Improved retrieval of sea ice total concentration from spaceborne passive microwave observations using numerical weather prediction model fields: An intercomparison of nine algorithms

The accuracy of the sea ice concentration estimates in polar regions is reduced by the effects of atmospheric emission and absorption. A method is presented where a fast atmospheric radiative transfer model and Numerical Weather Prediction (NWP) model fields are used to correct brightness temperatures before they enter the sea ice concentration algorithm. The skill of the method is a function of the errors in the NWP model fields modulated by the sensitivities of the sea ice concentration algorithm used. The NWP model fields representing the most significant atmospheric parameters, i.e. water vapour, cloud liquid water, surface temperature and wind speed over open water are evaluated using remote sensing data. For wind speed and total water vapour, it is found that the standard deviation of the difference is less than the RMS error quoted for the remote sensing algorithms. The best consistency is found for water vapour followed by wind speed. The NWP model cloud liquid water displays standard deviations much higher than the RMS error of the remote sensing algorithm and close to the total average content. Nine sea ice concentration algorithms are further evaluated in a sensitivity study to the above-mentioned atmospheric constituents using a detailed atmospheric radiative transfer model. The result shows that the class of algorithms based solely on the 19 and 37 GHz vertically polarised channels display the smallest sensitivity to all three atmospheric parameters: total water vapour, wind speed and cloud liquid water. Finally, it is demonstrated that this method overcomes many problems associated with conventional weather filtering over mixed ice-water and new-ice pixels and allows the retrieval of sea ice concentrations below 10%.     

Passive microwave algorithms for sea ice concentration: A comparison of two techniques

The most comprehensive large-scale characterization of the global sea ice cover so far has been provided by satellite passive microwave data. Accurate retrieval of ice concentrations from these data is important because of the sensitivity of surface flux (e.g., heat, salt, and water) calculations to small changes in the amount of open water (leads and polynyas) within the polar ice packs. Two algorithms that have been used for deriving ice concentrations from multichannel data are compared. One is the NASA Team algorithm and the other is the Bootstrap algorithm, both of which were developed at NASA's Goddard Space Flight Center. The two algorithms use different channel combinations, reference brightness temperatures, weather filters, and techniques. Analyses are made to evaluate the sensitivity of algorithm results to variations of emissivity and temperature with space and time. To assess the difference in the performance of the two algorithms, analyses were performed with data from both hemispheres and for all seasons. The results show only small differences in the central Arctic in winter but larger disagreements in the seasonal regions and in summer. In some areas in the Antarctic, the Bootstrap technique shows ice concentrations higher than those of the Team algorithm by as much as 25%; whereas, in other areas, it shows ice concentrations lower by as much as 30%. The differences in the results are caused by temperature effects, emissivity effects, and tie point differences. The Team and the Bootstrap results were compared with available Landsat, advanced very high resolution radiometer (AVHRR) and synthetic aperture radar (SAR) data. AVHRR, Landsat, and SAR data sets all yield higher concentrations than the passive microwave algorithms. Inconsistencies among results suggest the need for further validation studies.     

Some active and passive microwave signatures of Antarctic sea ice from mid-winter to spring 1991

Antarctic sea ice is often covered by a deep snow layer which acts as an emitter and a scatterer to microwave radiation leading to possible misinterpretations of ice signatures, particularly at high frequencies. The algorithms for ice identification, based on the observations of the Special Sensor Microwave Imager, at 19GHz (vertical and horizontal polarizations) and 37Ghz (vertical polarization), have proven to be inefficient for distinguishing new and old ice over the Antarctic Ocean. At an equivalent resolution and analysed on a weekly basis, complementary information can be obtained from active microwave measurements provided, at 5·3GHz (vertical polarization), by the Active Microwave Instrument, the scatterometer of ERS–1. Based on data obtained from the end of August to the end of November 1991, during the austral winter and spring radar backscatter is analysed as a function of the incidence angle. At low incidence angles, the derivative of the backscatter is closely related to the water concentration as derived from passive radiometry, whereas, at high incidence angles, the backscatter is mainly due to ice, as the water contribution is strongly reduced. During the whole period, stable features are apparent on the images obtained from the backscattering coefficients at 50°. On those images, higher values characterize the marginal ice zone, the polynya areas and the advected ice within the Ross Sea. At high incidence angles, the strong signatures of deformed/ rough ice depart significantly from the information classically extracted from the radiometers, the brightness temperatures as well as the derived products, polarization, spectral gradient ratios and concentration. It is therefore possible to classify the Antarctic ice cover into geographical clusters where the active microwave signatures can be attributed. to a peculiar ice type. Though those clusters are not totally identified, their stability and the coherence of their patterns show that they are related to geophysical structures. Four backscatter curves, simulating distinct behaviours over the Antarctic region, are proposed for sea water, marginal ice, first-year ice of the inner part of the pack and multi-year ice.     

