We thank anonymous Referee #1 for the constructive feedback and the concerns raised. In accordance with the form of the review, we formulate our responses to the concerns below. All of these responses will be incorporated into the revised text. Comments 12) - 17) will be addressed in the revised text without further remark.

1.) We agree with Reviewer#1 that highly temporal resolved observations from lysimeter are needed to provide reliable reference precipitation data. We are using highly temporal resolved lysimeter records to obtain hourly precipitation. Hourly precipitation data used here was compiled as followed: 1.) six weight observations per minute were averaged and stored as 1-minute lysimeter weight data, 2) application of a state-of-the-art noise reduction filter (AWAT), 3) we used the water balance equation to determine the precipitation or ET per 1-minute interval. Within this step we assume for each 1-minute time step, that only ET or precipitation can occur, which are related to either a decrease (ET) or increase of lysimeter weight (precipitation or NRE). 4.) In a next step we aggregate 1-minute precipitation data to hourly values. A more thoroughly description of the data processing procedure can be found in Schneider et al. 2021. We will provide a more precise description on the data processing procedure in the revised manuscript.

2.) Yes, non-rainfall events (NRE) are included in P in Eq. 2. We did not mention NRE in 2.2.2 because the gauges do not register them, except for disdrometers that can detect fog. As for 2.3, data from lysimeters with NRE are treated like any other hour with precipitation, so we did not mention them. In 2.4, P_ref is also calculated for NRE. We will introduce the term atmospheric water input (AWI) in the manuscript and in Eq. 2 to improve the distinction between NRE and regular precipitation.

3.) The methods for distinguishing between NRE and precipitation from lysimeter data without further sensor data are subject to uncertainties. We distinguished between NRE and precipitation in this investigation based on hourly data and mass increases of the lysimeter not concurrent with rain or snow were classified as dew according to Meissner et al. 2007 and Xiao et al., 2009. In previous studies, which focused on determination of non-rainfall water at the same research sites, similar rules have been applied to detect NRE (e.g., Groh et al. 2018, Groh et al. 2019, Forstner et al. 2021). For future studies, leaf wetness sensors and visibility device will be installed at the studied sites to better distinguish between NRE and precipitation and thus to reduce the possible uncertainties. Previous studies for Rollesbroich found that an NRE contributed between 0.013 and 0.017 mm h^-1 (Groh et al. 2019) compared to the minimum precipitation rate of 0.1 mm h^-1 considered in our study. Thus, NRE could possibility co-occur with rain events within 1 hour time intervals, but their overall magnitude at hourly base would be relatively small. Thus, we don’t expected that a co-occurrence of NRE and precipitation would distort the results.

4.) The size of the hexagonal area is approximately 49 m² in total, while the lysimeters cover an area of 6 m² and are each installed at a distance of 1.6 m from the nearest lysimeter. The gauges are installed within a range of about 7 m from the hexagon at all sites, except for the TB2 tipping bucket gauge at Rollesbroich, which is installed about 30 m from the nearest lysimeter. However, this gauge shows one of the best correlations with the lysimeter reference data compared to the other TB gauges. The mean of all observations per lysimeter for Ro, with a precipitation rate above 0.1mm h^-1, showed a maximum deviation of -0.015 mm h^-1 from P_ref (0.941 mm h^-1), and the standard deviation for the mean from all lysimeters was 0.011 mm h^-1. For Se, the max deviation from P_ref (0.819 mm h^-1) in the mean was 0.003 mm h^-1 and the standard deviation was 0.002 mm h^-1. In Dd, the max deviation of the mean average precipitation rate of all lysimeters from P_ref (0.822 mm h^-1) above 0.1 mm h^-1 was -0.007 mm h^-1, with a standard deviation of 0.007 mm h^-1. The differences between the statistics for the different sites could be explained by the distribution of the precipitation rates, since in Ro the precipitation rates have been generally higher (Table 1 and 2). Overall, it can be assumed that measurement errors due to the spatial scale can be neglected, since all devices are installed at a short distance, no shielding can become effective, the temporal resolution is 1 hour and a normal distribution of deviations from the reference due to small-scale spatial influences can be assumed.

