| 17 Apr 2026
The Canadian Surface Reanalysis (CaSR) v3.2 precipitation dataset: A 45-year high-resolution analysis for North America (1980–2024)
Abstract. The Canadian Surface Reanalysis (CaSR) includes an offline, high-resolution gridded precipitation reanalysis designed to provide accurate estimates across North America. This product, referred to as CaPA-24h, builds on the Canadian Precipitation Analysis (CaPA) system of Environment and Climate Change Canada (ECCC). It combines a dense network of daily surface observations with a background field from the CaSR dynamical component, using updated quality-control and assimilation procedures to filter spurious observations. This study evaluates the CaPA-24h fields produced in CaSR v3.2, together with their background field, and compares them with the previous version (v2.1) as well as with two independent datasets, ERA5-Land and PRISM. Results show substantial improvements in v3.2, particularly in data-sparse regions, with an enhanced representation of precipitation events of different intensities. Compared to ERA5-Land, CaPA-24h v3.2 provides more accurate seasonal and regional precipitation patterns, while evaluations against PRISM confirm this improved performance. However, biases persist in southern and western mountainous areas, especially for orographic precipitation. A first-time assessment of the hourly disaggregated product reveals limitations in the diurnal cycle representation, indicating the need for refined disaggregation methods and background field generation. Overall, CaPA-24h v3.2 delivers a reliable and well-established gridded precipitation dataset, offering a valuable resource for hydrological, climatological, and impact studies across North America.
This paper, submitted by Khedhaouiria et al., explains the methodology behind the recent Canadian Surface Reanalysis (CaSR) v3.2 precipitation dataset for North America and assesses its overall quality
General comment:
Developing a climate reanalysis is a task that is both meticulous and far-reaching, given the sheer volume of supporting data and the many pitfalls involved. Not to mention that Canada is a vast country, large parts of which are under-instrumented, requiring a tailored strategy. The work presented here, which focuses on total precipitation, is the result of ongoing improvements to Canadian hydroclimatic products. These products are particularly useful to the operational hydrology community.
Major comments:
It would have been helpful to state clearly from the outset that the database compiled relates to total precipitation, without attempts to identify the phase of the precipitation. I do not dispute the decision not to distinguish between liquid and solid precipitation, but rather the decision not to state this clearly or to justify it.
Section 6.1 – The manuscript draws on numerous databases, and their use is not always clear to the reader. This is particularly the case here, where numerous databases are evaluated and compared without specifying the source of the observations used as a reference. We must assume that these are the ones described in Figure 1, but they are then described as having been used for assimilation, not for evaluation. The same applies in Appendix A2: Details of assimilated data sets. In fact, the information is provided further on, in Section 6.2, line 364.
Section 6.4 – If I understand correctly, metrics are calculated for each grid point for the 17-year period from 2002 to 2018 in the PRISM database. Since the selected indicators are annual values, the time series for each point consists of 17 values. I am not sure that it is convincing to present KGE on such short time series. In this respect, in Figure 9, I have more confidence in the assessment of bias and correlation than in that of variability and therefore the KGE. Personally, I would not present the KGE and variability. It seems to me that the issue of the length of the time series deserves at least a detailed comment. What is written on line 482 seems insufficient to me.
Figure 11 – The diurnal cycle is essentially a local process. Is there an issue with combining values across large areas? Furthermore, the decision to present times in UTC complicates interpretation, as not everyone is familiar with how to perform the conversion.
The work presented here is partly motivated by hydrological applications. Indeed, the success of hydrological models depends largely on the quality of historical precipitation data – although this is not the only challenge they face. However, this precipitation alone is insufficient. At a minimum, an air temperature database is also required to clarify the phase of precipitation and to model the melting of snow accumulated on the ground – essential components of hydrology in a northern climate such as Canada’s. Although this is not the focus of this manuscript, it would be useful for potential users if this issue were addressed in the discussion or conclusion, for example by suggesting air temperature databases (and why not other hydrometeorological variables) compatible with CaSR v3.2.
Minor comments:
Line 50 – Please write out ‘NWP’ in full when it first appears
Line 135 – A brief note on the annual fluctuation in Figure 1 would be helpful. The explanation comes much later in the text, at lines 599 and 617. Also, Figure 1 would be easier to interpret if the y-axis started at zero.
Line 300 – The equations describing these metrics could be placed in the appendix with the others already there.
Figure 14 – I am not sure that this figure is strictly necessary.
Ligne 645 – Is it possible to provide an appreciation of the number of stations provided to the study by these different networks?