Northern Prairie Wildlife Research Center
The mallard model (MM) uses, as variable inputs, a file containing 161 values describing pond conditions from 15 March to 22 August. These values specify a step function of the proportion of the semi-permanent wetlands that contain water on each date (Mack 1991). Pond conditions on a given day influence the MM probability that a hen will initiate a nest on that day. The MM offers default choices of dry, average, or wet conditions, or the investigator can furnish site-specific wetland conditions with an alternative input file. The defaults describe conditions based on data collected in central North Dakota under typical long-term precipitation patterns. They do not portray the influence of sporadic rainfalls that occur most years. We reshaped the default step functions for pond conditions in an attempt to more closely reflect the changes in wetland conditions that occurred during the nesting season, 1990-1992.
Wetland abundance in a given spring depends on a number of factors of which precipitation in the previous year or years is important. An index to conserved soil moisture (Williams and Robertson 1965) has been used to model wetland conditions in a given year (Boyd 1981, Bethke and Nudds 1995). This complex index uses precipitation totals for 21 months preceding May and weights monthly totals differently depending on when precipitation is presumed to contribute the most to soil moisture. Larson (1995) also determined that the number of wet basins in the United States and Canadian prairies could be predicted with a model including precipitation totals for the previous fall and several other weather related variables. We used a similar but simplified approach that allowed us to model the influences of past precipitation as well as that occurring within a year on the proportion of wet basins existing during the season.
We assumed that precipitation since the previous 1 July influenced spring pond conditions, and that pond conditions were determined directly by the departure of precipitation totals from long-term values. We obtained daily (1989-1992) and long-term daily (1961-1990) precipitation values from the State Climatology Office for stations as close to each of our 3 study sites as possible (i.e., Ivanhoe - 1 mi east of the Ivanhoe study site [daily] and Tyler - 15 mi southeast of the Ivanhoe study site [long-term]; Lake Wilson - 10 miles northwest of the Slayton study site; and Windom - 15 miles southeast of the Windom study site). For each year and study site, we determined the proportional departure of the cumulative precipitation beginning the previous 1 July from the long-term cumulative for the same dates (%-DEP) (Figs. C-1 to C-3).
After consulting the local Minnesota Department of Natural Resources wildlife managers, we decided that wetland conditions on the 3 study sites at the beginning of each spring were generally dry in 1990 and average in 1991 and 1992. We then reshaped the MM dry and average pond condition step functions by multiplying year-, site-, and date-specific %-DEPs by the appropriate default value for each day of the nesting season (e.g., Fig. C-4). The new values were rounded to the nearest 10% to agree with the input precision used by the MM (Table C-1.).