Triple Collocation Technique Research Articles

Enhancing streamflow simulation accuracy in ungauged catchments via parameter calibration with triple collocation-based merged evapotranspiration and streamflow features

Estimating model parameters for accurate hydrological simulations poses a challenge in ungauged catchments. Evapotranspiration (ET) data derived from remote sensing (RS) holds promise for parameter calibration in ungauged catchments. Nonetheless, uncertainties inherent in raw RS-based ET data can pose challenges in the calibration process. This study evaluates two calibration methods mitigating the impact of this uncertainty. The first, a single-variable calibration, utilizes merged ET data based on the triple collocation technique, while the second incorporates streamflow features alongside ET data for multi-variable calibration. Applying the variable infiltration capacity model to 49 Chinese catchments revealed the superiority of the merged ET-based calibration scheme over the three raw RS ET-based calibrations, with improvements noted in the Nash-Sutcliffe efficiency (NSE) and Kling-Gupta efficiency (KGE) coefficients for simulated streamflow. On average, the NSE and KGE for the simulated streamflow demonstrate improvements ranging from 0.023 to 0.080 and 0.026 to 0.061, respectively. The multi-variable calibration schemes exhibit a significant enhancement in the reliability of streamflow simulations compared to the ET-based single-variable calibration schemes. On average, the NSE and KGE for the simulated streamflow exhibit improvements of 0.054 to 0.133 and 0.062 to 0.122, respectively. The performance of these calibrations was generally higher in humid catchments but showed more pronounced improvements in arid catchments when compared to raw RS ET-based calibrations. In conclusion, the proposed calibrations using streamflow features and/or merged ET data offer viable solutions for parameter calibration in ungauged catchments.

Triple collocation-based merging of multi-source gridded evapotranspiration data in the Nordic Region

Accurate evapotranspiration (ET) data are required for many hydro-meteorological applications. Compared with the traditional evaluation that requires in-situ measurements, the triple collocation (TC) technique estimates geophysical product errors without the need for ground truth, which is especially suitable over large areas lacking a dense in-situ network. However, violations of the zero-error cross-correlation (ECC) assumption are found to be the dominant sources of impairing the TC robustness. This study presents the first application of a TC-based merging framework that optimally considers ECC to merge multi-source gridded ET products in the Nordic Region during 2003–2018. The ECC estimates of each ET dataset pair calculated by the quadruple collocation approach are used to select the qualified triplets from four products, including FLUXCOM, Global Land Surface Satellite (GLASS), Global Land Evaporation and Amsterdam Model (GLEAM), and Penman-Monteith-Leuning Version 2 (PML-V2). Then the ET merged datasets are generated by weighting TC-based rescaled error variances of the parent datasets through least square merging. Finally, the accuracy of both the parent and the merged datasets are assessed with the Integrated Carbon Observation System (ICOS) flux data in the Nordic Region based on multiple statistical metrics. Results demonstrate that the ECC values provide intuitive evidence for filtering unqualified TC triplets. Both the absolute and relative error variances (signal-to-noise ratio) are considered for ET dataset evaluation. Overall PML-V2 has the best performance among the evaluated four products. Two merged ET datasets with the reference climatology of FLUXCOM outperform all parent products with the lowest errors by using ICOS data as reference among all sites – indicating the feasibility of TC technique for improving ET accuracy in the Nordic Region. This study also analyses the impacts of reference climatology selection on the TC merged results.

Spatiotemporal estimation of model error to improve soil moisture analysis in ensemble Kalman filter data assimilation

