Scatterometer Image Reconstruction Research Articles

Evaluating the effective resolution of enhanced resolution SMAP brightness temperature image products

The MEaSUREs Calibrated Enhanced-Resolution Passive Microwave Daily Equal-Area Scalable Earth Grid 2.0 Brightness Temperature (CETB) Earth System Data Record (ESDR) includes conventional- and enhanced-resolution radiometer brightness temperature (TB) images on standard, compatible grids from calibrated satellite radiometer measurements collected over a multi-decade period. Recently, the CETB team processed the first 4 years of enhanced resolution Soil Moisture Active Passive (SMAP) L-band (1.41 GHz) radiometer TB images. The CETB processing employs the radiometer form of the Scatterometer Image Reconstruction (rSIR) algorithm to create enhanced resolution images, which are posted on fine resolution grids. In this paper, we evaluate the effective resolution of the SMAP TB image products using coastline and island crossings. We similarly evaluate the effective resolution of the SMAP L1C_TB_E enhanced resolution product that is based on Backus-Gilbert processing. We present a comparison of the spatial resolution of the rSIR and L1C_TB_E enhanced resolution products with conventionally-processed (gridded) SMAP data. We find that the effective resolution of daily CETB rSIR SMAP TB images is slightly finer than that of L1C_TB_E and about 30% finer than conventionally processed data.

Resolution Enhancement of SMAP Passive Soil Moisture Estimates

The Soil Moisture Active Passive (SMAP) mission includes a unique combination of instruments intended to provide daily global soil moisture data with high accuracy and resolution. Due to radar instrument failure, the default resolution of the data product decreased from the intended 9 km to 36 km shortly after the mission started to return data. To improve this, we employed the Scatterometer Image Reconstruction algorithm in its radiometer form (rSIR) to enhance the resolution of the radiometer brightness temperature measurements from which the soil moisture was derived. This paper compares the soil moisture estimates created from the rSIR-enhanced brightness temperatures with SMAP project radiometer L2_SM_SP and SMAP-Sentinel L2_SM_P products reported on 9 km and 3 km grids, respectively. We find that the difference of the rSIR-enhanced passive soil moisture product is generally within 0.020 cm3 cm−3 RMS of the 9 km SMAP radiometer L2_SM_SP and 0.045 cm3 cm−3 RMS of the 3 km SMAP-Sentinel L2_SM_P soil moisture products. The accuracy of the rSIR soil moisture can be improved by including better antenna pattern correction methods applied to the input TB measurements.

Sensitivity Analysis of C and Ku-Band Scatterometers for River Water Level Estimation

Water-level monitoring is essential for analyzing hydrological extremes, environmental flows, sustainable development and management of water resources. Satellite altimeters have operationally been used to estimate river water levels (e.g., SARAL-Altika, Jason-2/3) with poor revisit times (10&#x2013;35 days). In this current study, we propose a method to estimate river water levels using scatterometer data with 2-day temporal resolution. Sensitivity analysis of radar frequency (Ku- and C-bands) for river water-level estimation was carried out. The concept of change in radar backscatter with an increase in soil wetness is used to develop a model function between basin water index (BWI) and observed river stage. High-resolution OceanSat Scatterometer (OSCAT) (Ku-band) and advanced scatterometer (ASCAT) (C-band) scatterometer image reconstruction (SIR), digital elevation model, and observed water stage datasets have been used in this study. We observed that C-band scatterometer shows better results than Ku-band in terms of Pearson correlation coefficient (<inline-formula> <tex-math notation="LaTeX">$R$ </tex-math></inline-formula>), Nash&#x2013;Sutcliffe Efficiency (NSE), and root mean square error (RMSE), which can be attributed to higher soil wetness sensitivity of the C-band compared with Ku-band scatterometers. <inline-formula> <tex-math notation="LaTeX">$R$ </tex-math></inline-formula> ranged between 0.93 and 0.97 for the C-band and between 0.89 and 0.95 for the Ku-band data. Similarly, NSE ranged from 0.84 to 0.95 for the C-band and between 0.56 and 0.95 for the Ku-band data over selected study sites in the Brahmaputra river basin. Model accuracy was found to be adequate for monitoring of river stage from a remote sensing platform. This study shows that scatterometer observations can be used for operational water-level estimations, which can be useful for water resource management.

