Mark Quilter ran the code for his study area and found that the metric_pixel_points.py placed a 'cold' calibration point in a riparian area and a 'hot' calibration point on a hillside. Seems that the code may not be excluding non-ag locations for calibration points.