USGS Home
Contact USGS
Search USGS

In the mid-1970s the USFWS implemented significant modifications that removed much of the previous subjective decision making and formalized the process by incorporating population and production data from the current year in all calculations of fall flights. Specifically, annual production information from the July survey, such as brood and late nesting data, is used each year to adjust base production rates established earlier for these areas. For those units not surveyed in July, constant production rates are used. These rates, together with the total spring population of ducks for each survey unit, provide a generalized assessment of the anticipated fall flight. This revised approach still assumes that recruitment is the same among all species, base production rates do not vary over time, and over-summer mortality is negligible. However, contrary to the earlier procedure, the allocation of the fall flight to various species is no longer attempted, and the resulting index simply reflects ducks in total. These limitations and others associated with this procedure obviously restrict its usefulness as an accurate estimator of the continental duck population in the fall. Yet it has never been considered as such, nor could it ever be easily substantiated even if used in this manner. By themselves, these indices have limited application. However, because of their relative comparability over time, due to consistent methods of calculation, they do have some utility in evaluating the status of duck populations. Together with many other sources of information, these forecasts provide some assistance to waterfowl managers in establishing annual hunting regulations. Unfortunately, these indices have frequently been used improperly as exact measures of fall duck numbers, resulting in some confusion among those unaware of the limitation of these values.

The second method of estimating fall flight is based on the use of predictive models and, in contrast to the generalized methods described above, is directed at one species, the Mallard. Despite an extensive data base, the best measure of the fall population is obtained only after the hunting season. Specifically, the age ratio in the fall population is used as an index to Mallard recruitment in the preceding breeding season. The first model described by Martin et al. (1979) is a regression equation in which the continental number of young in the fall is a function of the spring breeding population, the number of broods counted in July, and the number of July ponds in the southern prairie provinces. More recently, a modification of this general model has been used to predict Mallard fall age ratios each year. This later version circumvents earlier problems in the model by using the age ratio in the harvest, corrected for differential vulnerability, as the dependent variable and incorporates breeding ground information such as late nesting index, number of May ponds, and pond loss between May and July as explanatory variables.

A third regression is a nonlinear model proposed by Hammack and Brown (1974). The explanatory variables are breeding population and July ponds, and the model assumes density-dependent recruitment.

All of these models are dynamic in nature and, each year, the addition of data from the previous year results in new estimates of the model parameters. Predictions from these generalized models all agreed reasonably well but at times have yielded different forecasts of recruitment rates. In some years, a range of possible recruitment rates is considered or one estimate is selected based on other relevant biological data. Martin et al. (1979), using this production rate information along with spring breeding numbers and an estimate of adult survival over summer, outlined a generalized approach to forecasting the total Mallard population in fall. They also noted a high degree of agreement between projected fall flight indices and harvests in subsequent hunting seasons.

After each hunting season, the prediction of the continental production rate is checked against the age ratio information in the harvest, adjusted for differential vulnerability. The consistency of prediction does not mean that the results are accurate unless the estimate of fall age ratio, determined after the hunting season, is itself accurate. Munro and Kimball (1982) described the procedures used to estimate the preseason age and sex structure of the Mallard population from data on harvest, breeding population, and band recoveries. They also listed the assumptions required for the estimates to be without bias. Unfortunately, many of these assumptions are not met. As mentioned previously, the prediction of the fall population of Mallards also requires estimates of survival of adults during summer. Recent studies (Johnson and Sargeant 1977, Cowardin et al. 1985, Kirby and Cowardin 1986, Blohm et al. 1987) presented considerable evidence that summer survival rates used in the past were too high and should not have been treated as constant.

Despite the numerous biases and problems involved in the various data bases and methodology used for estimating breeding population and predicting fall flight in North America, the system is undoubtedly one of the most comprehensive and objective decision-making systems in wildlife management. The important consideration is that the biases and their possible impact on management decisions be recognized. The significance of these management decisions has promoted constant scrutiny of the system and encouraged all efforts to improve it.