| Deciduous broad-leaved forest is a part of the global ecological system and the economic forest. Caused by rapid growth and short rotation, its geographic distribution and biomass is always changing every year. With the improvement of the requirement for timeliness and accurate monitoring is increasing, remote sensing technology has gradually replaced the traditional method by artificial survey. But there are still some problems as follows:It is very difficult to obtain high resolution satellite data timely because of the influence of revisit period, forest phenology and the weather condition, which is currently the key technical problem in using the remote sensing technology to extract accurate forest distribution and estimate ecological parameter. In addition, the spatial resolution of some remote sensing images is too low to meet the demand of precision forestry operation. The last but not the least, the applicability and accuracy of the estimation model is limited when leaf area index (LAI) of forests is estimated with the method of vegetation index.The HJ satellite has the advantage of higher resolution and shorter imaging cycle, so we can make full use of the HJ satellite images for different periods to explore the phenological characteristics during vegetation growth period, identify various vegetation species and track vegetation growth simply and effectively. In this paper, taking full advantage of CCD remote sensing data with multispectral, higher phase, higher spatial resolution and easy to obtain, a normalized difference vegetation index (NDVI) difference rate recognition model of deciduous broad-leaved forest and LAI estimation model for its whole growth period were constructed combining with the field data. Moreover, taking Chuzhou in Anhui province as the study area, the distribution of deciduous broad-leaved forest was extracted. Then based on field data, taking the aspen(Populus) as an example, the LAI-NDVI estimation models for the leaf production period, flowering and fruit-bearing period, leaf constant period and leaf abscission period were constructed respectively and their applicability was analyzed. Finally, the LAI estimation model for the aspen forest’s whole growth period was established by comparative analysis and its applicability was also examined.The main research contents and conclusions of the paper are as follows:(1) The NDVI difference rate recognition model of deciduous broad-leaved forest was constructed, and its effectiveness and reliability for extraction was verified. The producer accuracy reached to83%and the user accuracy was90.6%.(2) Taking the aspen as an example, the LAI eastimation models for leaf expansion period and flowering, fruit-bearing period, leaf constant period and leaf abscission period were constructed. By comparative analysis, it was clear that the applicability of LAI estimation models for each period:the applicability performed well for leaf expansion period and flowering and fruit-bearing period, while the applicability was poor for leaf constant period and leaf abscission period due to the concentration of NDVI at high values of LAI and the influence of shrub and grass under the canopy.(3) Taking the aspen as an example, the LAI estimation model for its whole growth period was established using the LAI measured in all the periods and NDVI computed from HJ-CCD remote sensing images. Results showed that the correlation coefficient(R2) between the measured LAI and the estimated LAI was0.531, and the root mean square error (RMSE) was0.38. All of the above results indicate that this model can be used to estimate LAI for the leaf constant period and leaf abscission period, which makes up the limitation of LAI estimation model of single period. Finally, the applicability validation results show that this model can be used to estimate time series LAI. |
PDF Full Text Request