Adequacy of the daily TMPA 3B42 high-resolution satellite precipitation product for monitoring hydrometeorological hazards in the Southeast Sahel of Africa

Figures & data

Figure 1. Location of our study area. (a) Regional map for the vicinity of Gedaref City (open circle) in Southeast Sahel showing daily rainfall for 27 July 2006 ($\ge$5 mm), based on TMPA 3B42 estimates. Elevation contour interval is 200 m, with Gedaref State shown by the hatched area. (b) Local map showing the location of three cooperative gauge sites in addition to the principal WMO Gedaref gauge. Elevation contour interval is 50 m.

Figure 2. Daily precipitation time series for the TMPA 3B42 product (panel (a)) compared with GED WMO gauge values (panel (b)) for the 1999 − 2015 epoch. Panel (c) shows daily differences or deviations between pairs of the principal daily rainfall metrics of this study: gauge and 3B42, where (3B42-GED gauge).

Figure 3. Regression analysis between the principal daily rainfall metrics of this study (3B42 versus GED gauge values) for the period of record 1999 − 2015. Data pairs are purged of couplets having zeros in common. Dashed line is only for reference showing a unit slope.

Figure 4. Enlarged overlays, 5 July 2006 to 18 August 2006, of GED gauge and 3B42 daily time series, around the date 27 July 2006, the maximum storm-day of record of the 3B42 database for the period of record, 1999 − 2015. See Figure 1 for the areal coverage of the 27 July 2006 event.

Table 1. Compare estimated intensity of the seven maximum SPE events with highest ranked gauge observations: 1999−2015.

	GED gauge		TMPA 3B42 SG
Rank	Date	GED Gauge (mm)	Date	3B42 (mm)
1	19 August 1999	106	27 July 2006	118
2	17 July 1999	92	19 August 1999	102
3	20 July 2010	90	13 August 2003	101
4	5 October 2014	78	24 July 2005	100
5	10 July 2008	73	8 September 2007	86
6	19 July 2003	72	4 July 1999	80
7	26 July 2003	70	24 July 2003	78

Table 2. 1999–2015 Summary of 3B42 misassignment of daily rainfall. 77 days year-1. Annual bias = 2.2%.

		GED gauge	3B42	Units
1	Mean annual	631	645	mm year^-1
2	Σ(False alarms)	299		mm year^-1
3	Σ(Missed events)	−285		mm year^-1
4	Σ(AD)^a	584		mm year^-1
5	Metric for scale of daily event^b
6	RMSDVc	15.4	15.6	mm day^-1
7	RMSDd	13.1		mm day^-1
8	Metric for scale of daily event^b
9	Mean daily	8.2	8.4	mm day^-1
10	MAD^‖	7.6		mm day^-1

^a Σ(AD): Annual sum (total) of AD (absolute deviations).

^b Uses mean annual number of storm-days of 77.

^c RMSDV: RMS of daily values (DV).

^d RMSD: RMS of deviations (D).

‖ MAD: Mean absolute (daily) deviation.

Table 3. Cumulative statistics on the daily deviation of 3B42 from GED gauge (1301 events; 1999−2015) .

Fraction of number of storm-days		Fraction of annual total deviation by volume
Daily total (mm)	Fraction of storm-days^a	Daily total (mm)	Fraction of annual total anomaly^b
−38	0.01	−88	0.01
−32	0.02	−60	0.02
−21	0.05	−50	0.05
−12	0.10	−36	0.10
−3.2	0.25	−18	0.25
0.5	0.50	1.5	0.50
4.1	0.75	15	0.75
11	0.90	36	0.90
18	0.95	51	0.95
32	0.98	71	0.98
41	0.99	80	0.99

^a The mean annual number of daily storms is 77.

^b The annual total anomaly (Σ(MAD)) is 584 mm.

The total annual gauge rainfall is 631 mm.

Figure 5. Statistical distributions of the signed daily deviations of 3B42 from gauge values p plotted along the abscissa (x-axis), where the quantile with respect to the ordinate (y-axis) is the fraction of the respective base total, which is either the annual mean number of storm-days per year, or the total annual anomaly or misrepresented rainfall per year.

Table 4. Compare summary statistics for the principal GED data set for 1999–2015, with complementary data sets from three cooperative gauge sites for 2010–2013.

	1999–2015	2010–2013
Parameter	GED	GED	GHD	ERW	DOK
N	1301	280	257	256	358
r	0.50	0.60	0.34	0.21	0.35
Slope	0.50	0.48	0.23	0.15	0.25
Intercept	4.30	3.21	5.50	6.79	5.96
Bias	2%	−10%	−11%	−19%	−7%
CV	1.59	1.25	1.83	1.78	1.85
COM	0.93	0.81	1.11	1.20	1.05