BY LAND OR AIR: EVALUATION OF WADING BIRD COLONY CENSUSING USING AERIAL AND GROUND TECHNIQUES.

1. BY LAND OR AIR: EVALUATION OF WADING BIRD COLONY CENSUSING USING AERIAL AND GROUND TECHNIQUES. M. Clay Green (claygreen@txstate.edu), Department of Biology, Texas State University; Margaret C. Luent (ms.sherman@juno.com), Department of Biology, University of Louisiana; Clinton W. Jeske (clint_jeske@usgs.gov), U.S. Geological Survey, National Wetlands Research Center; Thomas C. Michot (thomas_michot@usgs.gov), U.S. Geological Survey, National Wetlands Research Center; Paul L. Leberg (leberg@louisiana.edu), Department of Biology, University of Louisiana. RESULTS We surveyed 39, 38, and 23 active colonies by using fixed-wing aircraft, helicopter, and ground observers respectively; all active colonies surveyed from the ground were also surveyed from the air. The fixed-wing aircraft surveys estimated a significantly greater number of mean breeding pairs per colony than the ground estimates (t57 = 2.903, P = 0.005; Table 1). We based these differences only on colonies that were surveyed using both approaches in the same survey period. Estimates for white (t53 = 2.236, P = 0.029) and dark (t54 = 2.302, P = 0.025) species were also significantly greater from fixed-wing aircraft than ground counts (Table 1). Surveys from fixed-wing aircraft lacked a measure of precision; however, an estimate of precision is possible by assuming that the error in the ground counts is relatively small. The proportion of variation in fixed-wing counts explained by the estimates from the ground counts was 0.14. This estimate of precision was higher for estimates of white-plumaged birds than for dark-plumaged birds; however, estimates of R2 were low for all species except great egrets. For the comparison between ground and helicopter surveys, 17 colonies were estimated by both survey methods during May 2005. There was no significant difference in estimated mean number of breeding pairs per colony between helicopter and ground surveys (t16 = 0.330, P = 0.75; Table 2). Additionally, there was no significant difference between estimates of white (t16 = -0.128, P = 0.90) and dark species (t16 = 1.038, P = 0.32) or between individual species (Table 2). The regressions of estimates from the ground and helicopter surveys resulted in generally high estimates of R2 (Table 4). Estimates were >0.70 for total numbers of nesting birds as well as for both white- and dark-plumaged birds; there was little difference in this measure of precision based on plumage. For ground count comparisons, there was a significant difference in probability of detection among species (F1,786 = 19.4, P < 0.001, Table 3) and between observers (F9,786 = 3.4, P < 0.001). The average probability that the 2 ground observers would detect an individual nest was 0.796 and 0.706 for observers 1 and 2 respectively. The use of marked subsample method revealed that ground counts conducted by a single observer would result in an average underestimation of nesting pairs by 21.8 %. Abstract: Aerial surveys are often used to quantify the sizes of waterbird colonies; however these surveys would benefit from a better understanding of associated biases. We compared counts of breeding pairs of waterbirds, in colonies across southern Louisiana, made from the ground, fixed-wing aircraft and a helicopter. We used a marked-subsample method for ground counting colonies to obtain estimates of error and visibility bias. Estimates from fixed-wing aircraft were higher than ground counts for the overall estimated number of breeding pairs and for both dark and white-plumaged species. Precision of fixed-wing aircraft estimates were low, based on the assumption that ground counts are closer to the true count, and fixed-wing aerial surveys appeared to overestimate numbers of nesting birds of some species; this bias often increased with the size of the colony. Helicopter estimates appeared to be more precise than fixed-wing aircraft estimates as the numbers of nesting pairs made from ground and helicopter surveys were very similar for all species observed. The use of ground counts by a single observer resulted in underestimated number of breeding pairs by 20 % on average. Use of marked-subsample method provided an estimate of the number of missed nests as well as an estimate of precision. These estimates represent a major advantage of marked-subsample ground counts over aerial methods; however, ground counts are difficult in large or remote colonies. Helicopter surveys and ground counts provide less biased, more precise estimates of breeding pairs than surveys made from fixed-wing aircraft. The most appropriate combination of survey approaches will be based on the need for precise and unbiased estimates, balanced with financial and logistical constraints. METHODS We used fixed-wing aircraft, helicopter, and ground counts to survey a total of 60 mixed-species waterbird colonies in southern Louisiana during May and June of 2004 and 2005. We identified a priori these colonies based on the likely ability to conduct simultaneous ground counts and colonies’ adjacency to one another to minimize travel time thereby maximizing the number of colonies we could visit in a given day. A total of 15 species of colonial waterbirds were surveyed (families Anhingidae, Pelicanidae, Ardeidae, Threskiornithidae). Aerial surveys Fixed-wing aircraft