“In remote sensing scientific research, the combination of ground observations and remote sensing data drives deep learning and innovates the paradigm of land surface parameter inversion. However, there are representative errors in ground observations, and the key is to obtain ground truth values that are infinitely close to the scale of remote sensing pixels. Experts discuss the key steps in obtaining ground truth at the pixel scale, explore challenges and solutions, and provide new insights and theoretical guidance for improving the accuracy and precision of the acquisition.”