As a key variable in the surface hydrological cycle and energy balance, soil moisture is often analyzed using data assimilation to improve precise and high spatiotemporal resolution from multisource data. However, accurate estimation of model error in soil moisture data assimilation is not sufficient in spatial and temporal dimensions. This paper proposes a spatiotemporal error estimation method that integrates the spatial and temporal dimensions of the study area into model error estimation. The in-situ data was regarded as the “true value” for the temporal dimensions model error estimation. The triple collocation (TC) technique was used to estimate the spatial error of the model. Based on the aforementioned estimation, a linear method to construct the relationship between spatial and temporal information was proposed to fuse the spatiotemporal model error. Result from the proposed method was compared against the overall model error estimation and spatial estimation methods in ensemble Kalman filter (EnKF) data assimilation. The ERA-Interim reanalysis data and in-situ soil moisture data from Washington State in 2016 were used to evaluate the performance of EnKF data assimilation. Using statistical metrics, including the root mean square error (RMSE), the mean bias error (MBE), and the Pearson correlation coefficient (i.e., R-value), the results show that our proposed approach, which uses spatiotemporal approach to model error estimation in EnKF data assimilation, yielded more accurate soil moisture estimates. When ERA-Interim reanalysis soil moisture data was used as the reference data, compared with the overall model error estimation EnKF Data Assimilation Experiment, RMSE and MBE of spatiotemporal model error estimation EnKF Data Assimilation Experiment were reduced by 0.0062 and 0.0061 m3 / m3, respectively, and R remained basically unchanged. When using in-situ measurements as reference data, RMSE and MBE decreased by 0.0034 and 0.0038 m3 / m3, respectively, and R increased by 0.0365. Compared with the other estimation approaches in the experiments, the spatiotemporal error estimation approach greatly improved the estimation accuracy of soil moisture analysis, yielding the closest values and variation trends with the in-situ measurements. Fully considering the temporal and spatial variation of model error and improving the estimation accuracy of model error can provide more accurate model information for data assimilation. The proposed method and framework in this paper can be used in model error estimation and data assimilation, particularly in the fields of numerical weather forecasting, natural environment monitoring, and geographical environment research.

Simple Models Outperform More Complex Big-Leaf Models of Daily Transpiration in Forested Biomes.

Transpiration makes up the bulk of total evaporation in forested environments yet remains challenging to predict at landscape-to-global scales. We harnessed independent estimates of daily transpiration derived from co-located sap flow and eddy-covariance measurement systems and applied the triple collocation technique to evaluate predictions from big leaf models requiring no calibration. In total, four models in 608 unique configurations were evaluated at 21 forested sites spanning a wide diversity of biophysical attributes and environmental backgrounds. We found that simpler models that neither explicitly represented aerodynamic forcing nor canopy conductance achieved higher accuracy and signal-to-noise levels when optimally configured (rRMSE=20%; R 2=0.89). Irrespective of model type, optimal configurations were those making use of key plant functional type dependent parameters, daily LAI, and constraints based on atmospheric moisture demand over soil moisture supply. Our findings have implications for more informed water resource management based on hydrological modeling and remote sensing.

Error Characterization of Significant Wave Heights in Multidecadal Satellite Altimeter Product, Model Hindcast, and In Situ Measurements Using the Triple Collocation Technique

Abstract Ocean wave measurements are of major importance for a number of applications including climate studies, ship routing, marine engineering, safety at sea, and coastal risk management. Depending on the scales and regions of interest, a variety of data sources may be considered (e.g., in situ data, Voluntary Observing Ship observations, altimeter records, numerical wave models), each one with its own characteristics in terms of sampling frequency, spatial coverage, accuracy, and cost. To combine multiple source of wave information (e.g., for data assimilation scheme in numerical weather prediction models), the error characteristics of each measurement system need to be defined. In this study, we use the triple collocation technique to estimate the random error variance of significant wave heights from a comprehensive collection of collocated in situ, altimeter, and model data. The in situ dataset is a selection of 122 platforms provided by the Copernicus Marine Service In Situ Thematic Center. The altimeter dataset is the ESA Sea State CCI version1 L2P product. The model dataset is the WW3-LOPS hindcast forced with bias-corrected ERA5 winds and an adjusted T475 parameterization of wave generation and dissipation. Compared to previous similar analyses, the extensive (∼250 000 entries) triple collocation dataset generated for this study provides some new insights on the error variability associated to differences in in situ platforms, satellite missions, sea state conditions, and seasonal variability.

Quasi-global machine learning-based soil moisture estimates at high spatio-temporal scales using CYGNSS and SMAP observations

