| 02 Dec 2025
Meteorological Evaluation of the MERRA-2 Reanalysis Dataset: Insights for the Indian Subcontinent
Abstract. MERRA-2 meteorological data is widely utilized across the Indian region to investigate various climatological phenomena, necessitating a thorough evaluation of its accuracy. This study evaluates the performance of MERRA-2 meteorological fields over the Indian region by combining radiosonde measurements with satellite observations from AIRS and TRMM, along with reanalysis data from NCEP/NCAR. Our analysis concentrated on important meteorological variables, such as temperature, precipitation, water vapor, wind components, and tropopause pressure, examining them in multiple seasons and pressure levels. MERRA-2 demonstrates comparable seasonal and spatial variations in temperature relative to AIRS observations, with strong correlations (r2 > 0.85) and root mean square errors (RMSE) ranging from 0.9 K to 2.5 K near the surface, decreasing to approximately 1 K at higher altitudes. However, MERRA-2 exhibits a cold bias closer to the surface and warm biases in the upper troposphere. Water vapor profiles reveal a wet bias, particularly in the lower to mid-troposphere, with RMSE increasing with altitude, from less than 20 % at 1000 hPa to more than 75 % at 300 hPa. Significant discrepancies are found in zonal wind estimates in the lower troposphere, especially over the Tibetan region, where MERRA-2 overestimates wind speeds. Below 700 hPa, Zonal winds show mean biases (MB) from −0.7 to 1.5 m s-1 and RMSEs between 0 m s-1 and 2.2 m s-1. Agreement improves above 700 hPa, with MBs ranging from −0.5 to 0.6 m s-1, and zonal wind estimates outperform meridional winds (RMSE: 0 m s-1 - 4.4 m s-1). MERRA-2 reasonably captures the spatial distribution and intensity of precipitation but overestimates rainfall over complex terrain during the summer monsoon by up to 20 mm d-1 compared to TRMM data. Tropopause pressure comparisons show good agreement with AIRS (MB: −2 to 3 hPa; RMSE: 2 hPa–4 hPa), though larger biases are evident against radiosonde data (MB: 11 hPa–29 hPa). These findings underscore the robustness of MERRA-2 in representing regional meteorological variability over the Indian region, while also highlighting specific biases, particularly in the lower troposphere and over complex terrain, that require careful consideration. As MERRA-2 data are frequently used as input for climate and chemical transport models, identifying and quantifying these biases is essential for improving model accuracy and enhancing the reliability of atmospheric simulations. This study offers critical insights for developing more robust modelling frameworks.
Meteorological Evaluation of the MERRA-2 Reanalysis Dataset: Insights for the Indian Subcontinent
Chakradhar Reddy Malasani, Basudev Swain, Ankit Patel, Arundathi Chandrasekharan, Aishwarya Singh, Nidhi L. Anchan, Rui Song, Amit Sharma, and Sachin S. Gunthe
This paper compares MERRA-2 meteorological variables (T, p, wind, humidity, precipitation) with satellite, radiosonde and further reanalysis data over India to investigate the accuracy of MERRA-2.
The paper is mostly clear and good to follow, with some small mistakes in punctuation and writing. The results section should be restructured and reformulated for a clearer understanding. The topic of the paper is highly relevant. However, the findings should be made more meaningful, e.g. by a more in-depth discussion of possible reasons for differences between datasets.
The weakness of the paper is the methodology. First, MERRA-2 fields are compared to datasets that are assimilated in MERRA-2. The MERRA-2 dataset is compared with AIRS satellite data that have been assimilated to MERRA-2. Furthermore, radiosonde observations are assimilated in MERRA-2, i.e., from the IGRA dataset. The radiosonde data that are used for the comparison in this study come from the same underlying station reports. Thus, the radiosonde comparative data set is also assimilated in MERRA-2 — through the operational data stream. Moreover, comparing a reanalysis with another reanalysis (NCAR/NCEP) might not reveal an answer about the accuracy of MERRA-2 as both reanalyses have a limited resolution and depend on the accuracy of the input data.