Enhanced-Resolution SMAP Brightness Temperature Image Products

The NASA-sponsored Calibrated Passive Micro- wave Daily Equal-Area Scalable Earth Grid 2.0 Brightness Temperature (CETB) Earth System Data Record Project team has generated a multisensor, multidecadal time series of high-resolution radiometer products designed to support climate studies. This project uses image reconstruction techniques to generate conventional and enhanced-resolution daily brightness temperature images on a standard set of map projections. Sensors included in CETB are the Aqua Advanced Microwave Scanning Radiometer–Earth Observing System (AMSR-E), Scanning Multichannel Microwave Radiometer, and all Special Sensor Microwave/Imager and Special Sensor Microwave Imager/Sounder radiometers. These span frequencies between 6 and 89 GHz. This paper considers the issues of adding the L-band (1.6 GHz) Soil Moisture Active Passive (SMAP) radiometer measurements to the CETB climate record, with emphasis on optimizing the reconstruction to provide the highest possible spatial resolution at the lowest noise level. SMAP radiometer reconstruction on SMAP-standard grids is also considered. Simulation is used to optimize the reconstruction, and the results confirmed using actual data. A comparison of the performance of the Backus–Gilbert approach and the radiometer form of the Scatterometer Image Reconstruction algorithm is provided. These are compared to the conventional drop-in-the-bucket gridded imaging.

Scatterometer Backscatter Imaging Using Backus–Gilbert Inversion

Wind scatterometer measurements are collected over an irregular grid, and processing is required to generate backscatter images on an Earth-centered grid. The most common algorithms used for this are “drop in the bucket” (DIB) and variations of the scatterometer image reconstruction (SIR) algorithm. These algorithms are also used for radiometer brightness temperature imaging. The Backus–Gilbert (BG) algorithm has been used for radiometer imaging but has not been applied to scatterometer backscatter imaging. In this paper, the application of BG to scatterometer backscatter imaging is explored and its performance is compared to DIB and SIR. Like SIR, optimally tuned BG is capable of producing higher resolution images than DIB, though its noise performance is slightly inferior to SIR’s. While BG and SIR produce similar results for radiometer data, the higher relative noise level of scatterometer data increases the differences between the SIR and BG algorithm performance, and limits the performance of BG relative to SIR in scatterometer imaging. Comparison of the SIR and BG algorithms in scatterometer imaging offers important insights into the inversion/reconstruction problem.

Spatial Resolution and Precision Properties of Scatterometer Reconstruction Algorithms

Various reconstruction methods have been used to enhance the spatial resolution of scatterometer data. Most of the image reconstructions are two-dimensional problems, which combine multiple passes of overlapping data over the temporally homogeneous surface, and thus are only suitable for land and ice applications. This paper attempts to address the one-dimensional reconstruction to enhance the azimuth resolution of scatterometer data using a single pass of observations. Since the range resolution determined by the on-board dechirping technique is generally up to several hundred meters, the one-dimensional reconstruction is adequate for certain near real-time ocean applications, such as the development of coastal scatterometer winds. Three well-known reconstruction algorithms, including additive algebraic reconstruction technique (AART), multiplicative algebraic reconstruction technique (MART), and scatterometer image reconstruction (SIR), are evaluated. The spatial resolution and the reconstruction precision resolved by each algorithm are separately analyzed using the local impulse response and Monte Carlo methods. The dependence of the spatial resolution and the reconstruction precision on a variety of parameters, such as the mean backscatter coefficient and its variance, the beamwidth of spatial response function (SRF), and the SRF function type, is evaluated using a simulation framework. In particular, the tradeoff between the spatial resolution and the reconstruction precision is examined for three algorithms. The results show that SIR offers the quickest convergence and lowest noise.

Comparison of SeaWinds Backscatter Imaging Algorithms.

This paper compares the performance and tradeoffs of various backscatter imaging algorithms for the SeaWinds scatterometer when multiple passes over a target are available. Reconstruction methods are compared with conventional gridding algorithms. In particular, the performance and tradeoffs in conventional 'drop in the bucket' (DIB) gridding at the intrinsic sensor resolution are compared to high-spatial-resolution imaging algorithms such as fine-resolution DIB and the scatterometer image reconstruction (SIR) that generate enhanced-resolution backscatter images. Various options for each algorithm are explored, including considering both linear and dB computation. The effects of sampling density and reconstruction quality versus time are explored. Both simulated and actual data results are considered. The results demonstrate the effectiveness of high-resolution reconstruction using SIR as well as its limitations and the limitations of DIB and fDIB.

