Observational Perspectives from U.S. Climate Reference Network (USCRN) and Cooperative Observer Program (COOP) Network: Temperature and Precipitation Comparison

Synthesis

1. The main analysis is about the differences between the two networks, which would be an inhomogeneity if the USCRN had replaced the COOP network and thus informs us about the nature of inhomogeneities. That is why this article is in the category “inhomogeneities”. It also contains one figure on trend difference between the USCRN data and homogenized COOP data, which is an new way to validation homogenization algorithms. The paper is thus also categorized as “validation”.
   Currently only one review available. So no quantitative synthesis can be made yet.

   Reply

Reviews

1. The article describes a study comparing 12 pairs of stations from the USCRN and the COOP network for temperature and precipitation located with 500 m of each other. The USCRN is a climate observational network with high quality standards and homogeneity, which was set up to study climate change in America and was completed in 2005. The COOP network is a historical network, which was set up for (agricultural) meteorology in the USA. From the COOP data the US Historical Climate Network (USHCN) is generated by homogenization.

   The naturally ventilated COOP temperature measurements (both Stevenson screen and multiplate) have a clearly stronger diurnal cycle. While the general picture suggests that radiation errors are important, for individual stations several other causes are discussed. The results are consistent with the change from Stevenson screens to MMTS systems. Seasonal effect were not just due to insolation, but also seasonal wind patterns (ventilation).

   In case of precipitation the USCRN measures 1.5% more precipitation than the COOP network, especially during winter (frozen precipitation). The USCRN reports 16 more precipitation days. This is due to a combination of less undercatchment (beter wind shields), but more evaporation losses. Again clear differences from station to station.

   To understand the reasons for the differences (inhomogeneities) it is helpful to study high resolution data. Especially in case of precipitation there can be differences due to differences in reading time from the official time. This problem was elegantly reduced in this study by studying rain events rather than daily rain sums. This makes the time scale a bit larger, but reduced a lot of short-term noise hindering the interpretation. Could this idea also help daily homogenisation (correction) methods? The article mentions it would be worthwhile to study the influence of errors in reading times and it may be better if the observers record their observing times every day to make further studies more accurate.

   In 2006 there was a small change in the way the minimum and maximum temperature were computed in the USCRN. They are planning improving the method to estimate precipitation amounts. That such changes are even necessary at reference stations illustrates that it is important to store the data in the rawest form available so that in case of changes in the processing it is possible to recompute older values.

   As a side result, the study also shows one figure comparing the annual mean temperature anomalies. Even if only 8 years were available, the changes in the temperature anomalies match very well and the USCRN even shows a bit more warming between 2005 and 2012, which is nowadays (in 2018) even clearer (although likely not yet significant). Together with other evidence, it does make a case that we should look more at cooling biases in the raw observations that may lead to the underestimation of actual warming.

   This is an important study because the stations are compared in the wild. Most studies on parallel data are about multiple instruments measuring side by side often at weather services with good maintenance and observations made by the same person. The results for this study, however, were quite similar to side by side studies.

   It is a well executed study, however for many of the reasons for the differences it was hard to go beyond noting that certain measurement problems would produce a similar pattern, without being able to pin it down.

   Impact on the larger scientific community. [60]
   This type of studies is really important to understand the measurement processes behind our climate observations. That is also why I started the Parallel Observations Science Team. Unfortunately, this requires a large number of studies because the instrumentation, operations and climate is different everywhere. Single studies thus unfortunately do not make much impact by themselves. For all of climatology the influence of these studies is indirect through higher quality homogenised datasets.

   Contribution to the scientific field of the journal. [70]
   For the homogenisation community it is important to have multiple independent lines of evidence. However, as noted above many such studies are needed to arrive at clear conclusions.

   The technical quality of the paper. [70]
   The study is well executed. Its main weaknesses are inherent in the limited amount of data available (only 12 pairs of stations). At the same time, this number is too large to go in depth to understand the differences in detail. The data is freely available in the published COOP and USCRN datasets, but the selected dataset that was studied here, not the analysis software was published.

   Reply