Second, only one year is used for comparison, which might be a small data basis and does not reflect climatological extremes. The comparison within this paper is conducted for the year 2010. I would include a larger number of years to be definitely sure about capturing climatological conditions. Moreover, a second comparison with years reflecting non-climatological years to know how MERRA-2 represents extreme situations.
In conclusion, the paper requires major revisions. For the comparison, please use datasets that are not assimilated into MERRA-2 and expand the comparison over several years that also cover extreme meteorological conditions.
General notes:
Minor notes:
L4: by combining with radiosonde measurements, satellite observations from AIRS and TRMM, and reanalysis data from NCEP/NCAR.
L4: concentrates
L12: zonal
L19: As MERRA-2 data are frequently used as input for climate and chemical transport models, identifying and quantifying these biases is essential for improving model accuracy and enhancing the reliability of atmospheric simulations: Use as first sentence of the introduction and drop the sentence here.
L38: Please also name some spaceborne observational systems - like for the campaigns - before talking about their advancements and limitations.
L39: Please explain the satellite-based problems with topography in more detail.
L42: While in-situ observations… This is not only true for in-situ but also for remote sensing observations -> Please reformulate: The beforementioned coverage gaps in observations need to be filled by complementary datasets.
L44: In this context, reanalysis datasets have emerged as essential tools for studying regional climate variability across India: Please give a reference.
L44: Do reanalysis data only serve for regional climate variability studies? This is a very specific example for application. They are also used for closing data gaps in case studies. Please add this and give references.
L46: across both land and oceanic regions -> Please rewrite: over land and ocean.
L46: Among the available datasets, the Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2) (Gelaro et al., 2017), which was developed by NASA, has become one of the most widely used datasets. -> Please add reference.
L48: MERRA-2 assimilates a lot of observational data sources to provide high-resolution meteorological fields (Gelaro et al., 2017). -> Drop this statement here and add it to the MERRA-2 data section in more detail by elaborating the data sources and the MERRA resolution.
L50: Beyond meteorological analysis: climate and chemical transport models are also meteorological analysis methods. -> Just drop the first part of the sentence: MERRA-2 plays also a vital role …
L53: While both regional and global climate models (GCMs) are valuable tools, their reliability hinges on the quality of the input meteorological data. -> Please reformulate: For CTMs and GCMs accurate input data for initialization and boundary conditions are necessary. For CTMs this input should additionally be highly resolved.
L54: Fine-scale regional processes, such as land-sea interactions and topographic complexity, are often poorly resolved in global models, which underscores the importance of providing accurate meteorological inputs -> I would say that the models have a poor spatial resolution. This has nothing to do with the accuracy of the input data. Just drop the sentence.
L 56: MERRA-2 is frequently used for this purpose due to its global coverage and temporal continuity. -> But this is also delivered by ERA5 – with even a better spatial resolution! Please give better reasons for taking MERRA-2 and point out the advantages of MERRA-2 more clearly (e.g., for aerosol analysis). And mention the temporal continuity and coverage already in line 49.
L 60: Moreover, MERRA-2 data is being used in providing meteorological input or boundary conditions for both regional and global CTMs focused on simulating atmospheric composition over India and other parts of the world (Anchan et al., 2024; Malasani et al., 2024; Swain et al., 2024). -> This was already mentioned before. Please drop this sentence.
L 62: Given this widespread application, it is imperative to evaluate the accuracy, consistency and reliability of MERRA-2 over the Indian subcontinent to ensure robust climate and air quality simulations. -> Please drop this sentence.
L66: The study seeks to analyze the performance and applicability of MERRA-2 in representing spatio-temporal meteorological conditions over the Indian subcontinent by identifying and quantifying its biases. -> Please clarify how. Biases compared to which data sets?
