the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
The High-resolution Intermediate Complexity Atmospheric Research (HICAR v1.0) Model Enables Fast Dynamic Downscaling to the Hectometer Scale
Dylan Stewart Reynolds
Ethan Gutmann
Bert Kruyt
Michael Haugeneder
Tobias Jonas
Franziska Gerber
Michael Lehning
Rebecca Mott
Abstract. High resolution (< 1 km) atmospheric modeling is increasingly used to study precipitation distributions in complex terrain and cryosphere-atmospheric processes. While this approach has yielded insightful results, studies over annual time-scales or at the spatial extents of watersheds remain unrealistic due to the computational costs of running most atmospheric models. In this paper we introduce a High-resolution variant of the Intermediate Complexity Atmospheric Research (ICAR) model, HICAR. We detail the model development that enabled HICAR simulations at the hectometer scale, including changes to the advection scheme and the wind solver. The latter uses near surface terrain parameters which allow HICAR to simulate complex topographic flow features. These model improvements clearly influence precipitation distributions at the ridge scale (50 m), suggesting that HICAR can approximate processes dependent on particle-flow interactions such as preferential deposition. A 250 m HICAR simulation over most of the Swiss Alps also shows monthly precipitation patterns similar to two different gridded precipitation products which assimilate available observations. Benchmarking runs show that HICAR uses 118x fewer computational resources than the WRF atmospheric model. This gain in efficiency makes dynamic downscaling accessible to ecohydrological research, where downscaled data is often required at hectometer resolution for whole basins at seasonal time scales. These results motivate further development of HICAR, including refinement of parameterizations used in the wind solver, and coupling of the model with an intermediate complexity snow model.
Dylan Stewart Reynolds et al.
Status: open (until 11 Apr 2023)
-
RC1: 'Comment on gmd-2023-16', Anonymous Referee #1, 07 Mar 2023
reply
-
RC2: 'Comment on gmd-2023-16', Junhong Lee, 12 Mar 2023
reply
This paper describes the improvement of dynamic downscaling model by introducing terrain descriptors, wind solver, and asynchronous I/O. The authors present the method of implementation well and its evaluation. Results show realistic wind and precipitation fields and astonishing computational efficiency.
The paper is well aligned with the model description of GMD, and shows a useful method to downscale wind and precipitation to the hectometer scale. It is well written, but the following points need to be addressed and I encourage its publication following a minor revision.
General comments.
Section 2.3: To be honest, I don’t fully understand how authors get the x118 speedup relative to the WRF model. HICAR employs RK3 integration to solve advection in addition to the physics parameterizations. But, Mielikainen et al. (2014) in GMDD (https://doi.org/10.5194/gmdd-7-8941-2014) showed that dynamic core takes 61.93% of the computation time of WRF. If I take into account the cumulus scheme in addition to dynamics, I expect the integration speed to increase roughly by a factor of 2 at maximum. Is it because of the difference in the compiler?
Line 252- 255: What is the rationale for using the YSU PBL scheme? Is it recommended to turn off the PBL scheme at the hectometer scale? Also, what do you use for the surface-layer scheme?
Line 398-399: Why do you see wet bias over the entire domain in the WRF model? Is it coming from parent domains?
Section 4.2: How much are snow and precipitation different between ICAR and HICAR? It would be very nice to have a discussion on this.
Specific comments.
Eq #1: What is the difference from the Laplacian operator?
Line 249: “schaer2002,Kruyt2022” - what are those?
Line 255: Add original citation of the YSU PBL scheme, Honeg et al. 2006.
Line 269: Please add the full name and citation for the COSMO model.
Section 3.2: More information is desirable. At least, the horizontal resolution of simulation data would be very helpful.
Section 3.4: The observation data section is in-between model configurations. I would place this section after section 3.5.
Section 4.2.2: Need consistency in the way to refer to OSHD OI, such as OSHD precipitation, OI product, or OSHD OI. Same for section 3.4.
Line 470: Did you compare HICAR to WRF compiled with the Cray compiler? The same compiler should be used to be a fair comparison.
Section 5: I find it would be nice if the authors discuss the possibility to use their model in the wind energy field. This is not a demand, merely a wish.
Citation: https://doi.org/10.5194/gmd-2023-16-RC2
Dylan Stewart Reynolds et al.
Model code and software
HICAR Source Code Dylan Reynolds, Ethan Gutmann, and Bert Kruyt https://github.com/d-reynolds/HICAR
Dylan Stewart Reynolds et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
295 | 78 | 10 | 383 | 7 | 5 |
- HTML: 295
- PDF: 78
- XML: 10
- Total: 383
- BibTeX: 7
- EndNote: 5
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1