In #10 we saw an overview which included the idea that a rainfall deficit is one part of a soil moisture deficit. But it’s the part we can measure. AR6 says, p.1573: Global studies generally show no significant trends in SPI [drought index for rainfall] time series (Orlowsky and Seneviratne, 2013; Spinoni et al., 2014), and in derived drought frequency and severity data (Spinoni et al., 2019), with very few regional exceptions.
I'm not clear. I've not delved into it at all, but here are some extracts from Trenberth 2014:
"The discrepancies between the previous studies are also prob- ably due to the precipitation data sets used. Sheffield et al.20 used four different precipitation data sets: CPC-Prec/L, GPCCv4, CRU TS3.10 and U. Delaware v2.01. Van der Schrier et al.21,29 used an improved version of the CRU data set (CRU TS3.10.01), whereas Dai4,5 explored other data sets (see Methods). The Climate Research Unit (CRU) has very recently updated their precipitation data set and analysis to version 3.21. The global mean land precipitation anomalies from several data sets (Fig. 1) are fairly consistent from 1950 to 1990 (although mean alignment is guaranteed for the base period 1961–1990). However, differences become readily apparent after 1991, when data from fewer stations are available for all data sets. The much greater number of Global Precipitation Climatology Centre (GPCC) stations does not guarantee improved coverage if the extra stations are all in the same area. It may be argued that fewer, more homogeneous, records provide more reliable time series and this has been the rationale behind the construction of the CRU data set. Many of the stations used by GPCC are not available for use by others. Coverage certainly affects the analyses of precipitation anomalies but continuity, or more strictly lack of temporal continuity, is more of an issue..
..The CRU TS3.10.01 data set, which has measurements from fewer than 1500 rain gauges in recent years, differs substantially from the GPCC and Global Precipitation Climatology Project (GPCP) precipitation products that have many more gauge data. This data set effectively had a ‘wet bias’ with respect to the other data sets in the global average since around 1996. The problem occurred especially at northern latitudes and the tropics. Maps for 2002 (not shown) reveal the biggest differences in tropical South America, Indonesia and parts of Africa.
This issue has been reduced in the newer version (CRU TS3.21), which has an increased number of stations in these regions and globally (>2400 stations during the 2000s)."
And then in the methods section:
"For GPCC up to 40,000 stations have been used although with a drop in the past decade and especially after 2009. The CRU data set typically has between 5000 and 7000 stations in CRU TS3.10.01, increasing somewhat up to about 10,000 at times for CRU TS3.21, but dropping after 1991.
However, coverage for CRU TS3.21 is a substantial improvement in the past decade, although 2011 values are still tentative. The CRU data set does not use the daily SYNOP sources (used by other data sets) because a determination of how many days can be used for a complete month has to be made.
All data sets show a reduction in station numbers in the last two decades, but this should not be taken to mean the network is degrading."
SoD, Do you have any idea why the number of data stations anomalies went way down over time? (Figure 1)
Seems like the number of data stations would increase over time.
John,
I'm not clear. I've not delved into it at all, but here are some extracts from Trenberth 2014:
"The discrepancies between the previous studies are also prob- ably due to the precipitation data sets used. Sheffield et al.20 used four different precipitation data sets: CPC-Prec/L, GPCCv4, CRU TS3.10 and U. Delaware v2.01. Van der Schrier et al.21,29 used an improved version of the CRU data set (CRU TS3.10.01), whereas Dai4,5 explored other data sets (see Methods). The Climate Research Unit (CRU) has very recently updated their precipitation data set and analysis to version 3.21. The global mean land precipitation anomalies from several data sets (Fig. 1) are fairly consistent from 1950 to 1990 (although mean alignment is guaranteed for the base period 1961–1990). However, differences become readily apparent after 1991, when data from fewer stations are available for all data sets. The much greater number of Global Precipitation Climatology Centre (GPCC) stations does not guarantee improved coverage if the extra stations are all in the same area. It may be argued that fewer, more homogeneous, records provide more reliable time series and this has been the rationale behind the construction of the CRU data set. Many of the stations used by GPCC are not available for use by others. Coverage certainly affects the analyses of precipitation anomalies but continuity, or more strictly lack of temporal continuity, is more of an issue..
..The CRU TS3.10.01 data set, which has measurements from fewer than 1500 rain gauges in recent years, differs substantially from the GPCC and Global Precipitation Climatology Project (GPCP) precipitation products that have many more gauge data. This data set effectively had a ‘wet bias’ with respect to the other data sets in the global average since around 1996. The problem occurred especially at northern latitudes and the tropics. Maps for 2002 (not shown) reveal the biggest differences in tropical South America, Indonesia and parts of Africa.
This issue has been reduced in the newer version (CRU TS3.21), which has an increased number of stations in these regions and globally (>2400 stations during the 2000s)."
And then in the methods section:
"For GPCC up to 40,000 stations have been used although with a drop in the past decade and especially after 2009. The CRU data set typically has between 5000 and 7000 stations in CRU TS3.10.01, increasing somewhat up to about 10,000 at times for CRU TS3.21, but dropping after 1991.
However, coverage for CRU TS3.21 is a substantial improvement in the past decade, although 2011 values are still tentative. The CRU data set does not use the daily SYNOP sources (used by other data sets) because a determination of how many days can be used for a complete month has to be made.
All data sets show a reduction in station numbers in the last two decades, but this should not be taken to mean the network is degrading."
Thank you.