行播作物地面方向性测量的视场不确定性分析

Footprint Uncertainty Analysis for Ground-based Multiangular Measurement of Row Crops

行播作物以其独特的几何结构介于离散与连续植被之间。地面测量此类地物的双向反射系数（Bidirectional Reflectance Factor，BRF）特征，不可回避视场变化所引起的不确定性问题。在Kimes垄行结构模型中加入等效视场的概念，对视场进行分解，从而建立了一个行结构多角度地面测量的视场不确定性分析模型，为定量分析视场变化所引起的BRF测量误差提供了可能。利用该模型较为全面地模拟分析了视场变化对视场内四组分比例及冠层BRF的影响。结果表明：①BRF误差基本独立于植被-土壤光谱对比度。②误差与观测天顶角之间的关系复杂，前向观测表现得尤为明显。③垂直观测视场满1个垄周期后，四组分比例及冠层BRF的误差可保持较小且稳定的状态；满2个垄周期，误差达到局部最小值（局部指垂直视场含2．5个垄周期以下，不排除视场更大，误差更小的可能性）。④垂直视场若仅含0．5个垄周期，BRF误差最大值一般可高达67．8％，最小值亦可达38．7％；满1个垄周期后，BRF误差极大值降至20％以下，极小值可控制在6％以内。其中视场为1个整周期，误差范围为6％～12％；2个整周期，误差范围为0．6％～3．9％。⑤垂直视场大小为1～2个垄周期之间的非整周期，四组分比例及冠层BRF误差总体上均稍高于1个整周期，故建议在实际测量过程中，测量高度若无法满足垂直视场为2个垄周期，可优先考虑1个整周期的情况。还通过非线性最优化函数将2个模型分别与黑河实验玉米地方向性观测实测数据进行拟合，得出的结果与模拟分析的结论一致，即在垂直视场内包含2个垄周期以上的生长初期，方向性测量无需考虑视场效应；若垂直视场内不足一个垄周期（生长中期），则有必要考虑视场的不确定性。

Row crop can be classified between homogeneous and heterogeneous canopy due to its own special geometric characteristics. The footprint uncertainty problem of this kind of canopy should not be neglected in the field directional measurement. This study introduces an equivalent footprint of sensor＇s field of view into the original Kimes model and develops a footprint uncertainty analysis model for multiangular in situ measurements of row crops through disassembling the equivalent footprint. Both the influences of footprint uncertainty on the four component proportions （ i. e. the proportion of sunlit and shaded vegetation, and sunlit and shaded soil） and canopy BRF have been analyzed ： （1）BRF error is nearly independent of the spectral contrast between vegetation and soil. （2）The relationship between BRF error and view zenith angle is very complicated with the forward observation worse than the backward. （3）Both the errors of four component proportions and canopy BRF keep relatively small and steady after the nadir footprint reaches more than 1 row period. When the nadir footprint is 2 row periods, both the errors arrive at local minimum. Here the ＇ local＇ means the error might be smaller when the nadir footprint contains more than 2.5 row periods which is beyond the discussion of this study. （4）The relative mean error range of BRF is 38.7% 67.8% when the nadir footprint contains only half row period. This range changes into 6% - 12% and 0. 6% 3.9% when the nadir footprint contains 1 and 2 row periods respectively. （5）Errors are somewhat higher when the nadir footprint contains non-integral period which is between 1 and 2 row periods than it contains just 1 period. So considering the one period situation before the nadir footprint reached 2 periods in the field experiment is the suggestion. This study also compares the proposed and the original Kimes model to the field directional observation of corn canopy in the Heihe River basin based on a multivariate constrained nonlinear optimization technique. The results are consistent with the simulation conclusions above, which include that the footprint uncertainty problem can be ignored when the nadir footprint reached more than 2 row periods and it＇s necessary to consider the problem if the nadir footprint contains less than 1 period.