International Soil Moisture Network) data were used as the in situ measurement. PBIAS (Percentage of BIAS), R (Pearson correlation coefficient), RMSE (Root Mean Square Error), ubRMSE (unbiased RMSE), MAE (Mean Absolute Error), and MBE (Mean Bias Error) were selected for evaluation. The performance of five products over six observation networks and various land cover types was compared, and the differences were analyzed at monthly, seasonal, and annual scales. The results show that SMAP had the smallest deviation with the ISMN data because PBIAS was around -0.13, and MBE was around -0.02 m3/m3. ESA CCI performed the best in almost all aspects; its R reached around 0.7, and RMSE was only around 0.07 m3/m3 at the three time scales. The performance of the AMSR2 products varied greatly across the time scales, and increasing errors and deviations showed from 2016 to 2020. The PBO_H2O and USCRN networks could reflect soil moisture characteristics in the continental US, while iRON performed poorly. The evaluation of the networks was closely related to spatial distributions. All products performed better over grasslands and shrublands with R, which was greater than 0.52, and ubRMSE was around 0.1 m3/m3, while products performed worse over forests, where PBIAS was less than -0.62, and RMSE was greater than 0.2 m3/m3, except for ESA CCI. From the boxplot, SMAP was close to the ISMN data with differences less than 0.004 m3/m3 between the median and lower quartiles.

An Examination of the SMAP Operational Soil Moisture Products Accuracy at the Tibetan Plateau

Surface soil moisture (SSM) plays an essential role in the Earth’s water cycle and land surface processes as well as in vegetative growth, ecological health, and ecosystem properties. Particularly, information on this parameter’s spatiotemporal variability at the Tibetan Plateau is of key importance to the study of climate and the impact of climate change due to it is distinctive characteristics in this area. The present study assesses the operational SSM products provided by the SMAP (Soil Moisture Active and Passive) satellite at the Tibetan Plateau, Naqu observational station, China. In particular, the globally distributed Level 3 operational products, SPL3SMP_36km and the Enhanced Passive SSM Product SPL3SMP_9km, are evaluated in two-phases. SSM and the surface temperature estimates by SPL3SMP_36km and SPL3SMP_9km are compared against corresponding ground data available at the Naqu observation network. All in all, the examined products captured the SSM dynamics in the studied area. The results showed that precipitation is the key driving source of SSM variability. SSM fluctuated significantly and was dependent on precipitation in the studied region. Statistical metrics, such as the root mean square error (RMSE), varied for SPL3SMP_36km and SPL3SMP_9km in the ranges of 0.036–0.083 m3/m3 and 0.074–0.097 m3/m3, respectively. The unbiased RMSE (ubRMSE) was higher than the SMAP uncertainty limit (0.04 m3/m3) in most cases. This study establishes some of the causes for the different performances of SMAP products, mainly, the ancillary input dataset parameterizations, and, specifically, the surface temperature parameterization schemes of SMAP retrieval algorithm is analyzed and discussed. Our research findings highlight, among others, the usefulness of those SSM products from SMAP, particularly in mesoscale studies, providing additional useful insights into the use of those products in practice in China and globally.

Evaluating Root-Zone Soil Moisture Products from GLEAM, GLDAS, and ERA5 Based on In Situ Observations and Triple Collocation Method over the Tibetan Plateau