L 77: 0.01 hPa. The
L 78: for surface-level mixing depth variables -> What are mixing depth variables?
Fig. 1: The study area and its topographical features are illustrated, along with the geographic positions of the radiosonde (RAOB) stations utilised for this analysis. Station codes corresponding to each site are detailed in Table 6. -> Please shorten the caption: Topography of the study area and locations of the analyzed radiosonde stations (RAOB) (red circles). Station codes are detailed in Table 6.
L 83: leading to better accuracy and representation of key meteorological parameters (McCarty et al., 2016) -> On which scale? Globally?
L 84: MERRA-2 products have been extensively used in a wide range of atmospheric chemistry and air quality studies across South Asia, demonstrating their ability to reasonably capture spatiotemporal patterns of chemical evolution (Kara and Elbir; Wen et al., 2022; Hamal et al., 2020). The reliability of these applications is fundamentally linked to the accuracy of the meteorological fields. The present work evaluates how well MERRA-2 meteorological data represent conditions over the Indian subcontinent (as shown in Fig. 1 ) by comparing it against multiple independent datasets, including radiosonde observations (RAOB), satellite retrievals from AIRS and TRMM, and reanalysis data from NCEP/NCAR. The analysis involves a detailed quantification of biases across these observational platforms to assess the consistency and reliability of MERRA-2 in capturing regional meteorological conditions. -> Drop these sentences here. Instead, add aspects that are not mentioned in the Introduction yet to the Introduction section.
L 95: https://www.ncei.noaa.gov/data/integrated-global-radiosonde-archive/ -> Please create an extra reference, add it to the reference list and cite properly. Add this sentence also to the data availability section.
L 96: were utilised -> are utilized (Fig. 1). Please add a reference here and write your analysis in present. Apply this to the entire paper.
L 97: high-altitude (500–1000 m) -> Please add the reference (Table 6).
Table 6: Please correct the layout! Why do you start with Table number 6? Change the numbering to 1.
L 97: Further details provided in Fig. 1 and Table 6. -> Drop the sentence.
L 99: remove round brackets and correct ; to ,
L 99: Are the observations at these pressure levels the specific values obtained at each height or are these observations means around each level?
L 102: for satellite retrieval validation -> Drop it since it is not the scope of your paper.
L 105: SPACEBORNE remote sensing FROM GESOSTATIONARY SATELITES provides an approach for continuous monitoring of atmospheric conditions COVERING LARGER AREAS THAN GROUND-BASED STATIONS.
L 107: AIRS is a hyperspectral infrared sounder aboard NASA’s Aqua satellite, consisting of 2378 infrared channels and 4 visible/near-infrared channels. With a swath width of 1652 km, AIRS achieves a horizontal resolution of 13.5 km and vertical resolution near 1 km. -> Please add references.
L 105: In this study, satellite-based observations from two major platforms- Tropical Rainfall Measuring Mission (TRMM) and Atmospheric Infrared Sounder (AIRS)- ARE USED.
L 112: More information is available at https://airs.jpl.nasa.gov/mission/overview/ and https://aqua.nasa.gov/content/airs. -> Please create an extra reference each, add them to the reference list and cite properly.
L 115: , which was launched in November 1997 to monitor rainfall and assess the associated latent heating in the Tropics and Subtropics.
L 118: The 3B42 product integrates multi-satellite observations and provides rainfall estimates, particularly over the Indian region. If you do not want to drop reliable you have to at least give a reference.
L 119: Data can be accessed at https://pmm.nasa.gov/data-access/downloads/trmm. -> Please create an extra reference, add it to the reference list and cite properly. Add this sentence also to the data availability section.
L 121: Reanalysis Dataset -> This section name is confusing since MERRA-2 is also a reanalysis dataset.
L 122: The MERRA-2
L 122: including the zonal (U) and meridional (V) components, -> Drop this part of the sentence.
