Curating 62 years of Walnut Gulch Experimental Watershed dataset:
Improving QAQC methods of rainfall and runoff measurements.
Abstract
The Walnut Gulch Experimental Watershed (WGEW) is the primary outdoor
hydrologic laboratory for the USDA-ARS’ Southwest Watershed Research
Center (SWRC). This site represents the Southwest semiarid environment
within the Long-term Agroecosystem Research (LTAR) network. The SWRC
maintains a collection of long-term hydro-climatic measurements from
WGEW, featuring an extensive archive of rainfall and runoff observations
from an ephemeral network of streams within the 149 km2 watershed. The
WGEW was established in 1953, and has continually developed and improved
quality assurance and quality control procedures to aid in the accuracy
and curation of the constantly growing datasets obtained from more than
100 rain gauges and 18 flumes, weirs, and gauged ponds. These efforts
have led to the development of a state-of-the-art database and data
visualization tools to aid in the curation of research-grade
hydrometeorologic datasets. This required development of automated
quality assurance and quality control tools to check and maintain the
data for 21st century research needs. We developed five tools to improve
the quality of rainfall and runoff database based on conventional
hydrologic principles and the relationships between them: 1)
precipitation is spatially correlated; 2) there is a temporal relation
between rainfall and runoff; 3) runoff is only a limit portion of
rainfall; and, 4) closer took of extreme events. Hence, we developed the
following methods that included the analysis of interpolated rainfall
maps at a daily time step, the association between rainfall and runoff
events, lag time, runoff coefficients, and multiple regression methods
to identify problematic events in the data archive. To visually inspect
and verify the errors, we developed a graphical tool that displays
relevant event hyetographs and hydrographs within a specific time
window. After flagging anomalous events, we evaluated the types of
errors using the original records and metadata information. The
implementation of these approaches resulted in developing a suite of
semi-automated QAQC tools that correctly detected 813 rainfall and 24
runoff events with erroneous timestamps that had passed all previous
quality checks.