Search

Polar sea ice plays a critical role in regulating the global climate. Seasonal variation in sea ice extent, however, coupled with the difficulties associated with in situ observations of polar sea ice, makes remote sensing the only practical way to estimate this important climatic variable on the space and time scales required. Unfortunately, accurate retrieval of sea ice extent from satellite data is a difficult task. Sea ice and high cold clouds have similar visible reflectance, but some other types of clouds can appear darker than sea ice. Moreover, strong atmospheric inversions and isothermal structures, both common in winter at some polar locations, further complicate the classification. This paper uses a combination of feed-forward neural networks and 1.6-[micro]m data from the new Chinese Fengyun-1C satellite to mitigate these difficulties. The 1.6-[micro]m data are especially useful for detecting illuminated water clouds in polar regions because 1) at 1.6 [micro]m, the reflectance of water droplets is significantly higher than that of snow or ice and 2) 1.6-[micro]m data are unaffected by atmospheric inversions. Validation data confirm the accuracy of the new classification technique. Application to other sensors with 1.6-[micro]m capabilities also is discussed. Index Terms--Arctic, climate, neural networks, remote sensing, sea ice.

Accurate detection of areal extent of snow in mountainous regions is important. Areal extent of snow is a useful climatic indicator. Moreover, snow melt is a major source of water supply for many arid regions (e.g., western United States, Morocco) and affects regional ecosystems. Unfortunately, accurate satellite retrievals of areal extent of snow have been difficult to achieve. Two approaches to effectively and accurately detect clear land, cloud, and areal extent of snow in satellite data are developed. A feed-forward neural network (FFNN) is used to classify individual images, and a recurrent NN is used to classify sequences of images. The continuous outputs of the NN, combined with a linear mixing model, provide support for mixed-pixel classification. Validation with independent in situ data confirms the classification accuracy (94% for feed-forward NN, 97% for recurrent NN). The combination of rapid temporal sampling (e.g., GOES) and a recurrent NN classifier is recommended (relative to an isolated scene (e.g., AVHRR) and a feed-forward NN classifier). Index Terms--Areal extent of snow, neural network, snow hydrology.

A new hybrid method for automatic clustering of satellite-observed natural scenes is presented. It uses a partitional clustering algorithm augmented by a hierarchical split-and-merge step at each iteration. The method also dynamically computes image-specific split-and-merge thresholds and can accommodate arbitrary information vectors. Better partitioning of the data and improved computational efficiency are achieved. The sensitivity of the method to the clustering parameters is presented, and the results show that the method is relatively insensitive to the choice of clustering parameters. Comparisons with the often used ISODATA algorithm show the effectiveness of the new approach. Index Terms--Clustering, image classification, remote sensing.

A Beowulf-type cluster can: (1) mitigate many issues associated with the analysis of large, complex remotely sensed data sets; (2) shorten the response time of operational agencies to crisis-management situations; and (3) expedite the reanalysis of large archives of satellite data. Whereas most Beowulf-type designs support modeling applications, the Parallel Image Processing Environment (PIPE) addresses the unique requirements of remote sensing applications. PIPE has four hierarchical layers: hardware, operating system, middleware and applications. Rocks, a middleware sublayer, manages the cluster. DIAL-developed interprocess communication and control daemons form the second middleware sublayer. They encapsulate user-defined applications and thereby support automated, user-transparent parallelization of satellite data analyses, implemented in the applications layer using generalized constructs. The daemons also monitor resource (computational and I/O) utilization on a node/thread basis, a feature not supported by other generally available monitoring utilities. The application support libraries are fully extensible, facilitate the reuse of modular and commonly used software functions in new applications and thereby reduce both the cost and time to implement new applications. Two applications (signal analysis, image classification) show PIPE's versatility and performance characteristics. PIPE is intrinsically scalable, reliable and can be incrementally implemented. A comparison with other embarrassingly parallel systems is also provided. Copyright © 2006 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

Polar sea ice plays a critical role in regulating the global climate. Seasonal variation in sea ice extent, however, coupled with the difficulties associated with in situ observations of polar sea ice, makes remote sensing the only practical way to estimate this important climatic variable on the space and time scales required. Unfortunately, accurate retrieval of sea ice extent from satellite data is a difficult task. Sea ice and high cold clouds have similar visible reflectance, but some other types of clouds can appear darker than sea ice. Moreover, strong atmospheric inversions and isothermal structures, both common in winter at some polar locations, further complicate the classification. This paper uses a combination of feed-forward neural networks and 1.6-µm data from the new Chinese Fengyun-1C satellite to mitigate these difficulties. The 1.6-µm data are especially useful for detecting illuminated water clouds in polar regions because 1) at 1.6 µm, the reflectance of water droplets is significantly higher than that of snow or ice and 2) 1.6-µm data are unaffected by atmospheric inversions. Validation data confirm the accuracy of the new classification technique. Application to other sensors with 1.6-µm capabilities also is discussed. [ABSTRACT FROM AUTHOR]