Whereas the Joint Typhoon Warning Center (JTWC) has ten track forecasts to 72 h, only four dynamical model forecasts are available at 96 h and 120 h. Forming a selective consensus (SCON) by proper removal of a likely erroneous track forecast is hypothesized to be more accurate than the non-selective consensus (NCON) of all four models. Conceptual models describing large track error mechanisms, which are related to known tropical cyclone motion processes being misrepresented in the dynamical fields, are applied to forecasts by the Navy Operational Global Atmospheric Prediction System (NOGAPS), U.S. Navy version of the Geophysical Fluid Dynamics Laboratory Model (GFDN), United Kingdom Meteorological Office (UKMO), and National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) during the 2005 western North Pacific typhoon season. A systematic error in the GFDN was identified in which the model built a false anticyclone downstream of the Tibetan Plateau, which explained over 50% of the large GFDN track errors. In the GFS model, 95% of the large errors occurred due to an incorrect depiction of the vertical structure of the tropical cyclone. The majority of NOGAPS and UKMO large errors were caused by an incorrect depiction of the midlatitude system evolutions. Characteristics of the erroneous forecast tracks and corresponding model fields are documented and illustrative case studies are presented. By applying rules of the Systematic Approach, the average SCON error was 222 n mi (382 n mi) less than NCON (JTWC) in 20% of all 120-h forecasts. / US Air Force (USAF) author.

The objective of this study was to enable the removal of developer depletion and diffusion effects (i.e., adjacency effects from noisy photographic images, thus, providing a potential for improvement in the reliability of restored object and aerial image estimates. The investigation was based on the use of a previously formulated image resortation model which characterized the exposure, latent image and development interactions of the photographic process in terms of statistical estimation theory. This study addressed the application and appropriate modification of the formulated model in the removal of adjacency effects from noisy images of selected line targets. Literature pertaining to the initial observations of adjacency effects, their recognition as a nonlinear chemical development effect and the pertinent models (forward) used to predict their effects was reviewed. This was followed with a review of the statistical restoration (inverse) model and its comparison to previously derived forward models. X-ray quanta exposures were then used to obtain noisy photographic images, free of optical scattering effects, for the purpose of empirically determining a chemical spread function to characterize chemical adjacency effects. Photographic images were obtained that contained lines of 0.010, 0.100 and 1.000 mm widths to enable comparison between the magnitude of the chemical spread function and the Eberhard effect. A segmented polynomial (cubic spline function) approach was used to calculate the chemical spread function. Separate light quanta exposures were used to obtain gross grain density sensitometric curves and noisy line images. The covering power relationship between mass of developed silver and diffuse density was empirically derived for Panatomic-X film processed without agitation in D-76 developer (diluted 1:1). Emperical verification of the statistical restoration model was achieved. Chemical adjacency effects were successfully removed from noisy line images using an appropriately scaled version of the statistical restoration model. The spatial frequency content of the noisy line images was approximately 1, 10, 15, 20, 25 and 40 cycles/mm. The proportionality factor, used to scale the chemical spread function required in the restoration model, was found to be equivalent to the ratio of the empirically derived chemical spread function and a magnitude estimate of the Eberhard effect. The maximum diffuse density correction for edge effects was found to be 0.14, or approximately 11.1% of the gross grain density level, 1.20. Similar diffuse density corrections for fine line images were found to range between 0.14 and 0.36, or approximately 11.1% to 28.3% of the gross grain density level associated with a specific line element.

Improvements have been made to the spatial coherence method for automatically determining cloud-free ocean radiances in satellite imagery by incorporating the spectral signatures of reflecting surfaces. The spatial coherence method relies on the fact that small-scale cloud-free regions typically exhibit uniform emission and uniform reflection. While small-scale overcast regions typically exhibit uniform emission, they often exhibit considerable variability in reflectance. On rare occasions the requirements of spatial uniformity are not met and errors are produced in estimated cloud-free radiances. The frequency of errors in identification of cloud-free and overcast pixels was assessed using two years of Advanced Very-High Resolution Radiometer (AVHRR) data from six regions of the globe. Significant improvement in the identification of cloud-free radiances is obtained by including a test of Q, the ratio of the AVHRR channel 2 (0.83-μm) reflectance to channel 1 (0.63-μm) reflectance. Q varies depending on whether the reflecting surface is cloud-free ocean, cloud-free land, or overcast by clouds. A study was conducted to determine the dependence of Q for overcast pixels on changes in season and geography. While some variation is evident due to satellite viewing angle and differences in atmospheric water vapor content, these effects are sufficiently small that constant thresholds may be used to help separate cloud-free and overcast pixels. The modified spatial coherence method uses the threshold for Q and radiance uniformity thresholds at 0.63-μm and 11-μm to identify cloud-free and overcast pixels. A sensitivity study was performed to determine the dependence of cloud-free ocean radiance estimates on the values of the uniformity thresholds. The results of the study indicate that using thresholds of 0.5% for the 0.63-μm reflectivity and 0.5 mWm⁻²sr⁻¹cm for the 11-μm radiance, produces cloud-free radiances that are rarely biased by more than 0.4% for reflectances at 0.63 μm and 0.4 K for the 11-μm brightness temperature. The uniformity and Q thresholds may be used for a large variety of scenes from different seasons and geographic areas. / Graduation date: 2000