Impact of intermediate ice concentration training data on sea ice concentration estimates from a convolutional neural network

Convolutional neural networks (CNNs) are being used increasingly for classification and regression tasks in remote sensing. However, a challenge in the remote sensing field is having a sufficient quantity of data to train a CNN, in particular, for conditions that are observed less frequently. In this study, a CNN is trained and tested to estimate sea ice concentration from synthetic aperture radar imagery (SAR). We first investigate the importance of including samples of intermediate ice concentration in the training data, which correspond to samples from the marginal ice zone (MIZ). In the present study, ice concentration from image analysis charts is used to provide training data labels. The MIZ for these ice charts are believed to be less accurate than those of the high and low concentrations, but nevertheless, our results support including these samples in the training data set. Additional experiments are then carried out increasing the number of MIZ sample, both from similar regions and a different ice region. It is found the MIZ is represented best in the test data when more samples from a similar region are included. Overall the results improve upon earlier studies, increasing the classification accuracy of the MIZ from 0.66 to 0.74.     

Correlation of precipitation estimates from spaceborne passive microwave sensors and weather radar imagery for BALTEX PIDCAP

This paper describes the evaluation of a combined radar and passive microwave dataset obtained during the PIDCAP study of the Baltic Sea Experiment (BALTEX), where three-dimensional volumes of data from the Gotland radar were obtained timed according to the overpasses of the DMSP-satellites F10 and F13. Both satellites are equipped with a Special Sensor Microwave/Imager (SSM/I), suitable for precipitation retrievals. We compare radar precipitation estimates, convolved to the native resolution of the SSM/I, at different altitudes with polarization and scattering indices ( S 85 ) derived from the SSM/I. For all 22 overpasses investigated here radar precipitation estimates at 3-4 km altitude correlate well with the SSM/I-derived S 85 (average correlation coefficient = 0.70). Although more directly linked to surface precipitation, polarization indices have been found to be less correlated with radar data, due to limitations inherent in the remote sensing of precipitation at higher latitudes. A stratification of the dataset into frontal and convective events revealed significant variations in these relationships for different types of precipitation events, thus reflecting different cloud microphysical processes associated with precipitation initialization. The relationship between S 85 and radar rain estimates at higher altitudes varies considerably for different convective and frontal events. The sensitivity of S 85 to radar-derived rain rate ranges from 3.1 K mm m 1 h m 1 for a strong convective event to about 25 K mm m 1 h m 1 for the frontal and about 70 mm m 1 h m 1 for the small-scale convective events. For extrapolated surface precipitation estimates, sensitivities decrease to 14 mm m 1 h m 1 and 25 mm m 1 h m 1 for frontal and small-scale convective precipitation, respectively.     

Development of a winter snow water equivalent algorithm using in situ passive microwave radiometry over snow-covered first-year sea ice

A snow water equivalent (SWE) algorithm has been developed for thin and thick snow using both in situ microwave measurements and snow thermophysical properties, collected over landfast snow covered first-year sea ice during the Canadian Arctic Shelf Exchange Study (CASES) overwintering mission from December 2003 to May 2004. Results showed that the behavior of brightness temperatures (T_bs) in thin snow covers was very different from those in a thick snowpack. Microwave SWE retrievals using the combination of T_b 19 GHz and air temperature (multiple regression) over thick snow are quite accurate, and showed very good agreement with the physical data (R² = 0.94) especially during the cooling period (i.e., from freeze up to the minimum air temperature recorded) where the snow is dry and cold. Thin snow SWE predictions also showed fairly good agreement with field data (R² = 0.70) during the cold season. The differences between retrieved and in situ SWE for both thin and thick snow cover are mainly attributable to the variations in air temperature, snow wetness and spatial heterogeneity in snow thickness.     