Global soil moisture mapping at high spatial and temporal resolution is important for various meteorological, hydrological, and agricultural applications. Recent research shows that the land surface reflection in the forward direction of Global Navigation Satellite System (GNSS) signals at L-band can convey high-resolution land surface information, including surface soil moisture. However, these signals are often affected by complex land surface characteristics and the bistatic nature of the GNSS-Reflectometry (GNSS-R) technique, resulting in a nonlinear relationship between the signals and surface soil moisture. In this work, a machine learning (ML) approach is used to map quasi-global soil moisture using bistatic reflectance observations acquired from the recently launched Cyclone GNSS (CYGNSS) mission. Specifically, several land surface parameters are obtained from remote sensing products and integrated with Soil Moisture Active Passive (SMAP) enhanced soil moisture retrievals to facilitate daily quasi-global CYGNSS soil moisture mapping at 9 km. Based on cross-validation against SMAP data, the ML algorithm is shown to be suitable for retrieving soil moisture from CYGNSS. Median values of unbiased root-mean-square-difference for the quasi-global coverage or regions with vegetation water content less than 5 kg/m2 are 0.0395 cm3/cm3 and 0.0320 cm3/cm3, respectively. Likewise, via independent evaluation against more than 100 in-situ sites, the algorithm is shown to have an unbiased root-mean-square-error of 0.0543 cm3/cm3. CYGNSS-based retrievals contain similar spatial variability as SMAP across different seasons. Moreover, through a robust triple collocation technique, the accuracy of CYGNSS soil moisture is relatively high over moderately vegetated regions with correlations ranging from 0.4 to 0.8. Based on these validation results, we argue that derived CYGNSS soil moisture estimates can supplement current global soil moisture databases and provide more frequent retrievals at 9 km.

Uncertainty Characterization of Ground‐Based, Satellite, and Reanalysis Snow Depth Products Using Extended Triple Collocation

AbstractThe optimal use of hemispheric‐scale snow depth (SD) products for various hydrometeorological applications requires a comprehensive assessment of their quality. Most previous validation studies of SD products adopted in situ observations as the ground truth, which may cause representativeness errors due to spatial scale mismatch between point‐based ground SD measurements and grid‐based SD products. The extended triple collocation (ETC) technique is a powerful tool to estimate the uncertainty of three independent data sets without assuming any one data source is an error‐free “truth” reference. This study first used the ETC to assess the uncertainty of three types of hemispheric‐scale SD products, including the ground‐based analysis Canadian Meteorological Centre (CMC), the satellite‐based Advanced Microwave Scanning Radiometer 2 (AMSR2), and the model‐based Global Land Data Assimilation System (GLDAS) SD products. Furthermore, the uncertainties of each SD product were analyzed using ETC metrics, that is, the correlation coefficient (R) and error standard deviations (STDs), with respect to several environmental and perturbing factors. Overall, the CMC outperforms the AMSR2 and GLDAS, with a higher R and a smaller STD. Considering multiple environmental and perturbing factors, the poorest performance of the three SD products is mainly found in densely vegetated regions, and they are strongly related to the forest cover fraction and surface roughness. Despite the above factors, the best performance for all three SD products is found over temperate climate regions. The results demonstrate the usefulness of the ETC approach to quantify the uncertainty of SD products particularly in remote regions with sparse in situ measurements.

Towards Consistent Soil Moisture Records from China’s FengYun-3 Microwave Observations

Soil moisture plays an essential role in the land-atmosphere interface. It has become necessary to develop quality large-scale soil moisture data from satellite observations for relevant applications in climate, hydrology, agriculture, etc. Specifically, microwave-based observations provide more consistent land surface records because they are unhindered by cloud conditions. The recent microwave radiometers onboard FY-3B, FY-3C and FY-3D satellites launched by China’s Meteorological Administration (CMA) extend the number of available microwave observations, covering late 2011 up until the present. These microwave observations have the potential to provide consistent global soil moisture records to date, filling the data gaps where soil moisture estimates are missing in the existing records. Along these lines, we studied the FY-3C to understand its added value due to its unique time of observation in a day (ascending: 22:15, descending: 10:15) absent from the existing satellite soil moisture records. Here, we used the triple collocation technique to optimize a benchmark retrieval model of land surface temperature (LST) tailored to the observation time of FY3C, by evaluating various soil moisture scenarios obtained with different bias-imposed LSTs from 2014 to 2016. The globally optimized LST was used as an input for the land parameter retrieval model (LPRM) algorithm to obtain optimized global soil moisture estimates. The obtained FY-3C soil moisture observations were evaluated with global in situ and reanalysis datasets relative to FY3B soil moisture products to understand their differences and consistencies. We found that the RMSEs of their anomalies were mostly concentrated between 0.05 and 0.15 m3 m−3, and correlation coefficients were between 0.4 and 0.7. The results showed that the FY-3C ascending data could better capture soil moisture dynamics than the FY-3B estimates. Both products were found to consistently complement the skill of each other over space and time globally. Finally, a linear combination approach that maximizes temporal correlations merged the ascending and descending soil moisture observations separately. The results indicated that superior soil moisture estimates are obtained from the combined product, which provides more reliable global soil moisture records both day and night. Therefore, this study aims to show that there is merit to the combined usage of the two FY-3 products, which will be extended to the FY-3D, to fill the gap in existing long-term global satellite soil moisture records.