surveys were conducted from a Cessna 185 (USGS), configured with a voice/global positioning system (GPS)/moving map system that links the GPS unit of the aircraft with onboard intercom system and onboard laptop computers. The rotary-wing aircraft survey was conducted from a Bell 206L (Southern Helicopters). Aerial surveys were conducted at approximately 150 m above ground level for both the helicopter and airplane surveys. Fixed-wing aircraft speeds during surveys averaged 157 km/hr while helicopter speeds range from 0 - 40 km/hr along elongated colonies (e.g., flooded treerows). We estimated the number of individual birds at each colony and assumed each individual represented a breeding pair. Ground counts using mark-subsample method We conducted ground counts on 23 waterbird colonies during May and June 2004-2005. Ground counts were conducted ± 10 days of matching aerial surveys to minimize changes in colony size and species composition. Two observers conducted ground counts for all colonies. Each observer counted the number of nests in each colony independently, however both observers conducted counts at the same time. Observers did not share findings during the ground surveys. We counted only occupied nests and assumed one nest equaled one breeding pair. For each nest counted, we plotted the nest location within the colony on a map. At completion of each count, we compared between observers the nest count, nest location, and species composition to determine which nests were counted by both observers. Statistical Analysis For both the helicopter and fixed-wing aerial surveys, we used unadjusted estimates of number of breeding pairs, of each species, to represent the size of each colony. For the ground counts using tandem observers (marked-subsample), we used Seber’s (1973) Peterson estimation to derive estimates of number of breeding pairs, by species, for each colony. The marked-subsample method also allowed us to estimate the probability of detection for each observer and estimate visibility bias (Magnusson et al. 1978, Caughley and Grice 1982, Choquenot 1995, Walter and Rusch 1997). For purposes of statistical analyses, we assumed that measurement errors of observations of the same colony made in different years or months were independent. We performed a series of paired t-tests to examine the null hypothesis that mean difference in the estimated number of breeding pairs for each species between survey methods was zero. We also examined differences between the mean numbers of species detected by each survey method using paired t-tests. We estimated bias for all significant differences between the ground counts and the two aerial survey methods. We assumed a priori that ground counts were closer to a true count than either of the aerial survey methods. Bias was calculated by subtracting the ground estimate from either aerial estimate and then dividing the product by the ground estimate. We used bisector regression because of the existence of measurement errors in both ground and aerial surveys. Bisector regression has been shown to perform significantly better than other regressions (e.g. reduced major-axis) that are commonly used when both X and Y axes contain errors (Isobe et al. 1990). We used the R2 of this relationship to determine the amount of variation in the aerial survey estimates that can be explained by the number of nesting pairs observed form the ground. A low R2 would suggest that the aerial surveys resulted in very imprecise estimates of the number of nesting pairs of a species. Using bisector regression, we determined if the slope of the relationship between the number of pairs estimated from ground counts and aerial surveys equaled 1 and the intercept equaled 0, the expectations if bias did not change with colony size. INTORUDCTION Conservation and management of wildlife species requires regular monitoring of populations to facilitate trend estimates and overall status of the species of interest (Caughley 1977, Sauer and Droege 1990). Trends are generally derived from multiple estimates of population size over a period of time. Estimations of population size are not easily conducted because many species exhibit morphological or behavioral traits that increase crypsis, thus rendering animals difficult to detect and count. Additionally, certain survey methods may be more appropriate for certain species; therefore, incorrect choice of survey methodologies for estimating population size for a given species may result in increased bias and reduced precision (Caughley 1977, Lancia et al. 1996). The development and application of biologically and statistically sound survey approaches are necessary to minimize biases thereby improving assessment of the status of populations. The coastal wetlands of Louisiana provide nesting sites for significant proportions of the continental populations of several species of colonial waterbirds (Martin and Lester 1990). Colony surveys with fixed-wing aircraft have been conducted at infrequent intervals in Louisiana since the 1970s (Portnoy 1977, 1978, Keller et al. 1984, Martin and Lester 1990, Michot et al. 2003, Green et al. 2006). Aerial surveys were used because they allowed single observers to survey large areas over short time periods; furthermore, many colonies are located in remote areas that would be difficult to access from the ground. Although use of fixed-wing aircraft