Abstract Root-zone soil moisture (RZSM) is an important variable in land–atmosphere interactions, notably affecting the global climate system. Contrary to satellite-based acquisition of surface soil moisture, RZSM is generally obtained from model-based simulations. In this study, in situ observations from the Naqu and Pali networks that represent different climatic conditions over the Tibetan Plateau (TP) and a triple collocation (TC) method are used to evaluate model-based RZSM products, including Global Land Evaporation Amsterdam Model (GLEAM) (versions 3.5a and 3.5b), Global Land Data Assimilation System (GLDAS) (versions 2.1 and 2.2), and the fifth-generation European Centre for Medium-Range Weather Forecasts reanalysis (ERA5). The evaluation results based on in situ observations indicate that all products tend to overestimate but could generally capture the temporal variation, and ERA5 exhibits the best performance with the highest R (0.875) and the lowest unbiased RMSE (ubRMSE; 0.015 m3 m−3) against in situ observations in the Naqu network. In the TC analysis, similar results are obtained: ERA5 has the best performance with the highest TC-derived R (0.785) over the entire TP, followed by GLEAM v3.5a (0.746) and GLDAS-2.1 (0.682). Meanwhile, GLEAM v3.5a and GLDAS-2.1 outperform GLEAM v3.5b and GLDAS-2.2 over the entire TP, respectively. Besides, possible error causes in evaluating these RZSM products are summarized, and the effectiveness of TC method is also evaluated with two dense networks, finding that TC method is reliable since TC-derived R is close to ground-derived R, with only 6.85% mean relative differences. These results using both in situ observations and TC method may provide a new perspective for the soil moisture product developers to further enhance the accuracy of model-based RZSM over the TP. Significance Statement The purpose of this study is to better understand the quality and applicability of GLEAM, GLDAS, and ERA5 RZSM products over the TP using both in situ observations and the triple collocation (TC) method, making it better applied to climate and hydrological research. This study provides four standard statistical metrics evaluation based on in situ observations, as well as the reliable metric, that is, correlation coefficient (R) derived from TC method, and highlights that TC-based evaluation could supplement the ground-based validation, especially over the data-scarce TP region.

Dual-Frequency Retrieval of Soil Moisture from L- and S-Band Radar Data for Corn and Soybean

Radar backscattering responds differently to soil moisture due to vegetation effects depending on the microwave frequency. The retrieval of soil moisture using a single frequency has been common. In this paper, we study how soil moisture retrieval performs using dual-frequency radar backscattering (L- and S-bands) compared with using L-band only. The dual-frequency inputs increase the amount of independent information, which is expected to reduce the uncertainty in estimating soil moisture. Forward scattering models for corn and soybean fields were previously generated and validated with the L-band for the SMAPVEX12 campaign: they are inverted as an independent test for the retrieval of soil moisture using the SMEX02 campaign data in this paper. It is demonstrated that L-band modeling of forward scattering processes is scalable at the S-band, in that the physics and parameters behind modeling the vegetation effects remain the same between L- and S-bands. Either L- or S-band single-frequency retrieval has reliable performance for soil moisture retrieval. Furthermore, averaging the retrieved soil moisture from both frequencies further improves the retrieval performance. The averaging avoids the determination of the weights of the L- and S-band sigma0 during the cost function minimization. The dual frequency retrieval is evaluated with the unbiased RMSE of 0.031 and 0.057 m3/m3 for corn and soybean, respectively, which are improvements by up to 0.010 and 0.004 m3/m3, compared with single-frequency cases. The findings here can apply to the upcoming NISAR mission featuring L- and S-bands.

Applicability evaluation of multiple sets of soil moisture data on the tibetan plateau

Based on the in situ data of the soil moisture-observation networks established at Maqu, Naqu, Ali, and Shiquanhe (Sq) on the Tibetan Plateau (TP), and using five evaluation indices [Pearson correlation coefficient (R), root mean square error (RMSE), mean deviation (bias), standard deviation ratio (SDV), and unbiased RMSE (ubRMSE)], the applicability of soil moisture datasets [COMBINED, ACTIVE, PASSIVE, ERA5, ERA5-Land (LAND), ERA-Interim (INTERIM), CLSM, and NOAH] was comprehensively evaluated. The results showed that, at the observation-network scale, ACTIVE exhibited the best applicability in Maqu (R = 0.704, ubRMSE = 0.040 m3/m3), COMBINED performed best in Naqu (R = 0.803, bias = 0.016 m3/m3), LAND displayed the best consistency with observations in Ali (R = 0.734, bias = −0.035 m3/m3), and ERA5 not only showed the best performance in Sq (R = 0.793, bias = −0.037 m3/m3) but also exhibited good results in the other three observation networks (R &amp;gt; 0.6). In a smaller-scale evaluation in Maqu, ACTIVE performed best, followed by ERA5. The COMBINED and PASSIVE products had serious gaps in Ali and Sq, and had the worst applicability in the western TP. In conclusion, considering the correlation results and temporal and spatial continuities, ERA5 is the most suitable soil moisture dataset for the TP.

