Mon, Aug 02, 2021:On Demand
Background/Question/Methods
Understanding the responses of plant population growth or plant demographic rates to temporal changes in climatic drivers is frustrated by prohibitive data requirements. The minimal length to quantify the effects of temporal climatic variation on plant population dynamics is generally assumed to be 20 years. Here, we use power analysis to test whether the responses of plant populations to temporal climatic variation can be estimated collecting three to five years of data over a large number of spatial replicates - defined as populations not connected by dispersal. If these spatial replicates share similar mean climate, their populations should have similar responses to temporal climatic variation. Our power analysis uses parameter estimates obtained from a recent synthesis performed on data from 162 plant populations. We simulate a simple linear regression by modifying i) the absolute number of data points, ii) the range of climatic anomalies, iii) the true effect of climate among spatial replicates, and iv) the size of the residual variance among spatial replicates. Finally, we quantify the spatial correlation of annual temperature and precipitation anomalies in North America. Uncorrelated climatic anomalies across space allow sampling a larger range of climatic anomalies in a shorter number of years.
Results/Conclusions These results show that favoring spatial over temporal replication provides an untapped opportunity to correctly estimate the effects of temporal climatic variation on plant populations. Sample size and the range of climatic anomalies act synergistically to increase statistical power, while the effects of varying the true effect of climate among spatial replicates are minimal. On the other hand, statistical power decreases substantially in case residual variance changes unpredictably among spatial replicates. Finally, we found that across North America, the spatial correlation of annual precipitation (but not temperature) anomalies decrease to an average of 0.5 at 250Km of distance between sites. Our power analysis shows that 30 years of sampling provide a power of 0.5. The same statistical power can be reached sampling 30 separate populations for just four years if three conditions are satisfied. First, all spatial replicates need to experience a similar climate, to ensure similar responses temporal climatic variation. Second, populations need to be divided in two groups, separated by at least 250Km of distance, to increase the range of climatic anomalies sampled. Third, the temporal variance in the response variable of all populations needs to be comparable.
Results/Conclusions These results show that favoring spatial over temporal replication provides an untapped opportunity to correctly estimate the effects of temporal climatic variation on plant populations. Sample size and the range of climatic anomalies act synergistically to increase statistical power, while the effects of varying the true effect of climate among spatial replicates are minimal. On the other hand, statistical power decreases substantially in case residual variance changes unpredictably among spatial replicates. Finally, we found that across North America, the spatial correlation of annual precipitation (but not temperature) anomalies decrease to an average of 0.5 at 250Km of distance between sites. Our power analysis shows that 30 years of sampling provide a power of 0.5. The same statistical power can be reached sampling 30 separate populations for just four years if three conditions are satisfied. First, all spatial replicates need to experience a similar climate, to ensure similar responses temporal climatic variation. Second, populations need to be divided in two groups, separated by at least 250Km of distance, to increase the range of climatic anomalies sampled. Third, the temporal variance in the response variable of all populations needs to be comparable.