Enhanced-Resolution Reconstruction of ASCAT Backscatter Measurements

A wind scatterometer measures the normalized radar cross section σ° of the Earth's surface in order to estimate the ocean near-surface wind. Each measurement is the weighted integral of σ° over an area, so samples of σ° are filtered by a measurement spatial response function (SRF). Enhanced-resolution σ° data may be produced from σ° measurements and their associated SRF using image reconstruction algorithms. These enhancedresolution data products are useful for land and ice studies in addition to high-resolution ocean wind retrieval. The spatial resolution and noise may be improved by combining data for multiple passes, trading off temporal resolution for spatial resolution. In this paper, we consider the production of enhanced-resolution σ° image reconstruction from the Advanced Scatterometer (ASCAT) on the MetOp satellites for land and ice regions. Two previously introduced image reconstruction algorithms, the weighted AVErage (AVE) and Scatterometer Image Reconstruction (SIR), are applied to ASCAT data, as well as a conventional gridding method. We select optimum values of the reconstruction parameters for ASCAT. The imaging algorithms are compared and evaluated, as quantified by the spatial resolution, pixel mean and variance, and pixel correlation of the produced images. The effective spatial resolution of the AVE and SIR reconstructed images is on the order of 15 to 20 km, an improvement over the nominal spatial resolution of 25 and 50 km for the spatially averaged swath-oriented ASCAT L1B products.

Optimum Image Formation for Spaceborne Microwave Radiometer Products.

This paper considers some of the issues of radiometer brightness image formation and reconstruction for use in the NASA-sponsored Calibrated Passive Microwave Daily Equal-Area Scalable Earth Grid 2.0 Brightness Temperature Earth System Data Record project, which generates a multisensor multidecadal time series of high-resolution radiometer products designed to support climate studies. Two primary reconstruction algorithms are considered: the Backus-Gilbert approach and the radiometer form of the scatterometer image reconstruction (SIR) algorithm. These are compared with the conventional drop-in-the-bucket (DIB) gridded image formation approach. Tradeoff study results for the various algorithm options are presented to select optimum values for the grid resolution, the number of SIR iterations, and the BG gamma parameter. We find that although both approaches are effective in improving the spatial resolution of the surface brightness temperature estimates compared to DIB, SIR requires significantly less computation. The sensitivity of the reconstruction to the accuracy of the measurement spatial response function (MRF) is explored. The partial reconstruction of the methods can tolerate errors in the description of the sensor measurement response function, which simplifies the processing of historic sensor data for which the MRF is not known as well as modern sensors. Simulation tradeoff results are confirmed using actual data.

Evaluation of the Interactive Multisensor Snow and Ice Mapping System (IMS) for monitoring sea ice phenology

We present an evaluation of the Interactive Multisensor Snow and Ice Mapping System (IMS) for monitoring northern hemisphere sea ice phenology. Analysts utilize a variety of datasets to manually derive the daily extent of snow, ice, water and land, available at both 24 and 4km. The 4km IMS product was assessed for 2004–2008 against several previously established melt/freeze algorithms using Scatterometer Image Reconstruction (SIR) SeaWinds/QuikSCAT (QuikSCAT) backscatter (σ°), Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E) brightness temperature (TB) measurements, data from the Special Sensor Microwave/Image data (SSM/I) and sea ice concentrations derived from DMSP Special Sensor Microwave/Imager–Special Sensor Microwave Imager Sounder (SSMI–SSMIS) data (NASATeam dataset). The resolution possible with the 4km IMS product allows for better spatial representation of sea ice along the coastlines, the ice edges and in the narrow channels of the Canadian Arctic Archipelago as compared to the microwave products. IMS detects open water earlier and freeze onset later than the automated microwave products, and also allows for the detection of opening, and the subsequent closing, of leads that the other datasets are unable to detect. Using RADARSAT-1 imagery for evaluation, IMS is shown to outperform the other datasets for the timing and extent of the first open water detection. IMS identified between 17 and 53% greater open water coverage than the other datasets in the narrow channels of the Northwest Passage (Barrow Strait). In order to further the use of IMS for sea ice applications, we derived two new spatial datasets using the full record of IMS data (4km: 2004–present, 24km: 1997–present): melt duration to open water (duration from melt onset detected with SSM/I passive microwave until open water detected by IMS) and first year ice cover duration (duration from freeze onset until open water, both detected by IMS).

