the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Evaluation of reanalysis soil moisture products using Cosmic Ray Neutron Sensor observations across the globe
Yanchen Zheng
Gemma Coxon
Ross Woods
Daniel Power
Miguel Angel Rico-Ramirez
David McJannet
Rafael Rosolem
Jianzhu Li
Ping Feng
Abstract. Accurate soil moisture information is vital for flood and drought predictions, crop growth and agricultural water management. Reanalysis soil moisture products with multi-decadal temporal coverage are gradually becoming a good alternative for providing global soil moisture data in various applications compared to in-situ measurements and satellite products. Much effort has been devoted to evaluating the performance of soil moisture products, yet the scale discrepancy between point measurements and grid cell soil moisture products limits the assessment quality. As the land surface and hydrological modelling community evolve towards the next generation of (sub)kilometer resolution models, Cosmic Ray Neutron Sensors (CRNS) that provide estimates of root-zone soil moisture at the field scale (~250 m radius from the sensor and up to 0.7 m deep), may consequently be more suitable for soil moisture product evaluation as they cover a relatively larger footprint, when compared to traditional methods. In this study, we perform a comprehensive evaluation of seven widely-used reanalysis soil moisture products (ERA5-Land, CFSv2, MERRA2, JRA55, GLDAS-Noah, CRA40 and GLEAM datasets) against 135 CRNS sites from the UK, Europe, USA and Australia. We evaluate the products using six metrics capturing different aspects of soil moisture dynamics. Results show that all reanalysis products exhibit good temporal correlation with the measurements, with the median of temporal correlation coefficient (R) values spanning from 0.69 to 0.79, though large deviations are found at sites with seasonally varying vegetation cover. Poor performance is observed across products for soil moisture anomalies timeseries, with R values varying from 0.49 to 0.70. The performance of reanalysis products differs greatly across regions, climate, land covers and topographic conditions. In general, all products tend to overestimate in arid climates and underestimate in humid regions as well as grassland. Most reanalysis products perform poorly in steep terrain. Relatively low temporal correlation and high Bias are detected in some sites from west of the UK, which might be associated with relatively low bulk density and high soil organic carbon. Overall, ERA5-Land, CFSv2, CRA40, GLEAM exhibit superior performance compared to MERRA2, GLDAS-Noah and JRA55. We recommend ERA5-Land and CFSv2 should be used in humid climates, whereas CRA40 and GLEAM perform better in arid regions. GLEAM is more effective in shrubland regions. Our findings also provide insights on directions for improvement of soil moisture products for product developers.
- Preprint
(1946 KB) - Metadata XML
-
Supplement
(1373 KB) - BibTeX
- EndNote
Yanchen Zheng et al.
Status: open (until 04 Dec 2023)
-
RC1: 'Comment on hess-2023-224', Anonymous Referee #1, 15 Nov 2023
reply
The authors present a noteworthy study that compares CRNS stations across multiple continents with remote sensing/reanalysis soil moisture products. Utilizing the CRSPY tool, the authors reprocess the raw CRNS data, addressing gaps in information, such as the atmospheric moisture correction (absent in the COSMOS USA network), and harmonize the dataset. Undoubtedly, this tool holds promise for advancing the CRNS method and enhancing the global CRNS community. The article is well-crafted, with comprehensive data documentation in both the appendix and online. However, I have some suggestions and comments that should be addressed before publication.
Major Suggestion:
I find some ambiguity regarding the reprocessing of CRNS data with CRSPY, particularly in the computation of the N0 parameter. The main article and supplemental material lack any mention of N0. To calculate N0, each site typically requires at least one gravimetric calibration survey, composed of 12-18 soil profiles within the footprint, sampled every 5 cm down to 30 cm. The gravimetric survey data is then weighted (following Kohli 2015/2017, etc.), and the Desilets 2010 function is inverted to determine N0. Most CRNS sites have gravimetric calibration data for one or more sample periods. Please include a detailed description of the process for estimating N0 at each site, specifying the gravimetric calibration dataset used (single or average of multiple calibrations, etc.). The accuracy of the N0 estimate is crucial, as incomplete or non-representative data may introduce significant bias in the comparison with reanalysis products. Best practices for N0 involve 2-3 calibration dates (Iwema 2015), though this is labor-intensive and not consistently implemented.
I am a bit confused by the description of soil organic carbon. Does this include both the organic carbon and mineral lattice water values? The variation in lattice water was found to be important for many of the original CRNS networks (Zreda 2012 COSMOS, Hawdon 2014 COSMOZ, etc.). The paper and appendix does not include the description of lattice water and should be clarified or added to the metadata.
Comment: The influence of rapidly growing crops on the CRNS observations remains a challenge. There have been several attempts to provide correction factors (either on N0 itself or on the moderated counts) but nothing definitive has been adopted by the community. The authors mention this in the limitations. I hope the community comes to a consensus soon about how best to deal with CRNS data in croplands. The influence of forest biomass seems to be an even more challenging problem but equally important. Â
Minor comment:
Figure 2. Label the 4 subplots a-d and list what geographical region they are from. This was not clear from the description.
