Northern Prairie Wildlife Research Center
The estimates derived from the remote-sensing-based system may have several biases. Testing for these biases requires special studies that were beyond the scope of our evaluation. Typical difficulties in estimating duck numbers and production over vast areas may preclude objective verification. Our purpose here is to alert the reader to some of the more important possible sources of errors.
The remote-sensing-based system is based on data from remote sensing for the identification of wetlands and uplands from aerial photographs and on data from aerial video for measuring annual change in the number of ponds and area of water. We made the assumption that mapping by the National Wetland Inventory was without errors of omission or commission. Although not tested, our experience in this and numerous other studies in the prairie pothole region revealed that such errors are few. Errors that we detected were usually errors of omission of small temporary wetland basins. These errors lead to underestimation of breeding-population sizes. Data on the number of ponds and on the area of water from video (sometimes poor quality) may contain errors from the omission of numbers of ponds (L. L. Strong, Northern Prairie Science Center, National Biological Service, Jamestown, N. Dak., unpublished data). These errors cause the underestimation of breeding-population sizes. They also cause the underestimation of A and, therefore, recruitment rate. We also point out that our method for estimating A relies on data from only 4 years in one study. We also make the assumption that the form of the regression curves for estimating the number of breeding pairs is the same among years and areas. This assumption was not examined. Adjustment of regressions by G may have introduced errors that we are currently unable to evaluate.
For the estimates of duck production, we used Conservation Reserve Program cover estimates from 1990 and applied them to all years. Conservation Reserve Program contracts were taken during 1987-90. Conservation Reserve Program cover is attractive to ducks with relatively high clutch success (Kantrud 1993, Reynolds et al.1994); therefore, our estimates of production in years prior to 1990 may be high.
All systems for the estimation of breeding-population sizes of ducks are subject to errors of biological interpretation such as determining whether an observed pair of ducks represents a resident or migrant pair and whether observed social groups such as lone males represent breeding pairs (Cowardin and Blohm 1992). We made no attempt to solve these problems but attempted to use methods that correspond to those used in other surveys so that the estimates are comparable.
Although our overall sample size (Table 1) was large, samples in some wetland-management-district-landownership strata were small, and variation in habitats among plots was great. Confidence limits on most estimates of habitat parameters were large. Cost probably prohibits large increases of sample size, although addi tional plots are currently added to the sample where sample sizes are minimal. At present, wetland mapping and digitizing in the entire prairie pothole region are progressing rapidly, and technology is available for adding uplands to the data. When data from the entire prairie pothole region become available, sampling will no longer be necessary for estimating numbers of wetland basins and wetland area. Our confidence limits for duck-population sizes and duck-production parameters reflect only the variation associated with our sample of 10.4-km² plots. For example, the confidence limits do not reflect the variance in the number of ducks or recruitment in individual plots. Thus, the confidence limits for these parameters are actually narrower than they should be.
Suggested Future Modifications
The remote-sensing-based system, although operational, should be considered a prototype that can be improved by various modifications and by the addition of new data. One simple modification would be a complete inventory of wetlands as described in the previous section. Annual estimation of the number of ponds will probably have to remain sample based because of cost and logistic problems in obtaining complete remote-sensing data over such a large area each spring. Satellite data with sufficient resolution to delineate the small ponds that ducks use may become available at a reasonable cost in the future, but obtaining data in the required narrow time frame will remain a problem. Advances in aerial videography techniques have been made since the remote-sensing-based system was initiated. We recommend that new methods be evaluated to increase the accuracy of the annual estimates of the number of ponds. We used ratio-estimation methods involving the amount of land area in each landownership class to improve our estimates of several parameters. Ratio estimates involving area of wetland or numbers of basins in each landownership class could further improve estimates of parameters such as area and number of ponds or numbers of breeding pairs. Such improvements will be possible when the inventory of wetlands is complete.
As use of the remote-sensing-based system continues, new data become available for improving the regression equations for estimating numbers of breeding pairs from pond areas. Furthermore, the classification of wetland basins should be modified to better represent major differences in duck use. For example, the current class lakes includes many types of lakes that receive different use by ducks. The data should be examined to determine whether enough information exists to construct regressions for various kinds of lakes. In addition, the remote-sensh~gbased system was designed primarily for estimating numbers of pairs on discrete basins. The remote-sensing-based system does not work well for estimating numbers of pairs on large lakes and in riverine habitat. Studies should be conducted to determine whether shoreline segments of lakes and reaches of rivers are preferable sampling units in these habitats. Such a change requires gathering data from these units to develop appropriate regression models for the remote-sensing-based system. When we designed the remote-sensing-based system, we could not link vector data from the National Wetland Inventory maps to the raster data from aerial videography. Therefore, ground crews estimated the proportion of a wetland basin containing water. These estimates were used to calculate G. The link between the two data sets is now available, and we recommend that aerial videography (or possibly photography) be used for determining G.
Overall Assessment of the System
Our first goal with the remote-sensing-based system was consistency of estimates among areas and years. We believe that this goal was accomplished, although major differences in habitat among waterfowl management districts demand some regional modification of methods. Our second goal was maximum use of existing data. We believe that this objective was accomplished, but implementation of the remote-sensing-based system clearly pointed out that certain data required by the system are scarce or lacking. This is especially true for reliable estimates of clutch success in some areas and in some habitats (Shaffer and Newton, in press). Our third goal was rapid execution of complex procedures by microcomputer and a resulting database that documents changes in the estimated parameters. This goal was partly accomplished. Because the remote-sensing-based system has been constantly evolving, continued modification of computer programs has been necessary. At the same time, technological advances in computer development and remote-sensing techniques have moved ahead of those used in the remote-sensing-based system.
Fluctuation-sometimes violent-of breeding-population sizes and production of prairie ducks causes problems for waterfowl managers. The manager usually attempts to manage such things with the amount and extent of harvest and the availability and quality of habitat. The success or failure of management is usually evaluated in terms of duck numbers and production. Johnson and Shaffer ( 1987) demonstrated the difficulty of separating weather factors from the numerous other causes of fluctuation in duck population sizes and the need for long-term surveys with consistent methodology. The remote-sensing-based system proposed here has the potential to provide better evaluation of management by providing data that are essential to understanding the interaction of various factors that cause fluctuation in duck population sizes.