Search

The possibility and advantages of assessing the uncertainty of individual boundary existence and location from statistical models relating it to neighboring geometry (polygon area, shape, and line length) and forest-type attributes is examined. Unlike traditional methods of spatial uncertainty assessment that are based on a comparison of multiple (costly) map realizations, the proposed method is based on a single map. However, the reference spatial uncertainty for the models isdetermined from an approach that employs multiple map realizations. Results of the proposed method explain 27 percent of the variance in boundary location and 61 percent of the boundary existence. The best model for boundary location uncertainty predicts boundary width both as a function of boundary length and the shape of neighboring polygons, while the best model for boundary existence is based on boundary length, polygon shape and area, and change in species composition and height attributes across the boundary. [ABSTRACT FROM AUTHOR]

In the creation of forest type maps, photointerpreters first identify features that are the most certain, and subsequently identify features that are less certain in order to complete the map. In the present study, a method of photointerpretation is proposed in which only those features that an interpreter believes to be -100 percent certain#are identified. This results in interpretations that are not polygonal and that do not cover an entire photograph but that are instead treated using spatial interpolation to produce surfaces containing fuzzy membership values (FMVs) for all map classes at all locations in the area under study. These FMV surfaces can then be manipulated to produce thematic maps in which areas have a known certainty. The method is demonstrated using a synthetic image. It is concluded that maps resulting from this alternative method of photointerpretation are no more or less accurate than those produced from conventional polygon-based photointerpretation procedures. However, the alternative method provides considerable time savings and also provides an estimate of thecertainty of each area on the final thematic map -- something that the conventional photointerpretation method does not produce. [ABSTRACT FROM AUTHOR]

Abstract: Soil Surface Roughness (SR) provides a representation of surface variability which can be an important factor in a range of modelling applications such as surface water flow and sediment/nutrient transport. Moreover, it is a crucial parameter for interpreting backscatter characteristics of Synthetic Aperture Radar (SAR) for agricultural application such as near-surface soil moisture retrieval. SR is typically estimated using manual profiles of height variation along short transects (1 to 3m). However, this approach can be very time consuming and often only a small number of transects can be measured in this way, which may not be adequate to characterize the spatial variability of SR across the landscape. This study investigated the feasibility of utilising airborne Light Detection and Ranging (LiDAR) observations as an alternative for mapping SR attributes across an agricultural environment in New South Wales, Australia. To that end, SR attributes were extracted from airborne LiDAR observations and compared with those extracted from an extensive ground survey of SR making use of manual pin-profilers. Results show that LiDAR-estimates of soil profile surface heights Root Mean Square (RMS) are both accurate (compared to manual profiles) and precise (repeatable stable estimates) for fields presenting bare or fallow conditions and either presenting no row structure or as long as the orientation of the LiDAR scan line is perpendicular to the row structure. In such cases results indicated a strong correlation between LiDAR-estimated and ground-measured RMS estimates (R2 >0.68, p<0.05), with an RMSE better than 0.81cm and bias smaller than 0.48cm from a 400m flight altitude. Moreover, estimates produced from repeat pass LiDAR datasets were consistent and highly correlated (R2 0.98) suggesting that the approach is precise and robust, provided that key tillage parameters (i.e. presence of vegetative material and row direction) can be pre-classified. LiDAR estimates of surface height RMS were shown to be accurate enough to allow the tracking of temporal changes in surface roughness due to farming activities. In contrast, LiDAR-derived surface Correlation Length (CL) estimates were not found to be a reliable proxy of the ground-measured CL. [Copyright &y& Elsevier]