Citation: https://doi.org/10.5194/hess-2023-224-RC1 -
RC2: 'Comment on hess-2023-224', Anonymous Referee #2, 20 Nov 2023
reply
The paper analyses the performance of global reanalysis soil moisture products against soil moisture obtained from cosmic ray neutron sensors. The results are analysed using different performance criteria, region specific and climate specific. The paper addresses highly relevent scientific questions, is exceptionally well structured, provides novel insights thereby presenting novel concepts and setting the global data sets into new context. The conclusions are reached, to the point and substantial.
However, there are two main improvements to be made before acceptance.
First: I suggest the authors consider to include SMAP Level 4 data as reanalysis product for evaluation. SMAP Level 4 comes as data assimilation product and is considered state-of-the-art soil moisture for the recent years. Without SMAP Level 4, the study seems incomplete.
Second: the terminology in Figure 10 as "highly recommended", "recommended" and "not recommended" must be adapted and if needed must be adapted throughout the text. All reanalysis products show high bias, and hence, the ERA5-Land cannot be highly recommended for humid sites as the authors have done in Figure 10. It should rather be recommended with care. I provide a suggestion below.
Â
Â
Minor comments:
Line 11: It is recommended to shorten the abstract. Lines 11-19 can be reduced to 3 lines. This will focus the reader on the methods, novelty, and key findings. These are mentioned in the abstract.ÂLine 21: Although UK has its own network, it reads awkwards to read "sites from UK, Europe, USA and Australia" as a list because the UK is included in Europe and should not be listed beside Europe as separate geogrphic entity. Please not that this list refers to CRNS networks COSMOS-UK, COSMOS-Europe...Â
Line 42: Please also list this newer review on reanalysis, were a definition of reanalysis products is given https://doi.org/10.1029/2020RG000715
Line 50: This is well formulated. However, SMAP Level 4 is missing in this list https://doi.org/10.1029/2019MS001729Â
Line 115: rephrase to "harmonized processing of CRNS datasets".
Line 132: Â Where is SMAP Level 4? Please give strong reason for not using the most recent Soil Moisture Product https://doi.org/10.1029/2019MS001729 SMAP Level 4 is a reanalysis product as it includes meteorological variables, and satellite observations that are used to update soil moisture in a data assimilation framework.Â
In fact, I strongly suggest to add SMAP Level 4. As such, the study is out-dated. Adding SMAP Level 4 should be a short exercise - multiplying the scientific impact of this study by a factor. Hopefully SMAP Level 4 performs best amongst all Reanalysis products. Although it is not 20 years length, it is the SM reanalysis product that will be used for recent years rather than any of the other products analyzed. The technical definition of 20+ years (line 42) by a reference of 2005 is not sufficient to not use the most up-to-date SM reanalysis product in this highly relevant global study.
Line 164: This depends on the source of uncertainty. Daily averaging causes loss of signal and proper filtering maintains signal while reducing uncertainty/noise. High measurement uncertainty can be compensated for by applying temporal filtering methods or simple daily averaging e.g. https://doi.org/10.3390/s22239143 .Â
Line 174: Rephrase towards more active voice: Â "Spatial scale matching"
Line 184: see comment before.
Figure 2: Add region names and a/b/c/d to the figure ( or names of the networks ).
Line 290: Clarify meaning of "Figure 4cd". And typically it is called "US sites" rather than "USA sites".
Line 306: Clarify the use of "main paper" which implies, there is a secondary paper.
Line 318: "low temporal correlation" directly contradicts the abstract line 22: "all reanalysis products exhibit good temporal correlation with
the measurements". In the abstract, I suggest to add "products generally exhibit" to weaken this statement in the abstract.Line 325: Please rephrase. Sites cannot be a reason for low performance. Reasons for low performance can only be process related or of technical nature. A site itself cannot be the reason for low performance.
Line 329: CFSv2 or CFSRv2?
Line 330: I contradict, a model cannot perform best in all statistical metrics except for bias. High bias produces high MSE. With a poor bias, the MSE should be poor as well. Please clarify.ÂLine 345: better compared to what? Please specify.
Line 383: Please add that vegetation correction were proposed e.g. https://doi.org/10.1002/2014WR016443 Â
Line 385: Please discuss in a sentence your study results with these results: https://doi.org/10.5194/hess-21-6201-2017
Line 407: lower performance is preferred to "worse" performance which comes with a very negative connotation
Line 423: Pleas be specific for which use these reanalysis products are recommended. The reader leaves the study with the question - for what can these SM reanalysis be recommended?Â
Figure 10: Consider using more positive scoring such as +++ (optimal), ++ (ok) and + (least). No one wants to see "not recommended". Also, ERA5-Land in Humid seems to be rather not recommended according to your results (high bias) but must be considered optimal given there is no better product.
Line 487: See above and add "generally" good agreement.
Line 498: There is no "balanced climate". What climate are you referring to here? I guess: temperate. Please correct.
Citation: https://doi.org/10.5194/hess-2023-224-RC2
Yanchen Zheng et al.
Yanchen Zheng et al.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
207 | 50 | 7 | 264 | 26 | 1 | 2 |
- HTML: 207
- PDF: 50
- XML: 7
- Total: 264
- Supplement: 26
- BibTeX: 1
- EndNote: 2
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1