Spatio-temporal precipitation error propagation in runoff modelling: a case study in central Sweden

Abstract

Abstract. The propagation of spatio-temporal errors in precipitation estimates to runoff errors in the output from the conceptual hydrological HBV model was investigated. The study region was the Gimån catchment in central Sweden, and the period year 2002. Five precipitation sources were considered: NWP model (H22), weather radar (RAD), precipitation gauges (PTH), and two versions of a mesoscale analysis system (M11, M22). To define the baseline estimates of precipitation and runoff, used to define seasonal precipitation and runoff biases, the mesoscale climate analysis M11 was used. The main precipitation biases were a systematic overestimation of precipitation by H22, in particular during winter and early spring, and a pronounced local overestimation by RAD during autumn, in the western part of the catchment. These overestimations in some cases exceeded 50% in terms of seasonal subcatchment relative accumulated volume bias, but generally the bias was within ±20%. The precipitation data from the different sources were used to drive the HBV model, set up and calibrated for two stations in Gimån, both for continuous simulation during 2002 and for forecasting of the spring flood peak. In summer, autumn and winter all sources agreed well. In spring H22 overestimated the accumulated runoff volume by ∼50% and peak discharge by almost 100%, owing to both overestimated snow depth and precipitation during the spring flood. PTH overestimated spring runoff volumes by ∼15% owing to overestimated winter precipitation. The results demonstrate how biases in precipitation estimates may exhibit a substantial space-time variability, and may further become either magnified or reduced when applied for hydrological purposes, depending on both temporal and spatial variations in the catchment. Thus, the uncertainty in precipitation estimates should preferably be specified as a function of both time and space.

Twelve year interannual and seasonal variability of stream carbon export from a boreal peatland catchment

“…The following seasonal classification was used: winter (January, February, March); spring (April, May, June); summer (July, August, September); and autumn (October, November, December). This seasonal classification has been used in other Swedish hydrologic studies [e.g., Olsson , ] and captures variability in the annual hydrologic regime (winter: low flow and typically ice covered, spring: high flow snowmelt, summer and autumn: low flows with high flow events associated with rain events), as well as differences in plant productivity, particularly between the summer and autumn periods [ Peichl et al , ]. For each year and season we calculated carbon export and mean flow‐weighted concentrations of TOC, DIC and CH 4 ‐C.…”

Propagation of uncertainties in interpolated rainfields to runoff errors

“…They concluded that the magnitude of errors in runoff (volume and peak) are nearly double that of rainfall volume errors for a 21.2-km 2 catchment, runoff errors amplifying as the total rainfall volume became smaller. Olsson (2006) considered five rainfall sources and used one as the true source, its associated runoff generated by a rainfall-runoff model also considered the true runoff. He observed discharge bias of up to 79% for catchments ranging between 134 and 2133 km 2 using a 1-day simulation timescale.…”