Northeastern Section - 48th Annual Meeting (18–20 March 2013)

Paper No. 2
Presentation Time: 2:10 PM


OLSON, Neil F., New Hampshire Geological Survey, PO Box 95, 29 Hazen Drive, Concord, NH 03302-0095,

Current maps of stream networks commonly underrepresent small headwater streams. These small streams are dynamic and important sites of hydrological and biological processes therefore their inclusion in estimates of river network length and area is important to understand and estimate river processes at the whole network scale. Unlike typically used methods for mapping streams such as flow accumulation thresholds which infer the location of streams, high-resolution digital elevation models (DEM) allow for direct detection of stream channels. A test of previously developed methods was applied in three watersheds encompassing 110 mi2 in coastal New Hampshire. The elevation ranges between 511 ft and 2 ft and land cover ranges from undeveloped forest to sub-urban development. Image-processing techniques known as morphologcial filters were applied to a Light Detection and Ranging (LIDAR) derived DEM, specifically Closing and Opening in combination. This analysis highlighted the geomorphic signatures of headwater streams. These features were traced downstream and converted to a stream network which was subsequently verified using aerial photography and field survey. The resulting stream network revealed greater detail than previously mapped National Hydrography Dataset (NHD) streams for the same area. The extracted network represented a 164% increase in total stream length over the three watersheds compared to NHD streams. Field visits to randomized sites on the extracted network resulted in only 2 false positive results out of a total 28 sites, where there was no stream on the ground although it was predicted by the model. Using a lower threshold for stream detection in the model, 3 false positives out of 18 total sites were detected. In addition,1 false negative was detected out of 34 random sites not on the extracted network but within the flow accumulation network which had flow present where it had not been predicted.. Including the 18 lower threshold sites, an additional 5 false positives were present. These methods are promising in their potential to objectively define streams in a wide variety of settings, including under tree canopy cover and in built environments.
  • Neil_Olson_LiDAR_Stream_NEGSA_2013.pdf (6.1 MB)