The use of a microwave backscatter model for retrieving soil moisture over bare soil

Abstract

Most attempts at predicting soil moisture from C-band microwave backscattering coefficients for bare soil are made by fitting experimental calibration relations obtained for limited ranges of incidence angle and soil surface roughness. In this paper, a more general approach is discussed using an inversion procedure to extend the use of a single experimental calibration relation to a wider range of incidence angle and surface roughness. A correcting function is proposed to normalize the backscattering coefficients to the conditions (incidence angle and surface roughness) of the calibration relation. This correcting function was derived from simulated data using the physical optics or KirchhofTs scatter model using the scalar approximation. Before discussing the inversion procedure, the backscattering coefficients calculated by the model have been compared with experimental data measured in the C-band, HH polarization and three incidence angles (Θ= 15°, 23°, 50°) under a wide range of surface soil moisture conditions (0.02H_v  0.35cm³ cm^-3) and for a single quite smooth soil surface roughness (0–011 _s  OOI4/n)m. The model was found to be experimentally validated from 15° to 23° of incidence and for surface soil moistures higher than 0-I0cm³cm^-3. For the inversion procedure, it is assumed to have a wider range of validity (15°  Θ 35° ) for ihc incidence angle. A sensitivity analysis of the model to errors on roughness parameter and incidence angle was performed in order to assess the feasability and suitability of the described inversion procedure.     

The continuum of operating modes for a passive model neuron.

Whether cortical neurons act as coincidence detectors or temporal integrators has implications for the way in which the cortex encodes information--by average firing rate or by precise timing of action potentials. In this study, we examine temporal coding by a simple passive-membrane model neuron responding to a full spectrum of multisynaptic input patterns, from highly coincident to temporally dispersed. The temporal precision of the model's action potentials varies continuously along the spectrum, depends very little on the number of synaptic inputs, and is shown to be tightly correlated with the mean slope of the membrane potential preceding the output spikes. These results are shown to be largely independent of the size of postsynaptic potentials, of random background synaptic activity, and of shape of the correlated multisynaptic input pattern. An experimental test involving membrane potential slope is suggested to help determine the basic operating mode of an observed cortical neuron.     

A heuristic for retrieving containers from a yard

This paper presents a three-phase heuristic to solve for an optimized working plan for a crane to retrieve all the containers from a given yard according to a given order. The optimization goal is to minimize the number of container movements, as well as the crane's working time. After generating an initial feasible movement sequence, the second phase reduces the length of the sequence by repeatedly formulating and generating a binary integer program. With another mixed integer program, phase three reduces the crane's working time by adjusting the movement sequence through iterations. Numerical testing results show that the heuristic is able to solve instances with more than 700 containers, which is within the range of real-world applications. Moreover, the number of movements approaches the lower bound in most cases, and the resulting movement sequence is efficient.     

A query language for retrieving information from a soil data bank

The paper presents specifications and implementation details of a query language designed for retrieving information from a soil data bank. The commands of the language are based on operations of relational algebra, and can be employed without previous programming experience. The language is part of the ARSIS (A Relational Soil Information System) system that is being developed in Greece.     

Passive microwave emission from smooth bare soils: Developing a simple model to predict near surface water content

MICRO-SWEAT is a coupled microwave emission and soilvegetation-atmosphere transfer scheme. It can predict, among other things, the L band (21 cm wavelength) microwave brightness temperature ( T B ) for a smooth bare soil from knowledge of the soil hydraulic properties and local meteorological data. T B is the product of the apparent emissivity ( e app ) and the effective physical temperature of the soil profile ( T eff ). The modelled relationship between e app and average 0-2 cm soil water content was used to develop a simple semi-empirical model, which predicts the near surface water content from knowledge of the particle size distribution and e app . T eff was found to be comparable to the soil temperature at 11 cm depth, and was estimated using a simple sinusoidal function based on thermal diffusivity to extrapolate the temperature at 11 cm depth from knowledge of the diurnal variation in air temperature. The dependence of T B on incidence angle was predicted by MICRO-SWEAT and also incorporated. This simple model was then successfully used to predict the 0-2 cm water content from knowledge of T B, the air temperature regime and the particle size distribution, for data collected at various look angles over three different soils using a truck-based radiometer.     

