ENVISAT's dual-polarization capability helps satellite-image analysts distinguish between 

 icebergs and ships (Lane et al., 2003). Icebergs tend to show up in the imagery only in HH mode, 

 whereas ships show up in both HH and HV. By looking at a target in both modes, one can 

 determine whether it is an iceberg or a ship. Figure 1 shows, in both HH and HV modes, an 

 example of a ship, which was confirmed by Provincial Aerospace Limited (PAL), an aerial 

 reconnaissance firm based in St. John's, Newfoundland. Figure 2 shows an example of an 

 iceberg in both HH and HV modes that was confirmed by C-CORE. Notice the ship appears in 

 both HH and HV modes, while the iceberg shows up only in the HH mode. 



In addition to visually comparing HH and HV imagery, C-CORE developed an iceberg-detection- 

 software (IDS) algorithm that classified both icebergs and ships from EN VIS AT imagery. The 

 data was sent to IIP in MANICE code but — because it needed further validation — not merged into 

 HP's BAPS database. However, IIP used the data for flight planning and decision making. 



Methods 



During the 2004 ice season, a cooperative experiment involving IIP, C-CORE, the Canadian Ice 

 Service (CIS), and PAL was conducted to validate the algorithm's ability to distinguish between 

 icebergs and ships. Ice Patrol and PAL conducted five under flights in April and May of 2004 

 and identified 101 icebergs and 41 ships. These results were compared to the MANICE output 

 generated by C-CORE. The MANICE output reported iceberg and ship positions from the 

 ENVISAT imagery based on the confidence of C-CORE's algorithm and imagery review. 

 Modeled environmental conditions from the U.S. Navy Fleet Numerical Meteorology and 

 Oceanography Center were used to compare on-scene wind speed and wave height during the 

 validation efforts. 



Because C-CORE needed a minimum of two weeks to order ENVISAT frames from the 

 European Space Agency, C-CORE requested that IIP indicate a satellite-reconnaissance area 14 

 days before the acquisition date. Images were obtained by C-CORE near real time, processed 

 through the IDS, quality controlled using the dual polarization modes, and sent to IIP in MANICE 

 code approximately three to four hours after satellite acquisition. Combined with few southern 

 icebergs and a late season opening, the required two-week lead time made planning satellite 

 acquisitions very difficult. Therefore, without regard to the IS mode, IIP acquired ENVISAT 

 frames in areas of the ocean where icebergs were predicted to be. Consequently, four of the five 

 validation flights were conducted over areas where the sensor was in the IS modes (4 and 5) not 

 optimal for iceberg detection (Table 2). 



Provincial Aerospace Limited and HP conducted validation flights to compare targets reported in 

 the C-CORE MANICE messages with confirmed target data. Ice Patrol conducted its flight at 

 patrol altitude to detect targets via radar and then descended to identify each target in the satellite- 

 acquisition swath. C-CORE, IIP, PAL, and CIS evaluated the data to determine the probability of 

 detection (POD) and probability of classification (POC) of the MANICE message produced by C- 

 CORE. Table 2 represents the data from the five validation flights correlated by C-CORE. 



Results 



Probability of detection is defined as the probability of C-CORE to report a target in MANICE 

 that was detected by aerial reconnaissance. Probability of classification is defined as the 



52 