Multiyear Arctic Sea Ice Classification Using QuikSCAT

Long-term trends in Arctic sea ice are of particular interest in studies of global temperature, climate change, and industrial application. This paper analyzes intra-annual and interannual trends in Ku-band backscatter over first-year (FY) and multiyear (MY) sea ice to develop a new sea-ice-type classification method. Histograms of backscatter are derived from high-resolution backscatter images created using the scatterometer image reconstruction (SIR) algorithm applied to measurements obtained by the SeaWinds instrument aboard QuikSCAT. The backscatter of FY and MY sea ice are clearly identifiable and are observed to vary seasonally. Using an average of the annual backscatter trends obtained from QuikSCAT, a classification of MY ice is obtained, which uses a time-dependent threshold value. Validation of the classification method is done using regional ice charts from the Canadian Ice Service. Differences in ice classification are found to be less than 6% for the winters of 2006-2007 and 2007-2008, and the end of 2008. Anomalies in the distribution of sea ice backscatter from year to year suggest a reduction in MY ice cover between 2003 and 2009 and an approximately equivalent increase in FY ice cover.

Reconstruction From Aperture-Filtered Samples With Application to Scatterometer Image Reconstruction

This paper approaches scatterometer image reconstruction as the inversion of a discrete noisy aperture-filtered sampling operation. Aperture-filtered sampling is presented and contrasted with conventional and irregular sampling. Discrete reconstruction from noise-free aperture-filtered samples is investigated and contrasted with conventional continuous reconstruction approaches. The discrete approach enables analytical treatment of the reconstruction grid resolution and the effective resolution imposed by the sampling and reconstruction operations. The noisy case is also explored. A reconstruction estimator based on maximum a posteriori (MAP) estimation is proposed to recover the conventional samples from noisy scatterometer measurements. This approach enables the scatterometer noise distribution to be appropriately accounted for in the reconstruction operation. The MAP and conventional reconstruction approaches are applied to the SeaWinds scatterometer and the Advanced Wind Scatterometer, and the effective resolution of the different methods is analyzed. The MAP approach produces results consistent with the well-established scatterometer image reconstruction (SIR) algorithm. The MAP approach significantly enhances the resolution at the expense of increased noise. Although a detailed noise-versus-resolution tradeoff analysis is beyond the scope of this paper, the new framework allows for a more general treatment than the ad hoc tuning parameters of the SIR algorithm.

Development of a water clear of sea ice detection algorithm from enhanced SeaWinds/QuikSCAT and AMSR-E measurements

We develop and evaluate water clear of sea ice (open water following ice cover) detection algorithms that make use of Scatterometer Image Reconstruction (SIR) SeaWinds/QuikSCAT (QuikSCAT) backscatter (σ°) and Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E) brightness temperature (T B) measurements. Algorithm validation was performed within Canadian Arctic waters using the Canadian Ice Service Digital Archive (CISDA) ice charts, NASATeam ice concentration estimates, extended AVHRR Polar Pathfinder (APP-x) albedo data, RADARSAT-1 imagery, and MODIS imagery. Results indicate that the temporal evolution of QuikSCAT σ°, AMSR-E polarization ratio (PR18), and AMSR-E vertical spectral gradient ratio (GR3618) can detect water clear of sea ice events, however mean differences due to frequency dependent characteristics of the data (spatial resolution; sensitivity to open water) were apparent. All water clear of sea ice algorithms are in good agreement with the timing and clearing patterns given by the CISDA. The QuikSCAT algorithm provided a more representative ice edge and more details on the ice clearing process due to higher spatial resolution, however, transient clearing events were better represented by the AMSR-E PR(18) or (GR3618) algorithm. By exploiting the strengths of each sensor, we found that a QuikSCAT and AMSR-E fused algorithm provide improved open water area estimates by as much as 11%. The fusion of QuikSCAT and AMSR-E PR(18) yielded in the most spatially representative open water detection. The residual surface of the water clear of sea ice algorithms was found to provide another measure of the average September minimum pan-Arctic sea ice extent within 6% of the NASATeam algorithm estimates.

