Empirical Models For Measuring The Leaf Area And Leaf Mass For Major Broadleaf Species In Northeast China

Rapidly and accurately estimating broadleaf leaf area(LA)and leaf mass(LM)is one of the basic work for plant function in a non-destructive conditions,which is key parameters for predicting plant growth,but empirical models for measuring LA and LM in different broadleaf species of saplings and adult trees across growing periods have rarely been proposed.In this study,we took broadleaf tree species Betula platyphylla,Betula costata,Tilia amurensis,Fraxinus mandshurica,Acer mono,Ulmus japonica,Ulmus laciniata,Acer tegmentosum as the study objects in a mixed broadleaved-Korean pine(Pinus koraiensis)forest.We constructed the separated models between LA or LM and leaf structure parameters(e.g.,leaf length,L,and leaf width,W.etc)for saplings and adult trees in May(leaf growing period),July(the stable period)and September(the senescent period),to explore whether species,life history and growing period had a significant effect on the selection of the separated model types and independent variables or not,and then the data from the eight broadleaf tree species were classified into two categories by L:W,in order to constructed the combined models for predicting LA and LM in different tree species,life histories and growing period.Finally,we evaluated the forecast accuracy of these separated models and combined models in predicting LA and LM,and to test whether the combined models was useful for predict LA and LM for other broadleaf species at the leaf scale.The results showed that:(1)The optimal separated model types for predicting LA and LM was significantly affected by species,life history and growing period.The selection of optimal independent variable for predicting LA were L for saplings of Ulmus laciniate in May and adult trees of Acer tegmentosum in July;and W for the saplings of Betula platyphylla in May and July and the saplings of Acer tegmentosum in May;the rest selection of optimal independent variable for predicting LA were all LW.However,the selection of optimal independent variable for predicting LM was more diverse due to species,life history and growing period.(2)The forecast accuracies of the separated models in predicting LA and LM were ranged from 83%to 96%and from 62%to 92%for eight broadleaf tree species,respectively;in contrast,the forecast accuracies of the combined models in predicting LA and LM were ranged from 83%to 93%and from 76%to 83%under different L:W categories,respectively.In addition,the combined models were suitable for predicting the LA and LM of six shrub species with maximum forecast accuracies of 93%and 85%,respectively,indicating that the combined empirical models we developed were generally effective in predicting LA and LM for other broadleaf species.(3)Furthermore,in order to rapidly and accurately(with forecast accuracies not less than 97%)estimating broadleaf LA and LM,the optimum number of leaf samples ranged from 11 to 14 for LA and from 59 to 62 for LM.These results would provide technical support for effectively estimating leaf traits and their dynamics change at a leaf scale in future.