Towards an improved land surface scheme for prairie landscapes

- We developed a new surface runoff algorithm (PDMROF) for prairie landscapes.
- We compared performance of PDMROF with a slope-based runoff algorithm (WATROF).
- PDMROF performed significantly better than WATROF at simulating stream flow.
- PDMROF captured the dynamic nature of contributing areas in a parsimonious manner.

SummaryThe prairie region of Canada and the United States is characterized by millions of small depressions of glacial origin called prairie potholes. The transfer of surface runoff in this landscape is mainly through a “fill and spill” mechanism among neighboring potholes. While non-contributing areas, that is small internally drained basins, are common on this landscape, during wet periods these areas can become hydrologically connected to larger regional drainage systems. Accurate prediction of prairie surface runoff generation and streamflow thus requires realistic representation of the dynamic threshold-mediated nature of these contributing areas. This paper presents a new prairie surface runoff generation algorithm for land surface schemes and large scale hydrological models that conceptualizes a hydrologic unit as a combination of variable and interacting storage elements. The proposed surface runoff generation algorithm uses a probability density function to represent the spatial variation of pothole storages and assumes a unique relationship between storage and the fractional contributing area for runoff (and hence amount of direct runoff generated) within a grid cell. In this paper the parameters that define this relationship are obtained by calibration against streamflow. The model was compared to an existing hydrology-land surface scheme (HLSS) applied to a typical Canadian prairie catchment, the Assiniboine River. The existing configuration is based on the Canadian Land Surface Scheme (CLASS) and WATROF (a physically-based overland and interflow scheme). The new configuration consists of CLASS coupled with the new PDMROF model. Results showed that the proposed surface runoff generation algorithm performed better at simulating streamflow, and appears to capture the dynamic nature of contributing areas in an effective and parsimonious manner. A pilot evaluation based on 1 m LiDAR data from a small (10 km2) experimental area suggests that the shape of the modeled storage-contributing area relationship is broadly consistent with that inferred from terrain analysis, under certain simplifying assumptions. The direct identification of storage-runoff parameters from terrain analysis is an outstanding challenge, and a promising area for future research.