Improving the Combination of Satellite Soil Moisture Data Sets by Considering Error Cross Correlation: A Comparison Between Triple Collocation (TC) and Extended Double Instrumental Variable (EIVD) Alternatives

Satellite-derived geophysical variables provide valuable information about the earth's functioning, but there are errors that limit their direct uses. Linearly combining two or more data sources is a simple and efficient method to reduce the uncertainty between the truth and observations. However, calculating the optimal weight for such a linear combination generally needs a reference “truth” that is rarely available in practical applications. To address this limitation, a triple collocation (TC) technique is often used to estimate data error by using a data triplet without the truth. The TC-based error variances are then applied to calculate the optimal weight calculation with an assumption of independence between errors. However, ignoring the error cross-correlations (ECCs) can lead to inaccurate optimal weight for a linear combination and hence may degrade the performance of calculations using merged products. Recently, an extended double instrumental variable (EIVD) method has been proposed to estimate ECC using only three products like TC. In this study, we examined the performance of an EIVD-based linear combination approach through two applications using synthetic data and actual satellite-derived soil moisture (SM) products. For direct comparison, the TC-based linear combination was also implemented using the same data sets. The verification showed that the EIVD-based products are generally better than the TC-based products and result in statistically significant differences in correlation coefficients. While the results here are based on SM, the proposed method is potentially applicable for reconstruction of other data sets that seldom have suitable references in large-scale regional settings.

Triple Collocation Analysis for Two Error-Correlated Datasets: Application to L-Band Brightness Temperatures over Land

The error characterization of satellite observations is crucial for blending observations from multiple platforms into a unique dataset and for assimilating them into numerical weather prediction models. In the last years, the triple collocation (TC) technique has been widely used to assess the quality of many geophysical variables acquired with different instruments and at different scales. This paper presents a new formulation of the triple collocation (Correlated Triple Collocation (CTC)) for the case of three datasets that resolve similar spatial scales, with two of them being error-correlated datasets. Besides, the formulation is designed to ensure fast convergence of the error estimators. This approach is of special interest in cases such that finding more than three datasets with uncorrelated errors is not possible and the amount of data is limited. First, a synthetic experiment has been carried out to assess the performance of CTC formulation. As an example of application, the error characterization of three collocated L-band brightness temperature (TB) measurements over land has been performed. Two of the datasets come from ESA (European Space Agency) SMOS (Soil Moisture and Ocean Salinity) mission: one is the reconstructed TB from the operational L1B v620 product, and the other is the reconstructed TB from the operational L1B v620 product resulting from application of an RFI (Radio Frequency Interference) mitigation technique, the nodal sampling (NS). The third is an independent dataset, the TB acquired by a NASA (National Aeronautics and Space Administration) SMAP (Soil Moisture Active Passive) radiometer. Our analysis shows that the application of NS leads to TB error reduction with respect to the current version of SMOS TB in 80% of the points in the global map, with an average reduction of approximately 1 K over RFI-free regions and approximately 1.45 K over strongly RFI-contaminated areas.

Multi-Collocation-Based Estimation of Wave Climate in a Non-Tidal Bay: The Case Study of Bagnoli-Coroglio Bay (Tyrrhenian Sea)

In this paper, the advantages of shaping a non-conventional triple collocation-based calibration of a wave propagation model is pointed out. Illustrated through a case study in the Bagnoli-Coroglio Bay (central Tyrrhenian Sea, Italy), a multi-comparison between numerical data and direct measurements have been carried out. The nearshore wave propagation model output has been compared with measurements from an acoustic Doppler current profiler (ADCP) and an innovative low-cost drifter-derived GPS-based wave buoy located outside the bay. The triple collocation—buoy, ADCP and virtual numerical point—make possible an implicit validation between instrumentations and between instrumentation and numerical model. The procedure presented here advocates for an alternative “two-step” strategy. Indeed, the triple collocation technique has been used solely to provide a first “rough” calibration of one numerical domain in which the input open boundary has been placed, so that the main wave direction is orthogonally aligned. The need for a fast and sufficiently accurate estimation of wave model parameters (first step) and then an ensemble of five different offshore boundary orientations have been considered, referencing for a more detailed calibration to a short time series of a GPS-buoy installed in the study area (second step). Such a stage involves the introduction of an enhancement factor for the European Centre for Medium-Range Weather Forecasts (ECMWF) dataset, used as input for the model. Finally, validation of the final model’s predictions has been carried out by comparing ADCP measurements in the bay. Despite some limitations, the results reveal that the approach is promising and an excellent correlation can be found, especially in terms of significant wave height.