L 123: NCEP/NCAR reanalysis. -> This is the first time you mention NCEP/NCAR. Thus, please write the full name.
L 125: It provides a consistent, high-resolution depiction of global atmospheric fields. -> Drop the sentence since more details follow.
L 126: The reanalysis product of NCEP/NCAR
L 126: over global grids (144 longitudinal × 73 latitudinal points), spanning 0°E–357.5°E -> These longitudes do not cover the whole globe.
L 128: via https://psl.noaa.gov -> Drop this here and move it to the data availability section.
L 130: I would move the methodology subsection to an extra section to clearer distinguish data and methodology.
L 131: against various ground-based and satellite-borne observations -> Please mention radiosondes as well.
L 138: including temperature, water vapor, wind components, and precipitation. -> Also mention pressure.
L 139: each station -> each measurement
L 141: station location -> measurement location
L 141: For the temporal comparisons, MERRA-2 outputs were averaged over time windows corresponding to the observational time stamps (0000 UTC and 1200 UTC). -> How long are these time windows? For the spatial comparison, the nearest location of MERRA-2 data is used. Wouldn’t it be more consistent to also take the ‘closest’ time stamp?
L 147: For water vapor retrievals, additional filtering was applied based on the methodology described by Olsen et al. (2005), discarding profiles with negative values or errors exceeding 50%. -> Why do negative values exist at all?
L 149: averaged values from AIRS.
L 150: , thus
L156: he -> The
L156: Drop (MB) since you already defined the abbreviation in the sentence before.
L 164: of observations -> of observed
L 174: And Unsystematic -> The unsystematic
L 176: found using Equation 5. -> calculated using Equ. 5.
Sect. Evaluation Methodology: You should mention that the evaluation is on a seasonal basis. Please also define the seasons (e.g., J/F/M -> winter) within the methodology section.
Sect. Results: Proposal for results subsections:
Fig. 2: Drop ‘for all the four seasons’. Please specify the rows. E.g., surface pressure (hPa) (first row). This is relevant for all figures!
L 178: The spatial distribution of seasonally averaged surface pressure (hPa), specific humidity (g kg−1), surface temperature (K) and precipitation (mm) derived from MERRA-2 for the each season of 2010—winter (DJF), spring (MAM), summer (JJA), and autumn (SON) is shown in Fig. 2. -> Please drop this sentence that just describes the figure. Instead, explain shortly that you analyze the MERRA-2 seasonal cycle and spatial variability first.
Sec. 3: Please improve the structure of the text. E.g., start with the analysis of the seasonal cycle for all variables. Follow with the spatial analysis afterwards using the same order of variables as before. (You started the analysis of the seasonal cycle with pressure, thus start the spatial analysis also with pressure).
L 181: However, compared to summer, regions north of 20°N experience somewhat larger variations, around 5 hPa. -> What does this mean? Please rewrite.
Fig. 2: It is very hard to capture differences between the pressure plots. Please create a separate plot for pressure that does not include the other variables and shows pressure differences compared to one season instead of absolute values. An anomaly plot enhances the readability of the plot.
L 183: with extreme values typically occurring in summer or winter -> with maxima in summer and minima in winter.
L 186: Regional differences in temperature magnitude across seasons reflect variations in solar heating over diverse landscapes and the influence of regional meteorological factors (Kumar et al., 2012). -> This is very general. Please explain the regional factors and the variations in solar heating more specifically.
L 189: (Weldeab et al., 2022).
L 189: Kerala region (°N, °E) -> Please specify the coordinates of the subregion. Apply this to all subregions mentioned in the paper.
L191: Surface pressure remains relatively constant in the Bay of Bengal and Arabian Sea, showing little seasonal variation. -> Please mention this already during the seasonal analysis.
L 198: The average wind speeds and vectors from MERRA-2 for all four seasons are shown in Fig. 3. -> Drop this sentence and start with the results immediately.
Fig. 3: Add Fig. 3 as an additional row to Fig. 2.