Thesis (M.S. in Meteorology and Physical Oceanography)--Naval Postgraduate School, September 2003. / Thesis advisor(s): Philip A. Durkee, Carlyle H. Wash. Includes bibliographical references (p. 49-51). Also available online.

Thesis (M.S. in Meteorology and Physical Oceanography)--Naval Postgraduate School, Dec. 2004. / Thesis Advisor(s): Philip A. Durkee. Includes bibliographical references (p. 35-36). Also available online.

No description available.

One of the problems associated with the automatic image processing of satellite photographs such as weather maps is the need for image registration; that is, the fitting of a map that has some translational and rotational bias to a known data base. This paper investigates a least square method of image registration using an image that has been converted into a boundary map with a pixel representation 1 for land, -1 for water and zero for cloud pixels. A sampled correlation array is constructed by shifting the weather map to locations on a given grid, centered around a sampled correlation peak, and performing an accumulation of the pixel-by-pixel comparisons between the weather map and its data base over the whole map or a smaller search window. A least square approximation 0 f the translational and rotational bias is performed using the data from this sampled correlation array, fitted against the shape of an elliptical cone.

The requirement for a fast automated correlation algorithm for registration of satellite images is discussed. An overview of current registration techniques is presented indicating a correlator, matching binary maps compressed from the original imagery, may provide the required throughput when implemented with a dedicated hardware/processor. An actual registration problem utilizing GOES digitally processed imagery is chosen and defined. The realization of a fast correlator, matching image input data with sampled data base reference image data in real time is considered.

Approved for public release, distribution is unlimited / This thesis assesses the performance of the Naval Postgraduate School Aerosol Optical Depth (NPS AOD) model utilizing very high spatial resolution QuickBird (QB) satellite data. QuickBird derived AOD results are compared to other satellite and ground based AOD results, specifically, AErosol RObotic NETwork (AERONET), MODerate resolution Imaging Spectroradiometer (MODIS), Multi-angle Imaging SpectroRadiometer (MISR), and Advanced Very High Resolution Radiometer (AVHRR). Data is collected around Sir Bu Nuair Island, United Arab Emirates in September 2004 as part of the UAE2 Campaign. Satellite measured radiances are calibrated and due to spatial resolution differences between sensors, modal radiances are calculated for areas matching the highest resolution sensor. The AOD model is based on AVHRR wavelengths; hence, the modal satellite measured radiances are linearly extrapolated to the effective wavelengths of AVHRR. Results show application of the NPS AOD model to QuickBird data yields findings that are consistent with other satellite and ground based retrievals. In general, the NPS AOD model works well for nadir and near-nadir view angles, but not for high zenith angles. / Lieutenant Commander, United States Navy

Meteorological satellite images at visible and infra-red wavelengths are an invaluable source of information on cloud systems because of their extensive coverage of the whole of the Earth's surface, providing data in areas that are only sparsely monitored, if at all, by other means. Although this information has been used subjectively by forecasters for many years, the lack of automatic, quantitative analysis techniques largely prevents its assimilation into numerical weather prediction (NWP) models, which are the basis of all modern weather forecasting. This thesis investigates the use of neural network techniques for the analysis of the images in order to make fuller use of the available data. The recognition of a particular type of cloud is dependent on the determination of a set of features from the satellite image spectral bands that will give discriminating information. This feature extraction and selection process is dealt with in detail, and a feature selection process based on the radial basis function (RBF) neural network is presented. The high-dimensional feature space is visualized on a two-dimensional plane by the use of three techniques: the Kohonen map, the Sammon map, and a recently-developed technique known as the Generative Topographic Mapping (GTM). Classification results using a multi-layer perceptron (MLP) and an RBF neural network are presented. The results of independently classifying each pixel in an image are compared with a method that makes use of contextual information, the Markov Random Field (MRF) model. The limitations of the pixel-based approach are highlighted, and a region-based approach is presented that enables the definition of large-scale regions of uniform cloud type. Two segmentation methods are used, the active contour (or snake) model, and the more recentlydeveloped level set technique. The latter method was found to provide many benefits over the former. The region-based approach will facilitate the assimilation of cloud system information into NWP models in the future.