| The amount of chlorophyll and water in a plant is essential to assess how well the plantgrows or whether it adapts to a specific living condition. One factor is that chlorophyll is theunique medium in photosynthesis, for it absorbs solar radiation and enables plants to synthesiscarbon dioxide and water into carbohydrates. In this sense, chlorophyll is usually regarded asa predictor of a plant’s nutrition state, photosynthetic capacity and growth stage. The otherfactor is that water, as a major compound used in photosynthesis, serves as another predictorof a plant’s photosynthetic efficiency and its overall growth. The significance of measuringthe amount of water in the leaves of a plant is also reflected in its productivity of potentialdrought or wildfire, in its direction to irrigation and in its productivity of the productivity inagronomy.Pterocarya stenoptera C.DC. and Pinus elliottii Engelm. are suitable candidate speciesfor ecological restoration in the Three Gorges Reservoir. In order to judgment their growthstate with rapid manner without loss and provide better theoretical and practical guidance forecological restoration, their chlorophyll concentration and leaf water content is assessedthrough the hyperspectral data in this dissertation. The main research contents are specified asfollows:(1)In order to estimate the leaf chlorophyll content of Pterocarya stenoptera C.DC. andPinus elliottii Engelm., two kinds of methods are used. The leaf chlorophyll content modelsbased on9common vegetation indexed were established, the leaf chlorophyll content modelsbased on the red edge position, kurtosis and skewness of Spectral curve were established also.The leaf water content models based on4sensitive vegetation indexed were established, inorder to estimate the leaf water content of Pterocarya stenoptera C.DC. and Pinus elliottiiEngelm.. The results showed that the inversion result of Pterocarya stenoptera C.DC. leafchlorophyll content on the spectral index are similar to Pinus elliottii Engelm., they bothsensitive to VOG1and ND705, and these models are better than the others. The coefficientcorrelations for Pterocarya stenoptera C.DC. were higher than for Pinus elliottii Engelm., theformer were up to0.865and0.841, respectively, and the latter only were0.762and0.765, respectively. For estimation of leaf chlorophyll content using spectral characteristicparameters, there were opposite effect for Pterocarya stenoptera C.DC. and Pinus elliottiiEngelm.. It was to Pterocarya stenoptera C.DC., its estimation effect of spectralcharacteristic parameters was inferior to the Optimal effect of spectral indices and itscorrelation coefficient was only about0.7, otherwise, for Pinus elliottii Engelm., thecorrelation coefficient of kurtosis and skewness were0.873and0.855,respectively, itobviously was superior to the Optimal effect of spectral indices. For the inversion results getfrom the spectral indices of leaf water content about Pterocarya stenoptera C.DC. and Pinuselliottii Engelm., the accuracy of their optimal models were very similar, and the correlationcoefficient was about0.7. But, they were built on different spectral indices, it was WI2forPterocarya stenoptera C.DC. and was II for Pinus elliottii Engelm..(2)Because artificial neural networks (ANNs) have strong capacities for non-linearmapping, it was used to predict leaf chlorophyll content and the leaf water content ofPterocarya stenoptera C.DC. and Pinus elliottii Engelm.. Firstly, the red edge position,kurtosis and skewness, which were computed by spectral curve at the range680-760nm ofPterocarya stenoptera C.DC. and Pinus elliottii Engelm., were considered the input variablesfor artificial neural networks with three lays, and the leaf chlorophyll content was simulatedby the artificial neural networks, the simulated values and measured values were used tolinear regression and the corresponding regression equation were establish. Secondly,hyperspectral data was reduced the dimension using principal component analysis, then thescores of principal component were entered the BP neural networks and the correspondingsimulated values were given. The regression equation about leaf chlorophyll content wasestablished. Finally, the same PCA-BP neural network was applied to estimate the leaf watercontent of Pterocarya stenoptera C.DC. and Pinus elliottii Engelm. and the correspondingestimated model were given, respectively. Compared these results, the following conclusionswere obtained. The estimation effect using spectral indices was not as good as using PCA-BPneural networks, it was very obviously for leaf chlorophyll content of Pterocarya stenopteraC.DC., the correlation coefficient of the former was0.864, however, the latter was0.935. Inabove all, the effect using artificial neural networks were better than the effect using thespectral indices or spectral character parameters regardless of Pterocarya stenoptera C.DC.and Pinus elliottii Engelm.. This conclusion was applied to estimate the leaf water content ofPterocarya stenoptera C.DC. and Pinus elliottii Engelm.. It also can proved from the modelsabout Pinus elliottii Engelm., the correlation coefficient base on PCA-BP neural networkswas more higher than the other methods, it was up to0.905.(3)Fuzzy neural networks was utilized to estimate leaf chlorophyll content of Pterocarya stenoptera C.DC. and Pinus elliottii Engelm..The red edge position, kurtosis and skewness,which were computed by spectral curve at range680-760nm of Pterocarya stenoptera C.DC.and Pinus elliottii Engelm., were considered the input variables for fuzzy artificial neuralnetworks and under the condition that membership functions was “plmf”, the fuzzy neuralnetworks was trained by some data. Then, the simulated value of leaf chlorophyll content ofPterocarya stenoptera C.DC. and Pinus elliottii Engelm. were obtained by utilizing, thetrained fuzzy artificial neural networks and inversion model were established. Compared withthe obtained results, the estimation effect using this method was better than the artificialneural networks. It could be proved from inversion effect of the model about Pinus elliottiiEngelm.. The correlation coefficient of artificial neural networks was0.935, thecorresponding coefficient was0.966for another condition. It can indicate that this methodwas feasible to estimate leaf chlorophyll content.(4) Wavelet analysis used to predict leaf chlorophyll content and the leaf water content ofPterocarya stenoptera C.DC. and Pinus elliottii Engelm.. Firstly, Wavelet decomposition rangewas determined, then wavelet basis function which were utilized to estimate leaf chlorophyllcontent and the leaf water content of Pterocarya stenoptera C.DC. and Pinus elliottii Engelm.The optimal decomposition scales of determined wavelet basis function were given. Finally,wavelet energy coefficient and wavelet coefficient, which were computed by choose waveletbasis function under the corresponding optimal decomposition scale, were used to build thelinear regression and multi-linear regression function of the leaf chlorophyll content and theleaf water content of Pterocarya stenoptera C.DC. and Pinus elliottii Engelm., respectively.From the results, the estimation effect by wavelet analysis were better than the spectralindices in general, especially for leaf water content, but it was not as good as the artificialneural networks and fuzzy neural networks. Meanwhile, there were differences for the endresults with different wavelet basis function and different decomposition scale. Even for thesame case, the results were obviously difference with different wavelet basis function, it couldfounded from the multi-linear regression of Pterocarya stenoptera C.DC.In summary, BP neural networks and Fuzzy neural networks were the best models in theestimation of leaf chlorophyll content and leaf water content about Pterocarya stenopteraC.DC.and Pinus elliottii Engelm.. We can use these two methods on the rapid manner withoutloss, then grapping the plant growth dynamics effectively and providing better theoretical andpractical guidance for ecological restoration in the Three Gorges Reservoir. |
PDF Full Text Request