Читать книгу Congo Basin Hydrology, Climate, and Biogeochemistry - Группа авторов - Страница 37

The data utilized in this study include two satellite estimates of precipitation and two gauge‐data sets. This array of indicators is considered for several reasons. For one, the satellite record includes only a relatively recent period while the gauge data is very sparse in recent years. Also, the various satellite products differ with respect to temporal and spatial resolution and are hence applicable to different analyses. Finally, because of uncertainties in all of the data products, the most realistic characterization of the Congo hydrologic regime can be gleaned by the combination.

Figure 3.3 Vertical profiles of omega (hPa/s × 10^–2, negative values = ascent) during MAM (bottom) and ON (top) averaged between 10°N and 10°S (from Nicholson, 2017, based on NCEP Version 1). Similar patterns are evident in MERRA 2 and ERA5.

Source: Jackson et al., 2009. ©American Meteorological Society. Used with permission.

Figure 3.4 The African Easterly Jet‐South: mean wind (m/s) at 600 hPa during October (from Jackson et al., 2009).

Source: Nicholson et al., 2019. © American Meteorological Society. Used with permission.

The lack of gauge availability in recent years is a serious problem. Besides the lack of raw rainfall records, this impacts most satellite estimates because they are merged with or adjusted by gauge data (Nicholson et al., 2019). Data were plentiful throughout equatorial Africa during the period 1947–1972 but gauge networks in the region have steadily declined since that time (Camberlin et al., 2019; Malhi et al., 2013; Nicholson et al., 2018a; Washington et al., 2013). Figure 3.6 illustrates the steady decline in gauge availability. It shows the stations available in the major archives, such as the NIC131 gauge archive of the author or GPCC (Schneider et al., 2015). For the period 1947–1972 far more stations than indicated were available, but not entered into the major archives. For example, during the 1940s through the 1960s some 500 stations were operative in the Democratic Republic of the Congo (DRC) (Bultot, 1971). For the more recent periods additional stations were operative in some countries, but for the DRC, Angola, and South Sudan all available stations are shown. During the 1970s, the available gauge network in the DRC had declined tremendously and by the early 21st century, only some 12 gauges were operative. The meteorological network in neighboring Angola suffered a similar fate, as the meteorological service barely functioned over several decades.

Figure 3.5 Mean vertical motion (omega: hPa/s × 10^–2) in SON at 850 hPa during mid‐afternoon and evening (from Jackson et al., 2009; ©American Meteorological Society. Used with permission).

Figure 3.6 Gauge network around the Congo Basin in four time periods (from Nicholson et al., 2018; © American Meteorological Society. Used with permission). Country boundaries are shown. In the map for 1999–2014, the letters DRC and A respectively identify the Democratic Republic of the Congo and Angola.

Gridded rainfall data sets do cover the recent years, despite the paucity of gauge data. Examples are the CRU data set (Harris et al., 2014) and the GPCC data set (Schneider et al., 2015). The gridding is based primarily on techniques that assume linear relationships between available stations, a weak assumption when the gaps are as large as those over the Congo Basin. Nicholson et al. (2018a) created a gridded data set using a spatial reconstruction technique that allows for more complex relationships among the available stations. This data set, which is at 2.5‐degree resolution, has been validated and covers the years 1921 to 2014. Termed NIC131‐gridded, this data set is available at monthly, seasonal, and annual time scales and can be obtained from the author.

Numerous satellite precipitation products cover equatorial Africa. Unfortunately, over the Congo Basin the rainfall estimates of the various products differ greatly, much more so than over other areas of equatorial Africa. This is illustrated in Figure 3.7, which shows rainfall estimates from several products for a single month (March 2001) and Figure 3.8, which depicts interannual variability based on the various products.

The nine satellite estimates presented in Figure 3.7 show some striking contrasts. These are most pronounced over the Democratic Republic of the Congo (DRC), where there are few gauge stations. CMORPH CRT (Xie et al., 2017) and TRMM 3B43 V7 (Huffman et al., 2007, Huffman & Bolvin, 2014) show numerous areas where rainfall is below 80 mm, while PERSIANN CDR (Ashouri et al., 2015), ARC2 (Novella & Thiaw, 2013), and TAMSAT V3 (Maidment et al., 2017) show rainfall on the order of 140 mm or more throughout most of the country. In general, rainfall increases southward, but in RFE (Love et al., 2004) this trend is reversed. The gauge products (NIC131‐gridded, GPCC) highlight drier conditions in the northwest, a feature clearly captured by CHIRPS2 (Funk et al., 2015), GPCP (Adler et al., 2003), and to a lesser extent TRMM 3B43 V7 and CMAP Enhanced (Xie et al. 2003; Xie & Arkin, 1997).

Figure 3.7 Maps of rainfall (mm/mo) for March, based on nine satellite products and three gauge products (GPCC, individual stations, NIC131‐gridded) (from Nicholson et al., 2019; © American Meteorological Society. Used with permission).

Figure 3.8 Interannual variability of rainfall (mm/yr) over the central Congo Basin and over a large portion of the Central African Republic (modified from Nicholson et al., 2019). Rainfall is averaged for March–April (left) and for October–November (right). The number of available gauge stations in each year is indicated below each graph, with the top graph indicate the number in GPCC and the bottom indicating the numbers in the NIC131 archive.

Source: Jackson et al., 2009. © American Meteorological Society. Used with permission.

Figure 3.8 shows the interannual variability of March/April and October/November rainfall over the Congo Basin and over the Central African Republic (CAR) to the north. The gauge network is dense over CAR and the satellite products are in good agreement with each other and with the NIC131‐gridded data set. For the Congo Basin, a fair amount of gauge data was available until the mid‐1990s, after which time there is wide disparity among the estimates and little agreement with the NIC131‐gridded data set. These results suggest that the reason for the poor performance of satellite products in this region is the paucity of gauge data.

When the products shown above were validated against gauge data (Camberlin et al., 2019; Nicholson et al., 2019), the best performing products appeared to be CHIRPS for mean rainfall, TRMM for daily rainfall, and CHIRPS2 and TRMM for interannual variability. Both products show generally good agreement with gauge data over the Congo on monthly time scales and thus are selected for use in this study.

Most rainfall analyses in this chapter are based on the CHIRPS2 satellite product (Funk et al., 2015). It has a spatial resolution of 0.05° × 0.05° and a daily temporal resolution. CHIRPS2 begins in 1981 and extends through 2019. However, TRMM 3B43 Version 7 is used to evaluate rainfall over the Amazon. It runs from 1998 to 2014 and has a spatial resolution of 0.25 degrees of latitude/longitude. Its temporal resolution is monthly. Both CHIRPS2 and TRMM 3B43 Version 7 have been extensively validated over equatorial Africa and show a close relationship to gauge rainfall (e.g., Camberlin et al., 2019; Nicholson et al., 2019). TRMM 3B42 V7 is used here to ascertain the diurnal cycle of rainfall. Its successor from the global precipitation measurement mission (GPM), IMERG, has much higher temporal and spatial resolution and is available since 2014. However, it has not been validated over the Congo Basin and for that reason TRMM is used instead.