Estimating surface soil moisture from satellite observations using a generalized regression neural network trained on sparse ground-based measurements in the continental U.S

This study attempted to develop a point-surface collaborative inversion (PSCI) method for the estimation of regional surface soil moisture (SSM) using a generalized regression neural network (GRNN) trained on sparse ground-based measurements. Specifically, GRNN was employed to establish a nonlinear relationship between ground-based measurements from sparse network stations (SNSs) and passive microwave observations from the Soil Moisture Active Passive (SMAP) satellite in the continental U.S. for April 2015 to March 2018. More importantly, the extended triple collocation (ETC) technique was applied to address the scale mismatch issue resulting from the small spatial support of ground-based measurements, whereby individual SNSs’ reliability at the SMAP coarse footprint scale could be determined before fed into GRNN. The 10-fold cross-validation results showed that the GRNN model trained on reliable SNSs obtained a satisfactory performance—the cross-validated R and unbiased RMSE values were 0.88 and 0.050 cm3 cm−3, respectively, which outperformed the back-propagation neural network (BPNN) and the other GRNN model trained on all SNSs. Furthermore, temporal and spatial comparisons between the PSCI-based SSM retrievals and other SSM datasets indicated the former agreed the best with ground measurements both in time and space, suggesting the proposed PSCI method had shown great potential in estimating reliable regional SSM climate records.

Direct Comparison and Triple Collocation: Which Is More Reliable in the Validation of Coarse‐Scale Satellite Surface Albedo Products

AbstractA rigorous and reliable validation is the prerequisite and basis to properly understand the accuracy of satellite surface albedo products. However, the representativeness error of point scale in situ measurements, which mainly results from the spatial mismatch between in situ and satellite measurements as well as the surface heterogeneities, generally hinders an effective validation of satellite surface albedo products based on the conventional direct comparison method. Triple collocation technique has been recognized as a powerful tool to estimate the errors of satellite products without requiring the truth. Here, we evaluated the effectiveness of both the direct comparison and extended triple collocation (ETC) methods in tackling with the representativeness error of in situ measurements and characterizing the performances of satellite surface albedo products (MCD43A3 [V006], GLASS, and NPP VIIRS albedos) given whether the in situ‐based pixel‐scale albedo was available or not. The ETC method is shown to be advantageous over the direct comparison based on single “point” in situ measurements, because it is less affected by representativeness errors. However, it shall not substitute for the direct comparison method based on in situ‐based pixel‐scale albedo measurements, because its effectiveness in solving representativeness errors and estimating product errors is very limited and dependent on the choice of the data in the triplets. Furthermore, the ETC technique cannot be used to estimate the true physical errors of satellite surface albedo products.

A mutual assessment of the uncertainties of digital elevation models using the triple collocation technique

ABSTRACT This study investigates the uncertainties of digital elevation models (DEMs) using the triple collocation (TC) method. DEMs from satellite missions are important for many geoscience disciplines and for economic benefits and are freely available. Validating DEMs is necessary to select an appropriate model for a given region and application. Provided certain assumptions about the error structure of any three data sets – measuring the same phenomenon – can be made, the TC approach can be used to provide an unbiased and scaled estimate of the error variances of the data sets, without requiring a reference data. We compared the TC approach to the traditional approach of using a reference data set using the Shuttle Radar Topography Mission version 4.1 (SRTM v4.1) DEM, ASTER (the Advanced Spaceborne Thermal Emission and Reflection Radiometer) GDEM (Global DEM) version 2, the 1 arc-minute global relief model (ETOPO1), a DEM compiled by the Survey and Mapping Division of Ghana (SMD DEM), and 545 ground control stations (GCSs). The error estimates for the DEMs via TC were considerably smaller than those obtained from the reference-based approach. As an example, the best performing DEM (SRTM v4.1) recorded a root-mean-square error (RMSE) of 15.601 m using the GCSs as reference, while its TC-derived accuracy was 6.517 m. We note that based on the results of the TC, the estimated error of the GCSs is approximately 14 m. Using a data set with an error of 14 m to validate other data sets is certainly bound to result in unfavorable results. Thus, we have demonstrated in this work that the TC approach is able to provide an unbiased error of DEMs. The approach is important even for regions where GCSs are highly accurate, but more so for regions with low-quality GCSs.

