Nathan L. Stephenson Phillip J. van Mantgem 2005 <p><span dir="ltr">1 </span><span dir="ltr">We </span><span dir="ltr">assess the use of simple, size-based matrix population models for projecting </span><span dir="ltr">population trends for six coniferous tree species in the Sierra Nevada, California. We </span><span dir="ltr">used demographic data from 16 673 trees in 15 permanent plots to create 17 separate </span><span dir="ltr">time-invariant, density-independent population projection models, and determined </span><span dir="ltr">differences between trends projected from initial surveys with a 5-year interval and </span><span dir="ltr">observed data during two subsequent 5-year time steps.</span></p><p><span dir="ltr">2 </span><span dir="ltr">We </span><span dir="ltr">detected departures from the assumptions of the matrix modelling approach in </span><span dir="ltr">terms of strong growth autocorrelations. We also found evidence of observation errors </span><span dir="ltr">f</span><span dir="ltr">or measurements of tree growth and, to a more limited degree, recruitment. Log linear </span><span dir="ltr">analysis provided evidence of significant temporal variation in demographic rates for </span><span dir="ltr">only two of the 17 populations.</span></p><p><span dir="ltr">3 </span><span dir="ltr">T</span><span dir="ltr">otal population sizes were strongly predicted by model projections, although </span><span dir="ltr">population dynamics were dominated by carryover from the previous 5-year time</span><span dir="ltr">step (i.e. there were few cases of recruitment or death). Fractional changes to overall </span><span dir="ltr">population sizes were less well predicted. Compared with a null model and a simple </span><span dir="ltr">demographic model lacking size structure, matrix model projections were better able to </span><span dir="ltr">predict total population sizes, although the differences were not statistically significant. </span><span dir="ltr">Matrix model projections were also able to predict short-term rates of survival, growth </span><span dir="ltr">and recruitment. Mortality frequencies were not well predicted.</span></p><p><span dir="ltr">4 </span><span dir="ltr">Our results suggest that simple size-structured models can accurately project future </span><span dir="ltr">short-term changes for some tree populations. However, not all populations were well </span><span dir="ltr">predicted and these simple models would probably become more inaccurate over longer </span><span dir="ltr">projection intervals. The predictive ability of these models would also be limited by </span><span dir="ltr">disturbance or other events that destabilize demographic rates.</span></p> application/pdf 10.1111/j.1365-2745.2005.01007.x en British Ecological Society The accuracy of matrix population model projections for coniferous trees in the Sierra Nevada, California article