Ground surface elevation changes over permafrost areas revealed by multiple GNSS interferometric reflectometry

Ground subsidence and uplift caused by the annual thawing and freezing of the active layer are important variables in permafrost studies. Global positioning system interferometric reflectometry (GPS-IR) has been successfully applied to retrieve the continuous ground surface movements in permafrost areas. However, only GPS signals were used in previous studies. In this study, using multiple global navigation satellite system (GNSS) signal-to-noise ratio (SNR) observations recorded by a GNSS station SG27 in Utqiaġvik, Alaska during the period from 2018 to 2021, we applied multiple GNSS-IR (multi-GNSS-IR) technique to the SNR data and obtained the complete and continuous ground surface elevation changes over the permafrost area at a daily interval in snow-free seasons in 2018 and 2019. The GLONASS-IR and Galileo-IR measurements agreed with the GPS-IR measurements at L1 frequency, which are the most consistent measurements among all multi-GNSS measurements, in terms of the overall subsidence trend but clearly showed periodic noises. We proposed a method to reconstruct the GLONASS- and Galileo-IR elevation changes by specifically grouping and fitting them with a composite model. Compared with GPS L1 results, the unbiased root mean square error (RMSE) of the reconstructed Galileo measurements reduced by 50.0% and 42.2% in 2018 and 2019, respectively, while the unbiased RMSE of the reconstructed GLONASS measurements decreased by 41.8% and 25.8% in 2018 and 2019, respectively. Fitting the composite model to the combined multi-GNSS-IR, we obtained seasonal displacements of − 3.27 ± 0.13 cm (R2 = 0.763) and − 10.56 ± 0.10 cm (R2 = 0.912) in 2018 and 2019, respectively. Moreover, we found that the abnormal summer heave was strongly correlated with rain events, implying hydrological effects on the ground surface elevation changes. Our study shows the feasibility of multi-GNSS-IR in permafrost areas for the first time. Multi-GNSS-IR opens up a great opportunity for us to investigate ground surface movements over permafrost areas with multi-source observations, which are important for our robust analysis and quantitative understanding of frozen ground dynamics under climate change.

Towards constraining soil and vegetation dynamics in land surface models: Modeling ASCAT backscatter incidence-angle dependence with a Deep Neural Network

A Deep Neural Network (DNN) is used to estimate the Advanced Scatterometer (ASCAT) C-band microwave normalized backscatter (σ40o), slope (σ′) and curvature (σ″) over France. The Interactions between Soil, Biosphere and Atmosphere (ISBA) land surface model (LSM) is used to produce land surface variables (LSVs) that are input to the DNN. The DNN is trained to simulate σ40o, σ′ and σ″ from 2007 to 2016. The predictive skill of the DNN is evaluated during an independent validation period from 2017 to 2019. Normalized sensitivity coefficients (NSCs) are computed to study the sensitivity of ASCAT observables to changes in LSVs as a function of time and space. Model performance yields a near-zeros bias in σ40o and σ′. The domain-averaged values of ρ are 0.84 and 0.85 for σ40o and σ′, compared to 0.58 for σ″. The domain-averaged unbiased RMSE is 8.6% of the dynamic range for σ40o and 13% for σ′, with land cover having some impact on model performance. NSC results show that the DNN-based model could reproduce the physical response of ASCAT observables to changes in LSVs. Results indicated that σ40o is sensitive to surface soil moisture and LAI and that these sensitivities vary with time, and are highly dependent on land cover type. The σ′ was shown to be sensitive to LAI, but also to root zone soil moisture due to the dependence of vegetation water content on soil moisture. The DNN could potentially serve as an observation operator in data assimilation to constrain soil and vegetation water dynamics in LSMs.

