Sample Size Requirements for In Situ Vegetation and Substrate Classifications in Shallow, Natural Nebraska Lakes

—We assessed the precision of visual estimates of vegetation and substrate along transects in 15 shallow, natural Nebraska lakes. Vegetation type (submergent or emergent), vegetation density (sparse, moderate, or dense), and substrate composition (percentage sand, muck, and clay; to the nearest 10%) were estimated at 25–70 sampling sites per lake by two independent observers. Observer agreement for vegetation type was 92%. Agreement ranged from 62.5% to 90.1% for substrate composition. Agreement was also high (72%) for vegetation density estimates. The relatively high agreement between estimates was likely attributable to the homogeneity of the lake habitats. Nearly 90% of the substrate sites were classified as 0% clay, and over 68% as either 0% or 100% sand. When habitats were homogeneous, less than 40 sampling sites per lake were required for 95% confidence that habitat composition was within 10% of the true mean, and over 100 sites were required when habitats were heterogeneous. Our results suggest that relatively high precision is attainable for vegetation and substrate mapping in shallow, natural lakes. Habitat classification is commonly used to assess differences over large scales (e.g., among lakes or streams). In addition, fish population characteristics, such as abundance and growth, can often be attributed to these large-scale differences. For example, bluegill Lepomis macrochirus growth was reduced at high abundances of submergent macrophytes in a Wisconsin lake (Trebitz et al. 1997), and largemouth bass Micropterus salmoides abundance increased with increasing aquatic vegetation abundance in Texas and South Dakota impoundments (Durocher et al. 1984; Guy and Willis 1991). Nebraska sandhill lakes with higher levels of emergent vegetation typically contained higherquality bluegill populations (Paukert et al. 2002). * Corresponding author: [email protected] 1 Present address: Southwest Biological Center, U.S. Geological Survey, 2255 North Gemini Drive, Flagstaff, Arizona 86001, USA. Received December 4, 2001; accepted April 1, 2002 Therefore, vegetation assessment is evidently needed for effective management of fisheries in lentic systems. Habitat assessment techniques are often designed to optimize time rather than accuracy (Platts et al. 1983) and therefore commonly employ subjective visual estimates. Previous work on visual estimation of stream habitat attributes has indicated that observer experience, habitat complexity, and habitat classification scale all affect precision (Platts et al. 1983; Roper and Scarnecchia 1995; Wang et al. 1996). However, research on precision of habitat classification has focused on lotic habitats and, to our knowledge, has not been assessed in lentic habitats. Our objective was to determine the precision of visually-estimated vegetation and substrate classifications in shallow, natural lakes. We focused on the classification of vegetation and substrate to allow assessment of fish community relationships. Therefore, we were more concerned with largescale precision (e.g., the coverage of dense emergent vegetation within a lake) than with microhabitat-scale precision. We wanted to determine whether our method was sufficiently precise to warrant its use as a rapid and effective vegetation and substrate assessment that fisheries and watershed managers could complete with limited time