-
Notifications
You must be signed in to change notification settings - Fork 1
/
Chapter_10_Comparing_Data_from_Different_Sources.qmd
84 lines (68 loc) · 4.54 KB
/
Chapter_10_Comparing_Data_from_Different_Sources.qmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
# Comparing Data from Different Sources
## Introduction
The satellite and reanalysis data, discussed in Chapter 9, provides a
wonderful resource that can supplement the historical station data that
is described in this guide. The satellite data is usually from the early
1980s, while some of the reanalysis data is from 1950. Table 10.1
summarises some sources of rainfall data:
-----------------------------------------------------------------------
Table 10.1
-----------------------------------------------------------------------
{width="6.268055555555556in"
height="1.9694444444444446in"}
-----------------------------------------------------------------------
Some of these products also include other elements, including
temperatures and ERA5 is for many elements.
These data are already used extensively. However, often users access
only one type, i.e. either station or satellite/reanalysis. This is
often either because that is what the researcher is comfortable with, or
only one type is easily available. We consider here how station and
satellite data can be compared and then perhaps used together. There are
a range of possible objectives from these comparisons including the
following:
a) The satellite/reanalysis data, from the same location as a ground
station, can perhaps be considered as an additional station. As
such, perhaps the data can be used to complete, or infill missing
values in the station data.
b) Similarly, perhaps this new (satellite) station could be used to
support the quality-control of the station data.
These objectives may be more interesting for countries where there is a
relatively sparse station network. Where the network is dense,
neighbouring ground stations may be used for these objectives.
c) The bonus is that the satellite data does provide a dense network.
For example, for CHIRPS the estimated daily rainfall data is on
roughly a 5km square, so the equivalent of about 400 (pseudo)
stations per square degree. Hence it provides estimated daily
rainfall data for the whole of Africa, and beyond with a pseudo
station that is always close to any given location.
Comparisons between station and gridded data must recognise that they
have not measured the same thing. Station data are measured at a point,
while gridded data represent an area. The size of the area depends on
the method with an example shown in Fig. 10.1a.
In Barbados, Fig. 10.1a the point shown is a station called Husbands,
the site of a regional climate centre. CIMH. The largest pixel is for
the ERA5 reanalysis data and the smallest is for CHIRPS. This figure
also shows that the pixel in coastal sites can sometimes be largely over
the ocean and hence a neighbouring pixel may be more relevant.
-------------------------------------------------------------------------------------------------------------
***Fig. 10.1a Pixel size for 3 methods in Barbados*** ***Fig. 10.1b Difference between gridded and point data for rainfall***
------------------------------------------------------ ------------------------------------------------------
{width="3.094673009623797in" {width="2.7712182852143483in"
height="2.188830927384077in"} height="2.506793525809274in"}
-------------------------------------------------------------------------------------------------------------
Fig 10.1b illustrates a reason for possible differences between area and
point data for rainfall. The sketch shows a cloud, and hence possibly
rain in part of the pixel, but not at the station in the top left. Hence
the station may be zero, while the gridded data notes some rain. Thus,
unless the satellite data are adjusted, we would expect more rain days
(and potentially less extreme values) than at a point. This feature is
particularly for rainfall, but may also be shown for other elements,
such as sunshine hours, where there may be zeros in the data.
The problem that is addressed in this chapter is essentially just the
comparison of two variables, i.e. 2 columns of data, where the first is
the station and the second is the satellite, or reanalysis data. This is
essentially the same problem as in forecasting, where the forecast is
compared with the actual data. Many of the methods are from software
that was originally constructed for the forecasting problem.
From a statistical point of view this problem is just the same as
comparing