280 



Fishery Bulletin 98(2) 



from aircraft are currently used in several fisheries as 

 indices of the abundance of small pelagic fishes and 

 a lidar-based system would have several advantages 

 over these passive methods. Our computation with a 

 deterministic model showed that a lidar survey may 

 be about twice as efficient in detecting schools as a 

 vision-based system during the night and five times 

 more efficient during the day. At night, a lidar will 

 detect more schools than an observer, but the differ- 

 ence is not huge because the very wide swath width 

 (1600 m) of our hypothetical aerial obsei-ver compen- 

 sated, to some degree, for the observer not seeing 

 farther beneath the surface. During the day, the effi- 

 ciency of lidar detection, in contrast to visual detec- 

 tion, increases greatly because schools inhabit deeper 

 water. In addition to increased detection efficiency, 

 lidar has several other advantages over aerial observ- 

 ers: lidar images can be better quantified than those 

 based on visual detection or cameras because the 

 school volume rather than school area can be esti- 

 mated, thereby improving the precision of the index; 

 in addition, detection is less dependent on sea state 

 and is little affected by sun angle or moon phases. On 

 the other hand, skilled fishermen working in aerial 

 surveys can identify species of schooling fish with 

 remarkable accuracy; a remote species identification 

 algorithm for a lidar will be difficult, if not impossible, 

 to develop. 



As with hydroacoustic methods, species identifica- 

 tion with lidar is a major concern. Even after 50 years 

 of hydroacoustic research, the only method for iden- 

 tifying acoustic targets with certainty is by securing 

 voucher specimens. Radiometric backscatter has no 

 magical properties in relation to those of acoustic back 

 scatter that might allow a rapid solution to the prob- 

 lem of species identificaton. The lesson learned from 

 hydroacoustics is that for species identification to be 

 a reality in lidar surveys, additional sensing systems 

 will be needed. That skillful humans make accurate 

 species identifications visually provides the hope that 

 species recognition algorithms eventually will be prac- 

 tical. We believe it will be possible over the long teiTn to 

 develop species recognition algorithms for lidar in com- 

 bination with advanced lidar signal processing, digi- 

 tal video cameras, and local knowledge, but at present 

 species identifications must depend upon combining 

 lidar survey data with other information. From the 

 lidar data, we could distinguish reliably between small 

 (about 30 cm length) and large (about 1-m) fish. Iden- 

 tification of intermediate lengths may become possible 

 with more practical experience. One possible approach 

 for obtaining additional information is to use visual 

 identifications of fish schools by aerial observers pro- 

 rated to lidar targets. Other possible approaches are to 

 combine airborne lidar survey with a research trawler 



that can provide voucher specimens or to combine air- 

 borne lidar with simultaneous sampling of fish eggs 

 from a research vessel (Checkley et al., 1997). The 

 latter approach has been used successfully in a test of 

 the NOAA lidar (Chumside, 1999). 



Future application of airborne lidar 



An airborne lidar survey could provide a census of 

 epipelagic fishes an order of magnitude faster than 

 that provided by ships, thus reducing costs in dol- 

 lars (based on 1999 dollar amounts) from about $100 

 per ship-survey mile to $3 per aerial-survey mile 

 (research ship cost=$12,000 per day, net ship speed 

 including stopping at stations=5 kn; airplane=$600 

 per hour at 200 kn). 



Faster surveys not only cost less but improve accu- 

 racy because steady state assumptions are reduced, 

 vessel avoidance is eliminated, and, most important, 

 high speed makes it practical to survey a much larger 

 area, thereby eliminating the errors associated with 

 partial coverage. No major technical barrier exists in 

 acquiring a suitable instrument; adequate fish detec- 

 tion lidars already exist. Fish-detecting lidars may 

 be purchased from one or more vendors or a radio- 

 metric lidar may be assembled from "off the shelf 

 components as has been done with the NOAA lidar 

 (Churnside and Hunter, 1996). However, to imple- 

 ment routine surveys, signal-processing algorithms 

 for rapid quantification of targets are needed, and if 

 the fish targets are to be converted to biomass, direct 

 calibrations of target strength will be needed. 



The depth limitation of lidar is not a major bar- 

 rier to implementation. Our analysis demonstrates, 

 as does our practical experience, that school detec- 

 tion depths of 30-40 m can be expected for California 

 coastal waters using off-the-shelf instrumentation. 

 In fact, more powerful systems are unlikely to do 

 much better owing to the rapid attenuation of signal 

 with depth. The 30-40 m depth limitation is less 

 important at night because most epipelagic fish 

 schools are found within the volume of water that is 

 to be detected by lidar. Our analysis demonstrated 

 that schools can be detected at night despite a much 

 lower packing density. To deal most effectively with 

 the fraction of undetected schools, survey design 

 should be based on line transect theory and should 

 require an estimate of the average vertical distribu- 

 tion of schools under the specific survey conditions 

 (region, species, season, time of day). 



In conclusion, the census of epipelagic fish schools 

 with airborne lidar would be practical and useful 

 today if three conditions could be met: assumptions 

 regarding species identity are acceptable; a line 

 transect survey design is used in conjunction with 