5.) The uncertainty estimation based on the standard deviation was only done for the lysimeter data. For the uncertainty ranges within the correlation plots, the uncertainties of the lysimeter measurements have been extended with an additional value of 5 % of P_ref to account for uncertainties of the gauge measurements. Therefore, the uncertainties calculated and discussed are only valid for the lysimeters, while the uncertainty ranges visually supporting the scatter plots are not computed for the respective gauges individually.

6.) Statistics (min, max, median, mean, quantiles) on the precipitation rates are given in Table 1 and Table 2. In general, the distribution of the precipitation rates is highly skewed. For Dedelow (Dd), Rollesbroich (Ro) and Selhausen (Se) the precipitation rates within the top 5 % quantile are responsible for 42.6, 38.5 and 35.7 % of the overall observed precipitation at the sites. However, only 40, 63 and 44 observations exceed 5 mm h^-1 at Dd, Ro and Se, respectively. For precipitation rates above 10 mm h^-1, only 4, 10 and 4 hours have been recorded for Dd, Ro and Se.

By mistake, 20 mm h^-1 has been described as the maximum observed precipitation intensity during the observation period across all sites. In Ro, the value was once exceeded with 23.21 mm h^-1 and in Dd also a single observation exceeded the value with 32.24 mm h^-1. However, this has very little effect on the uncertainty range, as for Ro the uncertainty for this datapoint is 1.62 % and for Dd 1.19 %. The error solely affected the lysimeter uncertainty ranges, since for all other analysis the correct values have been used.

Table 1: Distribution of precipitation rates determined by the lysimeters over the observation period for P_ref >= 0.1 mm h^-1, as used for the gauge comparison. 

Table 2: Distribution of precipitation rates determined by the lysimeters over the observation period for P_ref > 0 mm h^-1, including water from non-rainfall events.

7.) We agree with Reviewer#1 that the correction methods are based on empirical data. The method after Richter has been developed especially for different locations with scarce data coverage. Therefore, the parametrisation is done loosely based on specific location characteristics. Here, the measurement method is important since the original method has been exclusively developed for Hellmann-type gauges. For the Dynamic Correction Method, factors for precipitation type, wind speed and gauge model are defined. The only site-specific factor appears to be the wind speed, which varies according to the site, but since the method has been developed covering a range of wind speeds including those measured at the sites, the factors should cover the issue. Overall, the correction results have been similar for the devices of the same gauge type at the different sites.

8.) The corresponding passage will be adapted in the revised manuscript: "The low bias in hourly lysimeter measurements indicated the suitability of their arithmetic mean as a reference for comparing precipitation methods."

9.) Reference should be to Fig. 2 A, not 2 B. Here the hexagon area is marked across the six lysimeters in equidistance to the control shaft in the middle. Fig. 2 B shows two of the hexagonal lysimeter arrangements in a row. We will add a scale and more information to the plot.

10.) The sum is calculated after the application of Eq. 1, 2 and 3.

11.) n_ia are the lysimeters which have been inactive or did not provide reliable data during time interval i. We will adjust the naming of the variables to make them clearer.

Forstner, V., Groh, J., Vremec, M., Herndl, M., Vereecken, H., Gerke, H. H., Birk, S., and Pütz, T.: Response of water fluxes and biomass production to climate change in permanent grassland soil ecosystems, Hydrol. Earth Syst. Sci., 25, 6087– 6106, https://doi.org/10.5194/hess-25-6087-2021, 2021.

Groh, J., Slawitsch, V., Herndl, M., Graf, A., Vereecken, H., and Pütz, T.: Determining dew and hoar frost formation for a low mountain range and alpine grassland site by weighable lysimeter, Journal of Hydrology, 563, 372–381,

https://doi.org/10.1016/j.jhydrol.2018.06.009, 2018.

Groh, J., Pütz, T., Gerke, H. H., Vanderborght, J., and Vereecken, H.: Quantification and Prediction of Nighttime Evapotranspiration for Two Distinct Grassland Ecosystems, Water Resour. Res., 55, 2961–2975,

https://doi.org/10.1029/2018WR024072, 2019.

Meissner, R., Seeger, J., Rupp, H., Seyfarth, M., and Borg, H.: Measurement of dew, fog, and rime with a high-precision gravitation lysimeter, Z. Pflanzenernähr. Bodenk., 170, 335–344, https://doi.org/10.1002/jpln.200625002, 2007.