Evaluation and calibration of a high-resolution soil moisture product for wildfire prediction and management

Soil moisture deficit is a key variable used in operational fire prediction and management applications. In Australia, operational fire management practices use simple, empirical water balances models to estimate soil moisture deficit. The Bureau of Meteorology has recently developed a prototype, high-resolution, land surface modelling based, state-of-the-art soil moisture analyses for Australia. The present study examines this new product for use in operational fire prediction and management practices in Australia. The approach used is twofold. First, the new soil moisture product is evaluated against observations from ground based networks. Among the results, the mean Pearson’s correlation for surface soil moisture across the three in-situ networks is found to be between 0.78 and 0.85. Secondly, the study evaluate a few different calibration methods to facilitate the ready utilization of the new soil moisture product in the current operational fire prediction framework. The calibration approaches investigated here are: minimum-maximum matching, mean-variance matching and, cumulative distribution function matching. Validation of the calibrated products using extended triple collocation technique shows that the minimum-maximum method has the highest skill. Evaluation of the calibrated products against MODIS fire radiative power data highlights that large fires correspond to a drier soil in minimum-maximum outputs compared to other calibration results and the current operational method.

Error Characterization of Sea Surface Salinity Products Using Triple Collocation Analysis

The triple collocation (TC) technique allows the simultaneous calibration of three independent, collocated data sources, while providing an estimate of their accuracy. In this paper, the TC is adapted to validate different salinity data products along the tropical band. The representativeness error (the true variance resolved by the relatively high-resolution systems but not by the relatively low-resolution system) is accounted for in the validation process. A method based on the intercalibration capabilities of TC is used to estimate the representativeness error for each triplet, which is found to impact between 15% and 50% the error estimation of the different products. The method also sorts the different products in terms of their resolving spatiotemporal scales. Six salinity products (sorted from smaller to larger scales) used were: the in situ data from the Global Tropical Moored Buoy Array (TAO), the GLORYS2V3 ocean reanalysis output provided by Copernicus, the satellite-derived Aquarius Level 3 version 4 (AV4) and Soil Moisture and Ocean Salinity (SMOS) objectively analyzed (SOA) maps, and the climatology maps provided by the World Ocean Atlas (WOA). This calibration study is limited to the year 2013, a year when all the products were available. This validation approach aims to assess the quality of the different salinity products at the satellite-resolved spatiotemporal scales. The results show that, at the AV4 resolved scales, the Aquarius product has an error of 0.17, and outperforms TAO, GLORYS2V3, and the SOA maps. However, at the SOA resolved scales (which are coarser than those of the Aquarius product because of the large OA correlation radii used), the SMOS product has an error of 0.20, slightly lower than that of GLORYS2V3, Aquarius, and TAO. The WOA products show the highest errors. Higher order calibration may lead to a more accurate assessment of the quality of the climatological products.

Triple Collocation to Assess Classification Accuracy Without a Ground Truth in Case of Earthquake Damage Assessment

The assessment of satellite image classifications is usually carried out using a test sample assumed as the ground truth, from which a confusion matrix is derived. There are cases where the reference data, even those coming from a ground survey, are affected by errors and do not represent a reliable truth. In the field of geophysical parameter retrieval, the triple collocation (TC) technique is applied for validating remotely sensed products when the source of test data (e.g., ground data) does not represent a reliable reference. TC is able to retrieve the error variances of three systems observing the same target parameter, assuming that their errors are independent. In this paper, we exploit the same idea to test the classification accuracy in cases where the ground truth is not available. We extend the TC approach to the classification problem for a general number of classes, but we solve it numerically for a two-class problem (i.e., collapsed and noncollapsed buildings). The specific case refers to the detection of L’Aquila 2009 earthquake damage from very high-resolution optical data. The image classification, performed by exploiting an object-based analysis, is compared with those from two different ground surveys carried out after the earthquake by different teams and with different purposes. This paper demonstrates the power of the TC approach for assessing the classification accuracy with no reliable ground truth available, and provides an insight into the problem of assessing damage, from satellite and on ground, in a very critical and unsafe situation, like the one occurring after an earthquake. Moreover, it was found that the remotely sensed product can have an order of accuracy comparable to that of at least one of the ground surveys.