L 201: , contributing to substantial sea-salt aerosol production in the Arabian Sea -> Please drop this since you do not analyze aerosols. This is not the scope of the paper.
L 205: Wind vectors over the Himalayan region and Tibetan Plateau during winter are typically southwesterly (Zhu et al., 2024) -> Please shortly discuss why.
L 207: respectively.
L 215: comparing it against -> MERRA-2 against.
L 215: reanalysis datasets -> additional reanalysis datasets.
L215: This comprehensive assessment will help determine the reliability of MERRA-2 for use in climate modeling as initial and boundary condition data. -> Drop this statement. You already mentioned this a few times.
Fig. 4: Please increase the space between row 2 and 3 for better readability and swap the season labels with the temperature title. Please also capitalize the colorbar labels for a uniform design of your plots.
Sect. 3.1: Please clarify in the text why missing data exist in Fig. 4, why you analyze the data at 700 hPa, and how the comparison looks like at other levels.
L 218: temperature (K) and water vapor (g kg−1) -> Drop units.
L 219: Section 2.5 -> Sect. 2.5.
L 220: in increase in the magnitudes of temperatures -> an increase in the magnitude of temperature.
L 221: followed by a decrease -> A decrease from when to when?
L 221: A strong north-south gradient of temperatures is there in MERRA-2 than the AIRS temperatures over the Indian subcontinent. -> Please reformulate.
L 224: relationship -> correlation.
L 224: There is a strong correlation between AIRS and MERRA-2 temperatures in all seasons. -> Why do you now? Please verify your statement (e.g., because parameter XY is larger/smaller).
Fig. 5: Please change the following aspects: grey shading -> dashed lines; The middle panel -> second row; Kelvin is K not k; bottom panel -> third row; . at the end; refer to Sect. Methodology after naming the statistical metrics.
Fig. 5: Is a linear fit applicable for the summertime scatterplot? I would say the linear regression does not fit the data at all! Please reconsider the fit.
Sect. 3.1: Please shortly assume a reason for the vertical distribution of statistical parameters regarding the temperature.
L 227: The r2 > 0.85 for all seasons except summer, particularly below 850 to 925 hPa. -> Please reformulate.
Fig. 6: specific humidity (g/kg) -> Comparison of specific humidity …; Apply comments given for Fig. 5.
L 237: 300 hPa (Divakarla et al., 2006a).
L 237: There is good correlation in all seasons except summer, possibly due to large spatial variability in water vapor caused by the southwest monsoon. -> You say that MERRA-2 cannot capture spatial variability correctly. But Fig. 5 shows that MERRA-2 just overestimates humidity at the eastern coast. I would say it is less a problem of the spatial variation but more an overestimation at one specific location.
Table 2: Please just rewrite the caption from Table 1 and drop the second part of the sentence. Moreover, refer to the section where you mention why you analyze the data only up to 300 hPa.
L 240 : of MB, r2,
L 245: 20% ,
L 245: The MERRA-2 wet bias may lead to overestimation of hydroxyl radical concentrations, which could cause underestimation of various volatile organic compounds, thus affecting ozone concentrations. -> Drop this as it has nothing to do with your analysis.
L 247: these indicators -> which indicators?
L 247: When MERRA-2 serves as boundary conditions, errors in simulated water vapor are unlikely to significantly affect air quality modeling, provided other sources of error are absent. -> Please shift this statement from the results to the discussion.
Fig. 7: Please drop ‘White areas denote missing data’ since no white is shown.
Sect. 3.2 / Fig. 7: Swap either the text about the meridional wind with the one about the zonal wind or swap the respective rows in Fig. 7 to match the order of appearance of both variables.
L 252: The spatial patterns of meridional wind components from both NCEP and MERRA2 exhibit a high degree of similarity -> But not in the northern parts of India! Please correct the text.
Sect. 3.2: Please restructure the section: start with the meridional wind analysis and end with the zonal wind analysis.