A passive movement method for parameter estimation of a musculo-skeletal arm model incorporating a modified hill muscle model

In this paper we present an experimental method of parameterising the passive mechanical characteristics of the bicep and tricep muscles in vivo, by fitting the dynamics of a two muscle arm model incorporating anatomically meaningful and structurally identifiable modified Hill muscle models to measured elbow movements. Measurements of the passive flexion and extension of the elbow joint were obtained using 3D motion capture, from which the elbow angle trajectories were determined and used to obtain the spring constants and damping coefficients in the model through parameter estimation. Four healthy subjects were used in the experiments. Anatomical lengths and moment of inertia values of the subjects were determined by direct measurement and calculation. There was good reproducibility in the measured arm movement between trials, and similar joint angle trajectory characteristics were seen between subjects. Each subject had their own set of fitted parameter values determined and the results showed good agreement between measured and simulated data. The average fitted muscle parallel spring constant across all subjects was 143 N/m and the average fitted muscle parallel damping constant was 1.73 Ns/m. The passive movement method was proven to be successful, and can be applied to other joints in the human body, where muscles with similar actions are grouped together.     

A theory of wave scatter from an inhomogeneous medium with a slightly rough boundary and its application to sea ice

An analytical theory of electromagnetic wave scattering from an inhomogeneous medium with a slightly rough boundary surface is formulated. The inhomogeneity in the medium is assumed to vary continuously in the vertical direction. In addition, it is also assumed to have a small random variation in the horizontal direction. The medium is assumed to consist of two layers. Maxwell's equations are solved by using the small perturbation method together with Fourier transform technique. The resulting differential equations are solved by using WKB and variation of parameter methods. Field amplitudes in each medium are determined by taking boundary conditions into account. The expressions for first order polarized radar backscatter cross-section δ⁰ are obtained. An attempt is made to apply the developed theory to compute sea ice scatter. The complex permittivity of sea ice, which depends on both the temperature and salinity, varies with the depth of sea ice. In addition, there is certainly some variation in the horizontal direction. Thus, the developed model may be able to give useful estimates when applied to sea ice scattering. Numerical calculations are performed for polarized radar backscatter cross-section ($\text{σ}{}_{\mathrm{<mtext>vv</mtext>}}{}^{\mathrm{<mtext>0</mtext>}}\text{and}\text{σ}{}_{\mathrm{<mtext>HH</mtext>}}{}^{\mathrm{<mtext>O</mtext>}}$) at two frequencies, 13.3 GHz and 400 MHz. It can be shown that WKB method is applicable at both of these frequencies. These theoretical results are compared with the experimental results obtained from NASA Earth Resources Program mission 126. Theoretical results give the same absolute value of σ^O and the relative variation among the six ice types as is given by the experimental results.     

A microwave propagation model for estimation of effective attenuation coefficients in a vegetation canopy

This article presents a detailed theoretical model for propagation of microwave in a layer of vegetation medium. The vegetation medium is modeled by a layer of randomly distributed dielectric circular disks and cylinders, representing the leaves and the stems, respectively. Propagation of wave in such a medium is taken as a transport of energy problem, and is formulated by a Monte Carlo method. Interactions of wave with the vegetation components are treated as a sequence of multiple scattering events between the photon flux and the scatterers. A Monte Carlo algorithm is used to track these collision processes, and the energy of the photon leaving the vegetation layer is collected. With sufficient number of photon histories, the ensemble averages of the photon energy are used to calculate the attenuation coefficients of the wave through the vegetation layer. Our simulation results indicate that attenuation coefficients calculated using this multiple scattering model may differ from those given by a simple single scattering model. Effects on the attenuation coefficients by various parameters such as moisture content, scatterer's size, and volume fraction are studied. Comparisons of the calculated results are made with the measured data and good match is obtained.     