Comparing soil moisture anomalies from multiple independent sources over different regions across the globe

Abstract. Agricultural drought events can affect large regions across the world, implying the need for a suitable global tool for an accurate monitoring of this phenomenon. Soil moisture anomalies are considered a good metric to capture the occurrence of agricultural drought events, and they have become an important component of several operational drought monitoring systems. In the framework of the JRC Global Drought Observatory (GDO, http://edo.jrc.ec.europa.eu/gdo/), the suitability of three datasets as possible representations of root zone soil moisture anomalies has been evaluated: (1) the soil moisture from the Lisflood distributed hydrological model (namely LIS), (2) the remotely sensed Land Surface Temperature data from the MODIS satellite (namely LST), and (3) the ESA Climate Change Initiative combined passive/active microwave skin soil moisture dataset (namely CCI). Due to the independency of these three datasets, the triple collocation (TC) technique has been applied, aiming at quantifying the likely error associated with each dataset in comparison to the unknown true status of the system. TC analysis was performed on five macro-regions (namely North America, Europe, India, southern Africa and Australia) detected as suitable for the experiment, providing insight into the mutual relationship between these datasets as well as an assessment of the accuracy of each method. Even if no definitive statement on the spatial distribution of errors can be provided, a clear outcome of the TC analysis is the good performance of the remote sensing datasets, especially CCI, over dry regions such as Australia and southern Africa, whereas the outputs of LIS seem to be more reliable over areas that are well monitored through meteorological ground station networks, such as North America and Europe. In a global drought monitoring system, the results of the error analysis are used to design a weighted-average ensemble system that exploits the advantages of each dataset.

Error Characterization of Soil Moisture Satellite Products: Retrieving Error Cross-Correlation Through Extended Quadruple Collocation

The triple collocation (TC) technique is being increasingly used to validate soil moisture retrievals derived from different systems, like satellites, hydrological models, or in situ probes. In recent years, several extensions of this method were proposed in order to evaluate the error standard deviations of more than three systems and to soften the TC hypothesis. In this paper, a novel extended quadruple collocation (E-QC) method is proposed, in order to consider the possibility of a cross correlation between product errors, identifying automatically the couple of error cross-correlated systems. The method is applicable even to a larger number of collocated datasets, although it may be unfeasible to collect them in practice. A synthetic experiment showed promising results, concluding that the E-QC is able to individuate (if any) the pair of systems with cross-correlated errors. It correctly compensates for the latter contribution and accurately retrieves error standard deviations of each system, otherwise biased if cross correlation is not taken into account. The E-QC was applied to soil moisture retrievals provided by satellite (SMOS, ASCAT, and SMAP), model (ERA Interim), and in situ probes (ISMN). The E-QC method identified the presence of error cross-correlation between the satellite products. This was also confirmed by analyzing the five datasets all together. E-QC showed fair performances of satellite products, especially of SMAP, although not as good as in case the presence of error correlation is not correctly taken into account.

Estimating Random Errors of Scatterometer, Altimeter, and Model Wind Speed Data

Scatterometer and altimeter wind data are very important for data assimilation and verification of numerical weather prediction models. Standard deviation of absolute random errors can be estimated using the triple collocation technique. However, error correlations between various wind sources (e.g., due to data assimilation) complicate the error estimation. A method is used to alleviate the impact of error correlations between the scatterometer and the model that assimilates such data. Using twenty-two datasets of triplet composed of Jason-2 altimeter, Metop-A/B scatterometers (ASCAT-A/B, respectively), and ECMWF model analysis and forecasts (1 altimeter × 2 scatterometers × 11 model analysis and forecasts = 22 datasets) covering a period of two years from August 2013 to July 2015, the correlation coefficient between the errors of scatterometers and the model analysis was found to be about 0.33 for those datasets. This correlation reduces with forecast lead time until it almost vanishes at day seven. Altimeter and scatterometer errors are not correlated. The standard deviation of wind speed random errors of Jason-2, ASCAT-A/B, and the IFS analysis are estimated as 0.7, 0.8, and 0.9 m/s, respectively. As expected, there was no difference between ASCAT-A and ASCAT-B results.