A Novel Fusion Method for Generating Surface Soil Moisture Data With High Accuracy, High Spatial Resolution, and High Spatio‐Temporal Continuity

AbstractSurface soil moisture (SSM) has a considerable impact on land‐atmosphere exchanges of water and energy fluxes. However, due to the inherent deficiencies of remotely sensed data (e.g., cloud contamination in thermal remote sensing and coarse resolutions for microwave remote sensing), none of the current algorithms alone can provide daily and seamless field‐scale (30 m) SSM information. To fill the gap, we proposed a novel SSM fusion framework to Generate high Resolution, Accurate, Seamless data using Point‐Surface fusion (GRASPS) based on remotely sensed, reanalysis, and in‐situ data sets. First, 30 m seamless continuous SSM correlated variables (land surface temperature, NDVI, and albedo) were downscaled by enhanced spatial and temporal adaptive reflectance fusion model. Then, downscaled auxiliary variables and other background variables were input into a deep learning model to produce 30 m daily and seamless SSM fields. To further improve the SSM estimation accuracy, a pixel classification‐based bias correction method was developed. The GRASPS method was validated over an in situ soil moisture sensor network (HOAL network) in Austria. The average Pearson correlation coefficient, root mean square error (RMSE), unbiased RMSE (ubRMSE), bias, and mean absolute error (MAE) over all validation sites achieved 0.78, 0.048, 0.033, −0.001, and 0.041 , respectively. After bias correction, RMSE, ubRMSE, bias, and MAE decreased by 13%, 7%, 22%, and 18%, respectively. The proposed method maximizes the potential of data fusion and deep learning in generating field‐scale seamless SSM, which is promising for fine‐scale studies and applications in agricultural, hydrological, and environmental domains.

A global 1‐km downscaled SMAP soil moisture product based on thermal inertia theory

AbstractMicrowave remote sensing technology has been applied to produce soil moisture (SM) retrievals on a global scale for various studies and applications. However, due to the limitations of current technology, the native spatial resolution of currently available passive microwave SM products is on the order of tens of kilometers, and this resolution cannot be used to characterize SM variability on a regional scale. To overcome this limitation, a downscaling algorithm based on the thermal inertia theory–derived relationship between SM and temperature difference was developed using outputs from the Global Land Data Assimilation System–Noah Land Surface Model and the land long‐term data record–Advanced Very High Resolution Radiometer normalized difference vegetation index (NDVI) dataset and applied to the Aqua Moderate Resolution Imaging Spectroradiometer land surface temperature/NDVI data to produce a downscaled 1‐km Soil Moisture Active Passive (SMAP) radiometer daily SM product, respectively, at 6:00 a.m. and 6:00 p.m. on a global scale from 2015 to 2020. The evaluation results reveal that the downscaling model performs better in the middle or low latitudes than in high latitudes. It also performs better in warm months than in cold months. The in situ SM observations from dense networks around the world were used to validate the 1‐km and enhanced 9‐km SMAP SM data. The validation metrics indicated that both the 1‐km and 9‐km SM data have overall overestimation trends, and the unbiased RMSE (0.063 m3 m–3 on average), mean absolute error (0.052 m3 m–3 on average), and spatial standard deviation (0.025 m3 m–3 on average) of the 1 km data are generally more accurate than the metrics of the 9‐km SM data, which indicates that the downscaled data provide reliable observed SM information.

Using ensemble learning to take advantage of high-resolution radar backscatter in conjunction with surface features to disaggregate SMAP soil moisture product

ABSTRACT Remote sensing based retrieved of soil moisture from low frequency passive microwave observations is preferred in different aspects such as better spatial coverage and more measurement compared to traditional ground-based measurements. However, due to coarse spatial resolution of the observations, their applications are limited in local to regional studies. This paper provides a framework using random forest regression to disaggregate the daily SMAP enhanced soil moisture (SPL3SMP_E) utilizing several ancillary data to overcome the spatial resolution limits and cloudiness effects. Ancillaries were acquired from sentinel-1 radar, MODIS monthly NDVI, land cover, topography, and surface soil properties. To validate the downscaled results with 1-km spatial resolution, the OZNET soil moisture measurements and sparse TDR ground soil moisture measurements were collected from Murrumbidgee catchment (Australia) and Firozabad catchment (Iran), respectively. Downscaled soil moisture product unbiased root-mean-square error (UnbRMSE) of ensemble learning demonstrated a range of 0.023 and –0.07 cm3/cm3. The produced downscaled soil moisture exhibited better local heterogeneity when compared to the coarse data and tracked the dynamics of temporal changes in soil moisture. Furthermore, cumulative distribution function (CDF) analysis showed good accuracy of downscaled soil moisture in grassland and cropland. Taken together, the findings supported usefulness of the suggested methodology in downscaling the medium- resolution SMAP soil moisture product.