Estimates of sea surface nitrate concentrations from sea surface temperature and chlorophyll concentration in upwelling areas: A case study for the Benguela system

The knowledge of nitrate fields at global or regional scales in the ocean is fundamental for the study of oceanic biogeochemical processes, particularly those linked to new primary production. The estimate of nitrate concentrations from space is generally based on empirical inverse relationships between sea surface temperature (SST) and nitrate concentrations. These relationships, however, are often highly variable spatially and temporally, and hardly applicable to large areas (i.e., larger than a few degrees in latitude). In this paper we propose a new approach specifically developed for areas influenced by upwelling processes. It relates the nitrate concentration to the difference between SST and the estimated temperature of the upwelled water (variable with latitude and season), δT, which is an indicator of the time elapsed since upwelling. This approach is tested for the Benguela upwelling system, and algorithms are developed using in situ data provided by the World Ocean Database 2005 of the NOAA-NESDIS-National Oceanographic Data Center. The results reveal a significant improvement compared to the NO₃-SST relationships, and a single algorithm can be applied to the whole upwelling area (15 to 35°S). Further improvement is gained by coupling this approach with a method that derives sea surface nitrate concentrations from SST and surface chlorophyll a concentration using multiple regression analyses, as proposed by Goes et al. [Goes, Saino, Oaku, Jiang, (1999). Method for estimating sea surface nitrate concentrations from remotely sensed SST and chlorophyll a: A case study for the North Pacific Ocean using OCTS/ADEOS data. IEEE Transactions on Geoscience and Remote Sensing, 37, no. 3 II, 1633-1644].     

A model for an intelligent operating system for executing image understanding tasks on a reconfigurable parallel architecture

Parallel processing is one approach to achieving the large computational processing capabilities required by many real-time computing tasks. One of the problems that must be addressed in the use of reconfigurable multiprocessor systems is matching the architecture configuration to the algorithms to be executed. This paper presents a conceptual model that explores the potential of artificial intelligence tools, specifically expert systems, to design an Intelligent Operating System for multiprocessor systems. The target task is the implementation of image understanding systems on multiprocessor architectures. PASM is used as an example multiprocessor. The Intelligent Operating System concepts developed here could also be used to address other problems requiring real-time processing. An example image understanding task is presented to illustrate the concept of intelligent scheduling by the Intelligent Operating System. Also considered is the use of the conceptual model when developing an image understanding system in order to test different strategies for choosing algorithms, imposing execution order constraints, and integrating results from various algorithms.     

A downscaling method for distributing surface soil moisture within a microwave pixel: Application to the Monsoon '90 data

A downscaling method for the near-surface soil moisture retrieved from passive microwave sensors is applied to the PBMR data collected during the Monsoon '90 experiment. The downscaling method requires (1) the coarse resolution microwave observations, (2) the fine-scale distribution of soil temperature and (3) the fine-scale distribution of surface conditions composed of atmospheric forcing and the parameters involved in the modeling of land surface-atmosphere interactions. During the Monsoon '90 experiment, eight ground-based meteorological and flux stations were operating over the 150 km² study area simultaneously with the acquisition of the aircraft-based L-band PBMR data. The heterogeneous scene is hence composed of eight subpixels and the microwave pixel is generated by aggregating the microwave emission of all sites. The results indicate a good agreement between the downscaled and ground-based soil moisture as long as the intensity of solar radiation is sufficiently high to use the soil temperature as a tracer of the spatial variability of near-surface soil moisture.     

A non-stationary MRF model for image segmentation from a soft boundary map

In this paper, we address the problem of estimating a segmentation map into regions from a soft (or possibly probabilistic) boundary representation. For this purpose, we have defined a simple non-stationary MRF model with long-range pairwise interactions whose potentials are estimated from the likelihood of the presence of an edge at each considered pair of pixels. Another contribution of this paper is also to demonstrate that an efficient and interesting alternative strategy to complex and elaborate region-based segmentation models consists in averaging several (quickly estimated) soft contour maps (obtained, for example, with simpler edge-based segmentation models) and to use this MRF reconstruction model to achieve a reliable and accurate segmentation map into regions. This model has been successfully applied on the Berkeley image database. The experiments reported in this paper demonstrate that the proposed segmentation model from an edge map is reliable and that this segmentation strategy is efficient (in terms of visual evaluation and quantitative performance measures) and performs well compared with the best existing state-of-the-art segmentation methods recently proposed in the literature.     

A model for deployment of a freefall lifeboat from a moving ramp into waves

The equations of motion of a freefall lifeboat are formulated using Kane’s method. Deployment from a moving ramp is assumed to occur in a known plane and a 2D model suffices. A segmented approach is used to model ramp contact. A separate 3D model is presented for the water entry phase. The hydrodynamic loads at water entry are also modeled using a segmented approach. The equations are solved numerically using a standard Runge–Kutta MATLAB routine.