Greenland snow accumulation estimates from satellite radar scatterometer data

Data collected by the C band ERS‐2 wind scatterometer (EScat), the Ku band ADEOS‐1 NASA scatterometer (NSCAT), and the Ku band SeaWinds on QuikScat (QSCAT) satellite instruments are used to illustrate spatiotemporal variability in snow accumulation on the Greenland ice sheet. Microwave radar backscatter images of Greenland are derived using the scatterometer image reconstruction (SIR) method at 3‐day intervals over the periods 1991–1998 and 1996–1997 for EScat and NSCAT, respectively. The backscatter coefficient σ° normalized to 40° incidence, A, and gradient in backscatter, B, in the range 20°–60° are compared with historical snow accumulation data and recent measurements made in the Program for Arctic Regional Climate Assessment (PARCA) shallow snow pits. Empirical relationships derived from these comparisons reveal different exponential relationships between C and Ku band A values and dry snow zone mean annual accumulation, Q. Frequency difference images between overlapping scatterometer images suggest that C band data are more sensitive to snow layering and buried inhomogeneities, whereas Ku band data are more sensitive to volume scattering from recently accumulated snow. Direct comparisons between NSCAT B values and in situ Q measurements show a linear relationship between ln (Q) and B, with a negative rank correlation of R = −0.8. The root‐mean‐square residual in fitting regression line equation ln (Q) = 3.08 − 17.83B to the data is 0.05‐m snow water equivalent. This simple Ku band empirical relationship is exploited to investigate decadal changes in dry snow zone accumulation between Seasat (1978) and NSCAT (1996). Additional comparisons between NSCAT and recent QSCAT (1999) data reveal significant upslope shifts in the dry snow line along the southwestern flank of the ice sheet. Recent acceleration in the increase in intensity of scattering is observed in the percolation zone, suggesting increased melting between 2000‐ and 3000‐m elevation in the southern half of the ice sheet.

Image reconstruction and enhanced resolution imaging from irregular samples

While high resolution, regularly gridded observations are generally preferred in remote sensing, actual observations are often not evenly sampled and have lower-than-desired resolution. Hence, there is an interest in resolution enhancement and image reconstruction. This paper discusses a general theory and techniques for image reconstruction and creating enhanced resolution images from irregularly sampled data. Using irregular sampling theory, we consider how the frequency content in aperture function-attenuated sidelobes can be recovered from oversampled data using reconstruction techniques, thus taking advantage of the high frequency content of measurements made with nonideal aperture filters. We show that with minor modification, the algebraic reconstruction technique (ART) is functionally equivalent to Grochenig's (1992) irregular sampling reconstruction algorithm. Using simple Monte Carlo simulations, we compare and contrast the performance of additive ART, multiplicative ART, and the scatterometer image reconstruction (SIR) (a derivative of multiplicative ART) algorithms with and without noise. The reconstruction theory and techniques have applications with a variety of sensors and can enable enhanced resolution image production from many nonimaging sensors. The technique is illustrated with ERS-2 and SeaWinds scatterometer data.

Cryosphere applications of NSCAT data

Though designed to measure vector winds over the ocean, new imaging techniques facilitate the use of NASA scatterometer data (NSCAT) in cryosphere studies. NSCAT provides data of unprecedented coverage, resolution, and quality which, when coupled with the scatterometer image reconstruction with filtering (SIRF) algorithm, enables images of /spl sigma//sup o/ at resolutions approaching 8 km over stationary targets. Such images are useful in ice mapping and classification, and multidecadal studies are possible by comparison with Seasat Scatterometer (SASS) data. The utility of NSCAT data in polar ice studies is illustrated through a review of two cryosphere applications of NSCAT data: (1) sea-ice mapping and tracking and (2) ice-sheet change in Greenland and Antarctica. The wide swath and frequent revisit, coupled with incidence and azimuth angle diversity makes NSCAT data very effective in mapping the extent of sea-ice. In Greenland, snow and ice facies are clearly delineated on the basis of the seasonally dependent radar backscattering cross section, due to sensitivity of radar backscatter to diagenetic changes occurring at and beneath the surface. Comparison of NSCAT and SASS data enables study of change in Greenland between 1978 and 1996.

Spatial resolution enhancement of SSM/I data

One of the limitations in using Special Sensor Microwave/Imager (SSM/I) data for land and vegetation studies is the relatively low-spatial resolution. To ameliorate this limitation, resolution-enhancement algorithms can be applied to the data. In this paper, the Backus-Gilbert inversion (BGI) technique and the scatterometer image-reconstruction (SIR) algorithm are investigated as possible methods for creating enhanced resolution images from SSM/I data. The two algorithms are compared via both the simulation and the actual SSM/I data. The algorithms offer similar resolution enhancement, though SIR requires significantly less computation. Sample results over two land regions of South America are presented.