L 256: Correlation between datasets improves with altitude. -> This is a result shown in Fig. 8 and 9. However, the result is mentioned already before the analysis of the both figures. Moreover, the statement is not confirmed by Fig. 9 for the meridional winds.
L 263: The r2 and d
L 266: Summer exhibits the most improved values -> Please rewrite.
L 267: show similar behavior between datasets. -> So, you mean that discrepancies are small? Please rewrite.
L 270: 0.6, and RMSE
L 269: The wind speed evaluation criteria of Emery et al. (2001) were adopted -> What do you mean? Please clarify.
Table 3: Please refer to Fig. 1 at the end of the first sentence and add ‘.’ at the end of the second sentence.
Sect. 3.3: Do you use the total MERRA-2 precipitation, including snow and liquid? Please clarify in the MERRA-2 data section. Do satellites have different shortcomings in retrieving solid or liquid precipitation? Please mention this in the TRMM data section. Moreover, include these aspects in the discussion of your results.
L 281: The Himalayan regions and parts of eastern India receive higher precipitation in both summer and winter -> Fig. 10 shows higher precipitation in summer only, but not in winter, and over western India instead of eastern India.
L 282: Please drop ‘western disturbances’ and just write Westerlies.
L 282: This winter precipitation plays a crucial role in supporting rabi crops and sustaining glacier mass, which subsequently contributes to river flow during other seasons (YADAV et al., 2012). -> However, the figure shows no enhanced winter precipitation. Even if this would be the case, this statement should be part of the discussion section.
Sect. 3.3: Please focus more on the differences in summer, i.e., the high MEARRA-2 values over the Himalayan regions. Additionally, explain why the MERRA-2 precipitation amount is enhanced at the western coast.
L 287: levels that TRMM generally underestimates. -> Prove that TRMM underestimates these levels, e.g., through a reference. Couldn’t it also be that MERRA-2 overestimate precipitation at these levels?
L 289: Why is TRMM less accurate over complex terrain? Please explain and add a reference to the respective dataset section.
L 291: Previous studies have reported that both global and regional climate models encounter challenges in accurately simulating the South-Asian monsoon (Rajan and Desamsetti, 2021). These limitations are linked to the complex monsoon dynamics, diverse regional topography, and localized convection processes. Such conditions also contribute to reduced retrieval accuracy in TRMM during the summer monsoon, a period marked by strong spatial gradients in temperature and precipitation, prevalent warm cloud systems, and heterogeneous terrain (Indu and Nagesh Kumar, 2014; Shukla et al., 2019). -> This should be part of the discussion and not of the results. Moreover, this statement raises the question, whether TRMM can be used in summer at all. Please discuss whether the summer differences stem from TRMM inaccuracies or inaccuracies inMERRA-2.
L 293: localized convection processes -> Localized convection processes should be investigated by separately analyzing the convective precipitation component in MERRA-2. This would allow an assessment of whether MERRA-2 struggles to adequately represent convective precipitation. In contrast, deficiencies in the large-scale precipitation component would indicate that the representation of large-scale monsoon dynamics in MERRA-2 requires improvement. Such a separated analysis of precipitation processes would help identify which parameterizations in MERRA-2 could be improved and would be a great improvement of the paper.
Sect. 3.4: Why don’t you also compare the MERRA-2 humidity, pressure and wind data with respective radiosonde observations? Please discuss this in the radiosonde data section.
Sect. 3.4: Have you performed the spatial attribution of the radiosonde and MERRA-2 data separately at each height level, or were all radiosonde and MERRA-2 data attributed solely based on the radiosonde launch location? A height-dependent spatial attribution is necessary to account for horizontal advection of the radiosonde during ascent.
Fig. 11: Please explain the abbreviation RAOB in the caption; drop ‘are compared’; add a comma between 700 500 and 300 hPa levels.
L 298: Why have you selected these radiosonde stations? Explain why you don’t analyze the average over each site category.
