Evaluation of precipitation measurement methods using data from precision lysimeter network
Abstract. Accurate precipitation data are essential for assessing the water balance of ecosystems. Methods for point precipitation determination are influenced by wind, precipitation type and intensity and/or technical issues. High-precision weighable lysimeters provide precipitation measurements at ground level that are less affected by wind disturbances and are assumed to be relatively close to “true” precipitation. The problem in previous studies was that the biases on precipitation data introduced by different precipitation measurement methods were not comprehensively compared and quantified with those obtained by lysimeters under different climatic conditions.
The aim was to quantify measurement errors of standard precipitation gauges as compared to the lysimeter reference and to analyse the effect of precipitation correction algorithms on the gauge data quality. Both correction methods rely on empirical constants to account for known external influences on the measurements, following a generic and a site-specific approach. Reference precipitation data were obtained from high-precision weighable lysimeters of the TERrestial Environmental Observatories (TERENO)-SOILCan lysimeter network. Gauge types included tipping bucket gauges (TBs), weighable gauges (WGs), acoustical sensors (ASs), and optical laser disdrometers (LDs). The data were collected from 2015–2018 at three sites in Germany and compared with a temporal resolution of 1 hour for precipitation above a threshold of 0.1 mm h-1.
The results show that all investigated measurement methods underestimated the precipitation amounts relative to the lysimeter references for long-term precipitation totals with catch ratios (CRs) between 33–92 %. Data from ASs had overall biases of 0.25 to -0.07 mm h-1, while data from WGs and LDs showed the lowest measuring biases (-0.14 to -0.06 mm h-1 and 0.01 to 0.02 mm h-1). Two TBs showed systematic deviations with biases of -0.69 to -0.61 mm h-1, while other TBs were in the previously reported range with biases of -0.2 mm h-1. The site-specific and generic correction schemes reduced the hourly measuring bias by 0.13 and 0.08 mm h-1 for the TBs and by 0.09 and 0.07 mm h-1 for the WGs and increased long-term CRs by 14 and 9 % and by 10 and 11 %, respectively.
It could be shown that the lysimeter reference operated with minor uncertainties in long-term measurements under different climatic conditions. The results indicate that even with well-maintained and professionally operated stations, considerable precipitation measurement errors can occur, which generally lead to a loss of recorded precipitation amounts. Data from standard precipitation gauges therefore still represent potentially significant uncertainty factors. The results suggest that the application of relatively simple correction schemes, manual or automated data quality checks, instrument calibrations and/or adequate choice of observation periods can help improve the data quality of gauge-based measurements for water balance calculations, ecosystem modelling, water management, assessment of agricultural irrigation needs or radar-based precipitation analyses.
Tobias Schnepper et al.
Status: final response (author comments only)
RC1: 'Comment on hess-2022-370', Anonymous Referee #1, 25 Dec 2022
- AC1: 'Reply on RC1', Tobias Schnepper, 05 Feb 2023
RC2: 'Comment on hess-2022-370', Anonymous Referee #2, 07 Feb 2023
- AC2: 'Reply on RC2', Tobias Schnepper, 06 Mar 2023
Tobias Schnepper et al.
Tobias Schnepper et al.
Viewed (geographical distribution)
The paper by Tobias Schnepper et al. focuses on evaluating several precipitation measurement techniques based on data from the TERrestial Environmental Observatories (TERENO)-SOILCan lysimeter network as a reference.
The topic is important, and the paper is well-structured and written. I have, however, several concerns that require clarification and/or revisions, as described below:
1) Reliability of precipitation measurements from the lysimeters: The assumption of no ET during precipitation interval and vice versa seems valid only at very short time intervals. Within an hour interval, I assume precipitation and ET can co-occur, especially for very wet soil and high evaporative demands. I think the authors should support this assumption, which underlies the precipitation computation. Otherwise, the reliability of the reference data is questionable.
2) Eq. 2: I assume P here includes NRE? It is unclear why NRE is presented in Eq. 1 but ignored until section 2.5.1. How come? Please clarify this.
3) More on NRE: according to former studies, NRE is a primary source of error for standard gauges compared to lysimeter precipitation data. It is, therefore, crucial this component would be as accurate as possible. However, the rules to determine NRE cases are very “ad-hoc” and probably specific to a given site. Furthermore, is it reasonable to assume that within an hour interval, precipitation will be either NRE or not? Isn’t it possible to have rainfall and NRE together within the hour?
4) Spatial autocorrelation of rainfall: using the mean of all hexagon gauges as the precipitation reference for the other type of gauges ignores spatial variance of precipitation within the hexagon area. There is no information about the size of the hexagon; maybe it is too small, and the variance is neglectable. Still, in principle, the difference between the lysimeter spatial average and the gauge measurement may be related to their spatial scale.
5) Uncertainty estimation: is computed based on the standard deviation between measurements for each hour. This computation is based on the assumptions: 1) the precipitation measurements (for a given hour) are normally distributed with the same mean and variance for all gauges (the text says the first part of this statement), and 2) the data from the different gauges are independent. How can these two assumptions be justified?
6) The highest observed precipitation rate is 20 mm/h. Can you provide some info about this value so the reader would know what part of the precipitation rate distribution the analysis is covering?
7) Section 2.8: Precipitation data corrections: This section proposes correction procedures that seem very empirical. How much can we trust these methods in the general case? How applicable are they for locations different than their development?
8) The following sentence appears in the conclusion section (L677): “The arithmetic mean of the lysimeter measurements has proven to be an almost unbiased reference for the precipitation measurement method”. I don’t think this was proved, but rather the assumption was the basis for the error analysis. Please clarify or correct.
More specific minor comments:
9) L151 – what is the hexagon area?
10) L225: “iii) summing the minutely to hourly values” – do you mean the raw data or after the application of Eq 1,2, 3?
11) Eq 4 + L247: “and ððð is the number of lysimeters with missing data during time interval ð (-).”. It is not clear to me what is the definition of nia here.
12) L255: you should state that it is assumed the measurements are from a normal distribution with the same mean and standard deviation
13) It would be good to show the CDFs of hourly values of precipitation and ET.
14) Eq. 9: index i is missing
15) I think something is wrong with Eq. 10; please correct
16) Eq. 8-10: Use either small or capital letters consistently.
17) L350: Eq. 13 – should it be Eq. 14?