Schneider, J., Groh, J., Pütz, T., Helmig, R., Rothfuss, Y., Vereecken, H., and Vanderborght, J.: Prediction of soil evaporation measured with weighable lysimeters using the FAO Penman–Monteith method in combination with

Richards’ equation, Vadose Zone Journal, https://doi.org/10.1002/vzj2.20102, 2021.

Xiao, H., Meissner, R., Seeger, J., Rupp, H., and Borg, H.: Effect of vegetation type and growth stage on dewfall, determined with high precision weighing lysimeters at a site in northern Germany, Journal of Hydrology, 377, 43–49, https://doi.org/10.1016/j.jhydrol.2009.08.006, 2009.

We thank anonymous Referee #2 for the positive feedback and the constructive suggestions made. Below we formulate our responses to these general and detailed suggestions. Referenced figures are provided with the attached supplement.

We agree that both, the gauge data (y-axis) and lysimeter data (x-axis) are subject to uncertainties. We tested the application of the reduced major axis regression and found that it has only minor impacts on most of the correlation plots (e.g., Fig. 1), except for those involving data from acoustic sensors (Fig. 2). In our study, the regression line plays a subordinate role, and the statistics used would not be affected by a different regression method.  Based on our assumption that lysimeter data is suitable as reference data, we would also not assign equal attribution to deviations from the 1:1 agreement for gauge and lysimeter data. Therefore, we will continue to use the current regression method of least squares for the analysis.

To identify trends in the relationship between wind speed and catch rates, we consider data binning to be a promising method. We tested this approach by calculating the mean and median catch ratios per bin, with each bin containing data within a range of 0.1 m s^-1 for wind speeds ranging from 0.0 to 12.0 m s^-1. We applied this method to the entire dataset for an acoustic sensor with both 10-minute (Fig. 3 A) and hourly (Fig. 3 B) resolution. The results showed that the trends for the same gauge were similar between both temporal resolutions (Fig. 3). Overall, this approach provides clearer trends of the catch ratios as a function of wind speed (e.g., Fig. 3 and 4), compared to unbinned catching ratios and wind speed. Therefore, we will include in a revised version of the manuscript data binning for analysing the influence of wind speed on the catching ratios.

Data binning is also a useful method for structuring comprehensive rainfall intensity data and reducing the impact of a large number of low-intensity events on statistics. However, we believe that visualising all available data points provides a more concise understanding of the gauge's measurements for rainfall intensities. We would exclude gauge-dependent deviations from the reference line, such as the broad scattering around the reference line for acoustic sensor data. Additionally, to effectively bin the data, well-considered ranges of rainfall intensities for the respective categories would be required in order not to manipulate the statistics. Therefore, we will not consider data binning for the correlation plots of this publication.

The issue about potential uncertainties derived from ET on the data was discussed in the answer to referee 1 and will be further concluded in the revised document. Generally, the hourly lysimeter data were derived from minutely measurements which were processed with a state-of-the-art noise reducing filter and checked for weight increase (P or non-rainfall water) or weight decrease (ET). Afterwards the minutely data was separately aggregated to hourly values.

It is challenging to estimate the total amount of NRE without gap-filled data. However, studies have been conducted for the same research sites focusing on the quantification of (annual) dew formation (Forstner et al. 2021, Groh et al. 2019, Groh et al. 2018). We will include these data in our revised manuscript and consider it in the discussion.

The feedback for minor changes will be addressed directly in the revised manuscript.

Forstner, V., Groh, J., Vremec, M., Herndl, M., Vereecken, H., Gerke, H. H., Birk, S., and Pütz, T.: Response of water fluxes and biomass production to climate change in permanent grassland soil ecosystems, Hydrol. Earth Syst. Sci., 25, 6087–6106, https://doi.org/10.5194/hess-25-6087-2021, 2021.

Groh, J., Pütz, T., Gerke, H. H., Vanderborght, J., and Vereecken, H.: Quantification and Prediction of Nighttime Evapotranspiration for Two Distinct Grassland Ecosystems, Water Resour. Res., 55, 2961–2975, https://doi.org/10.1029/2018WR024072, 2019.

Groh, J., Slawitsch, V., Herndl, M., Graf, A., Vereecken, H., and Pütz, T.: Determining dew and hoar frost formation for a low mountain range and alpine grassland site by weighable lysimeter, Journal of Hydrology, 563, 372–381, https://doi.org/10.1016/j.jhydrol.2018.06.009, 2018.