Local weather conditions predict avian species richness in a coastal sage scrub habitat

Background/Question/Methods

Species richness contributes to understanding the importance of locations in terms of biodiversity and accurate estimates of richness are important.  Sources of error in estimating richness merit investigation. Bird activity levels and the detection of birds can be affected by weather. This study examined effects of weather variables on estimates of bird species richness and the probability of observing individual bird species. We performed 50 visits of 30 min duration at a 1.25 ha study site in Los Angeles County, California and identified avian species. We gathered data at the start of each visit to the field site on weather variables: cloud cover, dew point, relative humidity, temperature, and wind speed. The latter four variables were also measured 15 min into the visit to quantify rate of change. Linear models were employed to investigate relationships between weather variables (predictors) and bird species richness and presence/absence (response).

Results/Conclusions

Thirty-two species were detected, with a mean of 11.2 ± 0.3 (SE) species observed per visit. A generalized linear model identified three significant predictors (cloud cover, dew point at the start, and rate of change in relative humidity) and significantly explained 20.5% of the variation in number of species per visit. Species specific logistic models were identified that significantly predicted the presence or absence of six species. All of the weather variables measured in this study appeared in at least one significant model. Particularly, each significant model included one or more of the following: dewpoint at start, change in dewpoint, relative humidity at start, and change in relative humidity. The relative moisture content in the air appears to be an important factor influencing the detection of bird species. Taken together, our results indicate it is feasible to construct a model using local weather conditions to predict the presence or absence of particular species and to generate estimates of species richness. Further, our results suggest that location specific models may help to prevent under-estimates of species richness and to increase the efficiency of detection in situations where a limited number of visits are possible.