might be the most economical means of surveying these colonies, the importance of this region to several waterbird populations provides the impetus for understanding the degree to which species-specific estimates of nesting activity are biased. If these biases are great, relative to helicopter or ground counts, improvements in estimates might justify use of more expensive survey approaches. We also wanted to quantify error associated with ground counts by using a marked-subsample method for ground surveys based on the Lincoln-Peterson estimator. Specifically, our objectives were to 1) compare the estimated number of breeding pairs of waterbirds in colonies between fixed-wing aircraft, helicopter, and ground surveys, especially in reference to plumage coloration and body size, 2) evaluate the marked-subsampling estimator as a method for ground counts and, 3) examine the probability of detection and visibility bias (not reported here) associated with ground counts. DISCUSSION Comparisons of colony size estimates revealed fixed-wing aircraft surveys overestimated the total number of breeding pairs per colony regardless of plumage coloration. Fixed-wing aircraft surveys tended to be biased and poorly reflect counts made from the ground. Counts of large, conspicuous species (e.g., great egret and roseate spoonbill) appeared to be the least biased and most precise estimates made from fixed-wing aircraft. With the remaining species, fixed-wing aircraft estimates resulted in imprecise estimates of the actual number of breeding pairs as determined from ground counts. The tendency of biases associated with fixed-wing surveys to increase with the number of nesting pairs in a colony is especially problematic. Fixed-wing aircraft surveys appear to overestimate number of breeding pairs as colonies increase in size, especially colonies containing white-plumaged species.. Helicopter and ground surveys were remarkably similar for estimating total number of waterbirds as well as for some individual species. Mean number of breeding pairs was similar for several abundant species (e.g., great egret) as well as for less common species (e.g., snowy egret). Helicopter surveys appeared to provide substantially more precise and less biased estimates of most species in comparison to ground counts than fixed-wing surveys. Helicopter surveys of species nesting high in the canopy (e.g., neotropic cormorants and great egrets) were the most accurate. The differences in probabilities of detection between observers demonstrates the importance of using multiple observers in conducting colony counts, whether by using a marked subsample count as we employed or a simple double count. Single observers are likely to underestimate the number of nesting birds in ground counts, resulting in a biased count. Whenever feasible, we recommend managers conduct ground surveys to estimate colony size and increase the probability of detection of rarer species. When a survey involves a large number of colonies, or includes some colonies that are either inaccessible from the ground or are so expansive that ground surveys are not practical, we recommend managers use a combination of ground and aerial surveys to estimate colony size. For aerial methods, helicopter surveys appeared to be the least biased and most precise method. If examination of spatial and temporal dynamics of colonies is the goal, helicopter surveys would provide considerably better estimates of nesting pairs than those possible with fixed-wing aircraft. For ground surveys, we strongly recommend that managers incorporate the use of double observers for estimates and preferably use the marked-subsample method. Our results demonstrate differences between observers in detection probabilities, which could lead to inaccurate estimations of colony size from a single observer. Furthermore, the marked-subsample method provides estimations of error, probability of detection, and visibility bias associated with ground surveys. When a colony size is small (<500 breeding pairs) and the colony can be completely observed, we recommend the marked-subsample method for estimations of colony size. Ground survey colony estimates in relation to fixed-wing survey colony estimates Ground survey colony estimates in relation to helicopter survey colony estimates ACKNOWLEDGMENTS This project was funded by the Louisiana Department of Wildlife and Fisheries and Louisiana Natural Heritage Program. We especially thank G. Lester and I. Maxit, Louisiana Natural Heritage Program. We are indebted to B. Adams, M. Collins, A. Hitch, and K. Purcell, University of Louisiana at Lafayette, for assistance during ground surveys. We thank B. Seal, Southern Helicopters. We thank the following U. S. Fish and Wildlife Service (USWFS) biologists and National Wildlife Refuges (NWR) for assistance during ground surveys: W. Syron (Lacassine NWR), G. Harris (Cameron Prairie NWR), S. Reagan (Southwest Louisiana NWR complex), and P. Yakupzack (Mandalay NWR). We also thank J. Linscombe, Rockefeller State Wildlife Refuge, and W. Sweeny, White Lake Preserve. Geographic Information Systems (GIS) mapping was conducted using the GIS Laboratory at the National Aeronautical and Space Administration Regional Application Center, in Lafayette, Louisiana. We thank B. Vermillion, F. Weckerly, and K. Weeks for comments and suggestions to an earlier version of this manuscript.