L 299: Seasonal temperature variation is more pronounced at Delhi and Bhopal compared to Bhubaneshwar and Port Blair. Notable differences in surface temperature are observed between January and May at most sites, except at Port Blair. -> Explain shortly why.
L 301: Overall, MERRA-2 temperatures show good agreement with radiosonde observations across all pressure levels and seasons. -> I don’t support this statement for near-surface levels. Please correct the statement.
Fig. 12: Please drop ‘are compared’; refer to Table 6 after category sites (Table 6); explain the statistical metrics and refer to Sect. 2.6; decrease the x range of the RMSE plots.
L 303: correlation coefficient (r2)
L 309: Furthermore, discrepancies between the actual elevation of the station and the topography assigned to the 310 model (Table 6) can contribute to temperature biases in the reanalysis data. -> However, you compare temperature at pressure levels not at the surface, thus, wrong surface elevation should not matter.
L 314: Reichler et al. (2003)
Fig. 13: Please rewrite: from co-located AIRS and MERRA-2 for 2010; second row of the plot: hpa -> hPa; Please show difference plots between AIRS and MERRA-2 to improve the clarity of the presentation.
L 218: (Meng et al., 2021),
L 324: winter. The
L 327: atmospheric dynamics -> Please discuss this in more depth.
Sect. 3.5: Please discuss the reasons for the differences between AIRS and MERRA-2 in greater detail, with particular attention to how these differences vary across seasons.
Table 5: Please rewrite: Annual averages and standard deviations of tropopause pressure from radiosonde observations (RAOB), AIRS satellite observations, and MERRA-2 reanalyses data for different sites (Table 6);
Table 5: , with values rounded to whole numbers -> That is not true. Drop the sentence and give all numbers with the same number of decimal places.
L 338: region. The MERRA-2
L 340: during the year 2010, which is a climatologically representative year according to…
L 343: a pronounced-> a more pronounced.
L 346: likely linked to local atmospheric dynamics -> Are you sure about this? Or might the bad representation of boundary layer processes in MERRA-2 be the reason?
L 349: due to AIRS’ reduced sensitivity and simulation uncertainties -> Can you estimate which one is more pronounced?
L 352: Correlations between the datasets improve with altitude, with better agreement observed above 600 hPa -> First, please add more information about the datasets assimilated into NCAR in the respective data section. Second, if the datasets assimilated into NCAR and MERRA-2 and if the datasets used for boundary conditions are the same, similarities between both modelled wind fields at their boundaries (at high levels) will not surprise. If so, a good agreement says nothing about the accuracy of MERRA-2 and NCAR since the input data of both reanalyses might be inaccurate.
L 355: by Emery et al., 2001
L 359: likely reflecting limitations in capturing fine-scale orographic effects. -> Please specify orographic effects on convection. To gain confidence that convection driven by orography is the reason behind the differences, it would be great to analyze the large-scale advective and the convective part of MERRA-2 precipitation separately.
L 262: Radiosonde temperature observations from 35 stations provide a valuable independent benchmark -> Are they really independent or are they used for assimilation in MERRA-2?
L 371: with some systematic biases primarily occurring at lower atmospheric levels and over complex terrain -> I think this finding is relatively weak and predictable. I would expect a more meaningful conclusion.
L 372: AIRS retrievals serve as a valuable observational complement, particularly for characterizing the upper troposphere and tropopause. -> AIRS data are no complement because they are assimilated into MERRA-2!
L 378: Additionally, incorporating higher-resolution observational datasets could help resolve fine-scale variability, particularly during the summer monsoon season when atmospheric conditions are highly dynamic and spatially heterogeneous. -> I doubt this. Already the coarse resolution of MERAA-2 does not allow for resolving fine-scale convection and circulation. High resolution input data would not resolve this problem.
Table 6: Radiosonde -> radiosonde and add ‘.’ to finish the caption.