Abstract
The portable chlorophyll meter [Soil Plant Analysis Development (SPAD)] has been used successfully for measuring leaf‐nitrogen (N) of several crops. Determination of the appropriate sample size, in terms of number of plants to be sampled within each plot, has recently become a matter of concern. An insufficient sample size does not allow for the detection of small, but real differences between treatment means, whereas an excessively large sample size constitutes a waste of time and resources. In this study, SPAD meter data were collected at two sites. Each of these two field experiments was organized following a split‐plot design with three blocks and two treatment factors: four nitrogen levels (main plot factor) and six maize (Zea mays L.) hybrids (subplot factor), selected to represent a broad range of canopy types. The approach followed in collecting SPAD meter readings for the determination of an appropriate sample size consisted of sampling one leaf per plant and taking a single reading per leaf. Confidence intervals for the mean of SPAD meter readings and the associated required sample sizes for the variability observed were generated using a standard procedure. Taking a single reading per leaf near the midpoint of the leaf blade, a sample size of 15 to 20 plants provided a level of precision of 5% (about ± 2.8 SPAD meter units). The variability among and within hybrids was highest at the zero N fertilization level (kg ha−1), for which the leafy and non‐leafy reduced stature and leafy normal stature hybrids showed the largest required sample sizes at both sites. At the Ottawa site, where an N fertilization effect was observed, required sample sizes at the 0 N level were larger than at any other level, including the recommended 170 N level. In summary, a relationship between sample size and precision level is presented for maize researchers using the SPAD technology.
Acknowledgments
Appreciation is extended to Teshome Melkamu, Isabelle Martineau, and Peter Neave for their assistance in field data collection. This research was supported partially by the Natural Sciences and Engineering Research Council of Canada (NSERC) through a collaborative research grant. The senior author is grateful to the Brazilian Post‐Graduate Federal Agency (CAPES) for financial support.