Inversion Of Forest Parameters Using Microwave Remote Sensing Data

PolInSAR data are quite sensitive to geometrical morphology and vertical variation of vegetation.PolInSAR technology and datasets are used to estimate tree heights as well as other forest parameters.Currently,the studies are mainly on flat ground areas.Inversion tree height on the sloped plateau area is still relying on the light detection and ranging(Lidar)data and ground survey.Both approaches are not applicable to inventory forested areas with large spatial extent.The objective of this study is to develop an inversion algorithm of the tree height based on the weight of coherence coefficients,and to apply the method to sloped plateau area to assess its effectiveness and efficiency.The major research tasks and innovation include follows.1)Tree heights,the location of each tree(latitude,longitude,and elevation),and tree species of ten 30 m × 30 m plots in the study area,Zoige,China were acquired through fieldwork in 2015.Field data processing and analysis were completed.Two common tree height inversion algorithms converting interferometric phase to height and subtracting interferometric phase were applied in study area respectively.The results are compared with measured tree heights.The algorithm that converts interferometric phase to tree height,inevitably magnifies the height value twice in the conversion of the phase data to tree heights.The inverted heights are much taller than the measured one.The subtracting interferometric phase algorithm depends greatly on the position where the electromagnetic waves intercept with the tree canopy.In reality,this position is most likely located beyond the canopy.Thus,the inversion height can be much shorter than the measured one.With the underestimation in this algorithm,and the overestimation in the previous one,therefore,the combined use of both algorithms linearly is studied next.2)The tree height is the linear combination of the tree heights estimated by both above algorithms.The weights in the combined algorithm are based on the coherence values.The inverted tree heights of plots 1,3,5,6,7,8,9,and 10 are 16.6,23.0,21.7,36.3,23.9,13.7,27.9,and 20.1 m,respectively.The absolute errors are 13%,17%,36%,32%,12%,29%,22%,and 12% accordingly.Significant improvement in tree height estimation has been achieved when compared to results using each algorithm individually.3)Analyzing the inverted results by the slope values and vegetation types.The estimated tree heights in plots 2 and 4 are still much larger than the measured ones.The steep slope(? 39.5° of inclination angle)is attributed to the cause.With the increase of the slope,the topographic impact(foreshortening and overlay)increases greatly.According to the field observation,plot 6 consists of primarily deciduous trees,and the inverted height is of an absolute error of 32% when compared with the measurement.In other plots where the predominant trees are conifers,the absolute error is around 19% or less.The vegetation type can adversely affect the height estimation.Therefore,the classification of forested areas into different types such as(coniferous,deciduous,and mixed categories)should be helpful and be considered.