Development of a Gridded Observational Dataset for the Post-Processing of Simulations According to Hydrological Performance

Quebec was hit by major floods in 2017 and 2019, which demonstrated the need to better understand the risks of flooding in rivers in southern Quebec. In this context, the Government of Québec wants to be able to demarcate areas to support land use planning that takes climate change into account. 

Simulations of future flow are based in part on post-processed climate simulations. Grids of interpolated observations, in particular those of NRCan and the air quality and climate division of the MELCCFP, are typically used in the post-processing of climate simulations, the calibration of hydrological models and the reconstruction of historical flows. Many problems caused by these grids have been seen in recent years, including the underestimation of precipitation extremes.

Main objective: Produce a historical meteorological dataset (1970-2018) on a daily grid combining data from various sources (observations, reanalyses, satellite data)

Specific objectives: 

• Draw up an exhaustive list of available climate products

• Propose an approach combining observations with data from other climate products

• Evaluate the climate performance of the products mentioned above

• Evaluate the hydrological performance of the products mentioned above

• Develop an adapted version of the methodology that was created for the production of the MSWEP precipitation grids, combining data from multiple sources.

• Develop an algorithm for the production of temperature grids combining data from multiple sources.

• Create daily grids of precipitation, minimum temperatures and maximum temperatures. The dataset, named GRNCH-QC, covers the time period from 1970 to 2019 at a spatial resolution of 0.10.

• Evaluate the climate performance of GRNCH-QC by comparing it, according to various indices, to the grids from the air quality and climate division of the MELCCFP, to the MSWEP V2 grids and to the individual grids integrated into GRNCH-QC.

• Evaluate the hydrological performance of GRNCH-QC for the calibration and validation of GR4J and HMETS.

Table 1 below provides an overview of the climate products used to produce GRNH-QC. These include station data (S), reanalysis data (RA), and satellite data (SAT). Given the variation in time coverage between the different products, GRNCH-QC was divided into four sub-periods (P1: 1970-80, P2: 1980-2000, P3: 2001-17, P4: 2018-19). The temperature products considered during each of the sub-periods appear in red and the precipitation products considered appear in blue. The P3 period is the one with the greatest diversity of climate products.

Figure 1 below shows a comparison between the daily precipitation time series observed at 40 stations in Quebec and the GRNCH-QC, GCQ (climate grids from the air quality and climate division of the MELCCFP) and MSWEP V2 datasets. (The series are taken at the grid points corresponding to the 40 stations for these three datasets.) Note that for the purposes of this comparison, the 40 stations were removed from the production of GRNCH-QC, although they are obviously included in GCQ and, possibly, in MSWEP v2 (it is not possible to verify this). Three indices are used: the correlation, the root mean square error and the ratio of variances. Each of the box-and-whisker plots shows the distribution of the results for the 40 sites (grid points/stations). The results show that GRNCH-QC surpasses MSWEP and offers performance that is close to that of GCQ (which includes all 40 stations).

Figure 2 below provides an overview of the results obtained (validation ongoing) with the HMETS and GR4J hydrological models in 105 watersheds in Quebec. The calibration metric used is the KGE. Each point has a KGE difference depending on whether GRNCH-QC or GCQ data is used. The predominance of the colour red shows the advantage of GRNCH-QC, especially in watersheds located in regions where data at the stations is scarce, such as in northern Quebec and Labrador.

This project was funded by Ouranos as part of the support for the MELCC's INFO-Crue initiative.

703800