TO BE (ACCURATE) OR NOT TO BE (ACCURATE): CONSTRAINING ERROR OF BOULDER MEASUREMENTS BETWEEN GOOGLE EARTH AND PHOTOGRAMMETRY IN THE EASTERN HIMALAYA

With remote sensing becoming more useful and Google Earth (GE) being one of the more accessible tools, it is important to understand how close measurements taken on GE are to real-life measurements. This is particularly important for locations with limited access, like the eastern Himalaya. In this region, there are over a hundred boulder bars, which are potentially linked to megaflooding or regional landsliding. Before we can interpret the impact that these bars have on the region, we must first measure the boulders in them, however, it is impossible to access most of these bars. We combine field data and remote GE imagery from one boulder bar near Tuting on the Yarlung-Siang River to constrain the uncertainty between the two measurements. We collected images to use for photogrammetry and created a structure from motion (SfM) model of a swath down the center of the bar. We measure the same boulders in both the SfM model and GE images to compare measurements and constrain uncertainty. In calibrating our uncertainty, we measured each boulder multiple times in each data set. We found that between these two data sets, uncertainty is +/- 5.5 cm or within 0.5-10% of each measured boulder. As the grain size increases, the percent error decreases. Once the grain size reaches >1.5 m, the percent error reduces to 5% and continues to decrease as the grain size increases. We believe this uncertainty comes from two main places: the quality of the GE images and human error. Shadows from steep topography, cloud cover, and high channel sinuosity could reduce the quality of these images, increasing the uncertainty between these two measurements. To reduce human bias and error and to make our methods replicable and systematic, we randomly selected each boulder to measure using a grid system and only had two people measuring boulders. While both of these sources of uncertainty will affect the accuracy of measurements from GE, the small percent error (esp. for boulders >1.5m) allows us to have confidence that our remote GE measurements of boulder bars in the eastern Himalaya accurately represents the size of boulders in this region. To further minimize error, we would suggest a return to the field and gather physical measurements of boulders with a range of grain sizes giving us physical data to fine-tune our error and allow us to look at a broader range of grain sizes.