A hybrid visual estimation method for the collection of ground truth fractional coverage data in a humid tropical environment

A substantial body of research exists exploring the spectral unmixing of remotely sensed image data. Specifically, we refer to the attempts and successes to model the percent vegetation cover (2-dimensional horizontal density) within a pixel, known as fractional coverage (fc). With this paper, we present a hybrid visual estimation method for fc field data collection in the complex landscapes found in humid tropical environments. The method includes a scalable theoretical model of fc, integrates the visual estimation technique with hemispherical photography collection, and is conducted over a systematic ground collection area. We present results from a case study conducted in the humid tropical region of Ecuador. Specifically, we report on the relationship between fc data modeled using a linear NDVI transformation and observed fc data collected using our hybrid visual estimation method. udy found a significant, positive linear relationship (β = 0.795, r2 > 0.84, and p < 0.001) between modeled and observed fc values. Because the accuracy of both modeled and observed values are unknown, a full validation of the proposed method of collection is not possible. Therefore, we conduct an error assessment, identifying limitations in the modeling method (e.g., non-linear relationship between modeled and true values and potential for saturation) and hybrid ground-truth collection method (e.g., subjectivity of visual estimation and positional errors in the ground collection area) that explain the deviation from a 1:1 relationship. We believe the proposed method of ground truth data collection is a significant contribution towards efforts to validate biophysical information gained from remotely sensed data.