Comparative Analysis of GEDI’s Elevation Accuracy from the First and Second Data Product Releases over Inland Waterbodies

Spaceborne LiDAR altimetry has been demonstrated to be an essential source of data for the estimation and monitoring of inland water level variations. In this study, water level estimates from the Global Ecosystem Dynamics Investigation (GEDI) were validated against in situ gauge station records over Lake Geneva for the period between April 2019 and September 2020. The performances of the first and second releases (V1 and V2, respectively) of the GEDI data products were compared, and the effects on the accuracy of the instrumental and environmental factors were analyzed in order to discern the most accurate GEDI acquisitions. The respective influences of five parameters were analyzed in this study: (1) the signal-over-noise ratio (SNR); (2) the width of the water surface peak within the waveform (gwidth); (3) the amplitude of the water surface peak within the waveform (A); (4) the viewing angle of GEDI (VA); and (5) the acquiring beam. Results indicated that all these factors, except the acquiring beam, had an effect on the accuracy of GEDI elevations. Nonetheless, using VA as a filtering criterion was demonstrated to be the best compromise between retained shot count and water level estimation accuracy. Indeed, by choosing the shots with a VA ≤ 3.5°, 74.6% of the shots (after an initial filter) were retained with accuracies similar to choosing A &gt; 400 (46.2% retained shots), SNR &gt; 15 dB (63.3% retained shots), or gwidth &lt; 10 bins (46.5% of retained shots). Finally, the comparison between V1 and V2 elevations showed that V2, overall, provided elevations with a more constant, but higher, bias and fewer deviations to the in situ data than V1. Indeed, by choosing GEDI shots with VA ≤ 3.5°, the unbiased RMSE (ubRMSE) of GEDI elevations was 27.1 cm with V2 (r = 0.66) and 42.8 cm with V1 (r = 0.34). Results also show that the accuracy of GEDI (ubRMSE) does not seem to depend on the beam number and GEDI acquisition dates for the most accurate GEDI acquisitions (VA ≤ 3.5°). Regarding the bias, a higher value was observed with V2, but with lower variability (54 cm) in comparison to V1 (35 cm). Finally, the bias showed a slight dependence on beam GEDI number and strong dependence on GEDI dates.

Quantifying Uncertainties in Passive Microwave Remote Sensing of Soil Moisture via a Bayesian Probabilistic Inversion Method

Improving the accuracy of remotely sensed soil moisture (SM) is a challenging and popular topic. Quantifying the uncertainty in the SM inversion process and enhancing the confidence of SM retrieval are promising ways to address this challenge but have received little attention. We present a Bayesian probabilistic inversion algorithm that can simultaneously retrieve SM, surface roughness, and vegetation optical depth data and quantify the uncertainty in the inversion. The proposed algorithm is evaluated using airborne polarimetric L-band multibeam radiometer (PLMR) observations. We use three combinations, 3-angular observations at V-polarization (3CV), 3-angular observations at H-polarization (3CH) and 6-channel observations (6CA), to identify the optimal configuration for SM retrieval by taking advantage of the PLMR’s dual-polarization and multiple angles. Uncertainties are quantified by introducing multiple uncertainty quantification metrics into Bayesian posterior distributions of SM retrievals. The estimates are validated against multiscale ground-based measurements, including manual measurements and wireless sensor network (WSN) measurements, and the spatial representativeness of the ground-based reference regarding the validation of pixel-scale SM retrievals is discussed. The 6CA attempt yields the best SM estimates (correlation coefficient (R) ≥ 0.864, root mean square error (RMSE) ≤0.04 m3/m3, and unbiased RMSE (unRMSE) ≤ 0.035 m3/m3), while the 3CH attempt yields the lowest uncertainty. Additionally, dense manual measurements are more representative than sparsely distributed WSN measurements. Overall, combining dual-polarized observations yields the best SM estimates but introduces additional uncertainty. This study highlights uncertainties quantification in SM inversion and thus provides confidence in SM inversion, facilitating improved SM retrieval algorithms.