| Leaf area index (LAI) is considered as the key variable for the net primary productivity and the carbon cycle assessment that is widely used in physiological and biogeochemical studies. It is not only the most important biophysical variable characterizing vegetation abundance and distribution across the landscape but also a part of the essential climate variable. Black locust (Robinia pseudoacacia L.) plays a vital role in the vegetation rehabilitation in the fragile ecosystem of the Loess Plateau. We analyzed how the tree condtion, tree density, stand age, and slope aspect influenced the LAI variance of black locust plantations. The dynamics of LAI were modeled by applying the two-compartment models. The LAI values were also estimated using spectral and textural information of Quickbird imagery and the high accuracy was obtained. The research on the LAI of black locust plantations could provide information for the evaluation of forest condition and productivity in Loess Plateau region, and also has important significance for the forest management of black locust plantations. The main results obtained were summarized as follows:(1) Influence of tree condition on the LAI of black locust plantationsThe LAI values of black locust plantation increased firstly and then decreased. The maximum LAI value was recorded in June and the minimum value in October. Except for September, healthy plantations of May, June, July and October had higher LAI values than below-average and poor health plantations. The mean LAI values of healthy plantations were significantly larger than those of below-average and plantations in poor health; however, it was difficult to distinguish below-average plantations from those in poor health’.(2) Two-compartment models for describing the temporal variation in the LAI of black locust standsSimple linear regression equation was used to build the relationship between stand age and LAI in different slope aspect. The R2between stand age and LAI was unstable and the largest value of R2reached0.91. July was determined to be the best month of measuring the LAI value of black locust plantations. Polynomial model which was L(t)=M+Nt+Ut2+Vt3(where M, N, U, V were empirical coefficients) was appied to predict the dynamic of LAI and the maximum value of R2reached0.91. On the basis of polynomial model, compartment model which was L2=βe-k2t-k1αe-k1t/(k1-k2)(where, α and β were empirical coefficients, k1and k2were increasing rate and decreasing rate respectively, t indicated month) was also used to model the dynamic of LAI. The maximum value of fitting coefficients reached0.93. Equivalence between polynomial model and compartment model was vertified by using numerical method to offer the theoretical basis for the polynomial model. AI of black locust plantations. According to the result of compartment model, the stands’ k1values were substantially greater than their k2values in most cases. Two separate stages of leaf dynamic were identified:one covering the period between emergence and leaf saturation, and another extending from saturation to senescence. The strongest increases in LAI were observed in11-15year-old stands on shady slopes (k1=4.76) and in young (<10years old) stands on sunny slopes (k1=3.94). The largest k2values were observed in mature stands (21-30years old) both on sunny and shady aspect. With the exception of the youngest stands (<10years of age), trees on shady slopes generally had higher k1and lower k2values than those on sunny slopes. The results obtained show that two-compartment models can be used to describe leaf growth in black locust stands, and that stand age and slope aspect both have strong effects on the rate and magnitude of the changes in LAI over the course of the growing season.(3) Performance of raw data processing and spectral vegetation indices (SVIs) of Quickbird imageryFour techniques including simple linear regression equation, second-degree polynomials equation, power equation and exponential model were applied to develop to describe the relationship between spectrum estimated from Quickbird imagery and52field measurements of LAI. The coefficients of determination for the relationships between the field LAI values and the spectral data for bands b1(blue), b2(green), b3(red) from the Quickbird data were found to be poor individually. However, stronger correlations were observed for band4(near infrared), for which the adjusted r2value was0.66. Overall, the SVIs offered better performance than was achieved by just looking at the reflectance from a single spectral band; the highest adjusted r2values were achieved using SAVI, NDVI, EVI and MSAVI (r2=0.68). SR and ARVI had the poorest performance with the r2=0.339and r2=0.359, respectively. The performance of second-degree polynomials equation was better than other three models.(4) Performance of texture processing of Quickbird imageryWe employed four techniques including simple linear regression equation, second-degree polynomials equation, power equation and exponential model to describe the relationship between texture parameters and52field measurements of LAI. We identified specific texture parameters (Angular Second Moment and Entropy parameters) were the most useful for estimating LAI values which could explain70%of the observed variation in the field LAI data. The Contrast (CON) parameters performed poorest which could merely explain53.7%of the observed variation in the field data. The r2increased with the window sizes for most texture parameter (eg. HOM, VAR, ENT, and ASM). By contrast, lager window sizes were selected to applied Dissimilarity (DIS) and Contrast (CON) for LAI estimation of black locust. There was no obvious difference among four models used for LAI retrievals. The performance of exponential model was little better than other three models.(5) Performance of SVIs in conjunction with texture parameters of Quickbird imageryFour techniques including simple linear regression equation, second-degree polynomials equation, power equation and exponential model were employed to describe the relationship between combined SVIs-texture parameters and52field measurements of LAI. The closest agreement between measured and estimated LAI values was achieved by using models based on a combination of MSAVI and ASM (adjusted r2=0.84, RMSE=0.41). The LAI estimation accuracy was improved compared to mere ASM, COR, and HOM when ASM, COR and HOM were combined with SVIs. To some extent, the accuracy of SVIs to estimate LAI was improved with the combination of CON, DIS, VAR and SVIs. The combination of ENTwith SVIs invariably yielded adjusted r2values that were lower than those achieved using SVIs alone. The combination of HOM, ASM and COR with SVIs gained the higher r2than those achieved using HOM, ASM and COR alone. The performances of CON, DIS and VAR were improved when combining with partly SVIs. The combination of Entropy data with SVIs invariably yielded adjusted r2values that were lower than those achieved using ENT alone... |
PDF Full Text Request