| Monitoring forest resources change is the core and basic work of forestry investigation and forestry construction.Finding out the status of forest resources and exchanging dynamic information is an important basis for China to formulate national ecological security planning,forestry development planning and sustainable production and management.From the point of view of forest resource change factors,the forest type is the first basic element of the forest resource survey;the vegetation coverage is a direct manifestation of the vegetation coverage status of the forest area,and is an important indicator of the quality and function of forest vegetation ecosystems;and forest above-ground biomass(AGB),productivity quantitatively reflects the carbon sequestration capacity of forest vegetation.These four parameters are all direct indicators of the changes in forest resources and important measurement factors.They are also important indicators of forest quality.The Great Khingan Forest Region of Inner Mongolia and Genhe Forest Reserve were selected as experimental and verified areas respectively.Based on the previous forest researches and the evaluation and measurement of major remote sensing data origin,the thesis selected the representative active or passive multi-mode remote sensing and other measurement means to solve the key technologies bottlenecks of forest resource parameters remote sensing such as behind square of forest types,vegetation coverage,AGB,and productivity.The innovation of this research is the integration of multi-source observation data and the adoption of multi-mode remote sensing monitoring technology and mechanism models of spatial-temporal continuous forest resources information,thus building a multisource integration monitoring mechanism and improving the precision and time-effectiveness of monitoring regional forest resource information,which guarantee the deep analysis of the change rules of forest resources.The main conclusions of this study are as follows:(1)Integrated on the multi-resource remote sensing data and ground survey data,the research conducted the rapid,dynamic and fine-grained identification of forestry land categories.This thesis analyzed the changes of the two-temporal forest land cover types(including subdivision types of forests)and the whole regional types.Based on a variety of reference data,including field survey data,continuous inventory data of forest resources,and high-resolution satellite-borne multi-spectrum,on-board CCD images,etc.,accuracy verification was performed;the results showed that,at the pixel scale,RF outperforms SVM,and GF-1 data was superior(total accuracy:88.06%;Kappa coefficient:0.87)to TM data(total accuracy:84.44%;Kappa coefficient:0.83).Both of them could identify forest land types precisely.The monitoring result was indicated that from 2009 to 2014,the areas of broad-leaved forests,coniferous forests,and mixed forests increased,and the number of sparse forest lands decreased slightly.Forest land(specifically the broad-leaved woodland,coniferous forest land,mixed forest land,shrubbery land,sparse forest land)increased from 9.61997 million hectares to 9.73688 million hectares,with an increment of 0.011691 million hectares;the area of arbor forests increased.0.09422million hectares(1.215%).(2)GF-1 data preceded TM data on the monitoring of forest vegetation coverage.Based on above TM data in 2009 and GF-1 data in 2014,a dimidiate pixel model suitable for regional scale was constructed to invert the vegetation coverage of regional forest.The inversion results in 2014 were verified with the vegetation coverage survey data.The R~2 was0.766,and the root mean square error RMSE was 0.051.The inversion results show that the average coverage in 2009 was 72.73%(79.75%of the vegetation coverage in the medium grade)and 82.78%in 2014(the highest level of vegetation coverage in the forest area is the majority,44.60%).Compared with the environment factor,these two phases of vegetation coverages are all highly related to the elevation height of the terrain and the Pearson correlation coefficient R~2in 2009 and 2014 were 0.930 and 0.914 respectively.Viewed from the monitoring effect,CF-1data was advantageous on extracting vegetation coverage,and could be used as the significant data source of integrated monitoring means.(3)k-NN algorithm could rapidly estimate the above-ground biomass.Based on feature information such as vegetation coverage and the above two optical and P-band SAR data(ALOS-1 PALSAR in 2009 and ALOS-2 PALSAR in 2014),the optimization of the learning machine(such as k-NN)algorithm is focused on solving the optimal problem.With the feature fusion rapid extraction technology,a forest AGB inversion method is proposed for multi-source data with large areas and high feature dimensions.The results show that:at the plot scale,the AGB inversion results in 2009 were R~2=0.56,RMSE=25.95 t/ha;in 2014,R~2=0.64;RMSE=24.55 t/ha.In 2009,the average of the estimates was higher than the calculation result from the survey data(prediction:81.59 t/ha,measured:78.64 t/ha);and the 2014 inversion mean was lower than the survey data(prediction:79.63 t/ha;Measured:82.48 t/ha).From the regional scale,the average forest AGB in 2009 was 88.33 t/ha;and 94.61 t/ha in 2014;the average AGB growth was 6.28 t/ha.(4)Biome-BGC model could precisely simulate the changes of NPP.Based on the Biome-BGC model and long-term sequence(such as from 2003 to 2012)of multi-source data,after the optimization of the model parameterization scheme,the research simulated the forest productivity(NPP)and verified the results of NPP based on the spatially distributed tree core data.The results showed that the Biome-BGC model could simulate the changes of NPP precisely.The study area presents spatial differences of NPP between the north(higher)and the south(lower),and that due to the forest fires in the central and eastern China,the average NPP value in this part is relatively low.The NPP values of different forest types were significantly different.The average annual NPP values were in the order of coniferous forest>broad-leaved forest>mixed forest.The interannual variation of NPP in the growing season was more responsive to temperature and solar radiation than the response to precipitation and relative humidity.Further analysis of the interannual variation of forest type NPP and meteorological factors in the growing season of the study area revealed that the temperature was the major climatic factors that play a leading role in the changes of regional NPP;and the main control factor of coniferous forests is the solar radiation.Based on the multi-remote sensing data and observation data,this research formed a rapid and all-around preliminary scheme of multi-source data integrated monitoring on the four major parameters of area change of forests,vegetation coverage,above-ground biomass and forest NPP,and obtained a highly efficient and accurate result.Meanwhile,the research provided a key reference scheme for dynamic monitoring of forest resources. |
PDF Full Text Request