OCTOBER 2012 A D E Y E M I A N D J O E R G 1855

Analysis of Water Vapor over Using Radiosonde and Satellite Data

BABATUNDE ADEYEMI Department of Physics, Federal University of Technology, , Nigeria

SCHULZ JOERG Department of Climate Monitoring, Deutscher Wetterdienst, Offenbach, Germany

(Manuscript received 13 April 2011, in final form 13 April 2012)

ABSTRACT

The Satellite Application Facility on Climate Monitoring (CM-SAF) focuses on retrieving geophysical parameters from satellite data by using inversion schemes that are based on radiative transfer theory. In this study, the focus is on the daily mean vertically integrated water vapor [i.e., total precipitable water vapor (TPW)] and the monthly mean layered vertically integrated water vapor [i.e., low-level precipitable water vapor (LPW), midlevel precipitable water vapor (MPW), and upper-level precipitable water vapor] obtained from CM-SAF Germany for the 2004–08 period and for 20 radiosonde stations distributed over Nigeria. The mean annual cycle shows a structure with two maxima, in May and September, respectively, for all low- and midlevel water vapor in the coastal and Guinea savannah regions, whereas at the midland and Sahelian regions, an almost constant maximum lasting between May and September occurs. The two-maxima structure manifested by TPW, LPW, and MPW at the coastal and Guinea savannah regions and the single maximum exhibited by them at the midland and the Sahelian regions are in synchronism with the movement of the intertropical discontinuity. In addition, comparisons have been made between the precipitable water from CM-SAF (CM-SAF PWV) and precipitable water from radiosonde (RS-PWV). Comparisons of the products at the seasonal scale show that CM-SAF PWV at all levels is larger than that of RS-PWV. In addition, a seasonal intercomparison between them shows that CM-SAF PWV and RS-PWV agree significantly at all of the stations that were investigated.

1. Introduction accurately as possible. The water vapor–greenhouse feedback is considered the strongest positive feedback Water vapor is fundamental to the transfer of energy in the climate system. Recent climate-model simulations in the atmosphere (Rocken et al. 1997; Li et al. 2003). It (Rind et al. 1991; Del Genio et al. 1991; Shine and Sinha is one of the most important and most abundant green- 1991) lend theoretical support to a positive feedback house gases in the earth’s atmosphere, keeping the tem- among surface temperature, tropospheric water vapor, perature of the surface of the earth above the freezing and the greenhouse effect. Observational studies of the level. Atmospheric water vapor plays a key role in the feedback (Raval and Ramanathan 1989; Stephens and hydrological cycle, which in turn has a fundamental im- Greenwald 1991; Duvel and Breon 1991) have relied pact on the earth’s climate. The distribution of water on correlations between satellite measurements of vapor varies greatly both in space and time, with values column water vapor over open oceans and sea surface ranging from about 50 mm near the equator to less than temperature. one-tenth as much at the poles (Mockler 1995). These In geodesy, tropospheric delay in radio signal due to variations can lead to sudden changes in local weather. water vapor is known to be a major source of error for To develop accurate weather prediction and global geodetic observations from very-long-baseline in- climate models, it is vital to monitor water vapor as terferometry, radar altimetry, the global positioning system, and interferometric synthetic aperture radar Corresponding author address: Dr. B. Adeyemi, Dept. of Phys- (INSAR). INSAR techniques have been used to study ics, Federal University of Technology, Akure 34000, Nigeria. water vapor (Hanssen et al. 2001) despite the fact that E-mail: [email protected] water vapor serves as a major limitation to it. The

DOI: 10.1175/JAMC-D-11-0119.1

Ó 2012 American Meteorological Society Unauthenticated | Downloaded 09/29/21 01:26 PM UTC 1856 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 51 knowledge of the amount of atmospheric water vapor is 2. Data also essential in appreciating the significant roles it plays in the attenuation of microwaves in the atmosphere. a. Environmental setting of the study area Atmospheric water vapor and oxygen absorb radio For the purpose of this study, Nigeria was divided into waves in the frequency range of 100–50 000 MHz (Ajayi four zones reflecting areas that have climates controlled and Adeyemi 2009). Information on the climatic varia- by different mechanisms as explained by Olaniran and tion of water vapor will therefore be of great help in Sumner (1989) (see Fig. 1). radio engineering and communication. These areas are Precipitable water vapor (PWV) estimates can be obtained in a number of ways ranging from in situ 1) the coastal zone (CO; 4.68–6.58N, 3.78–7.18E), dom- measurements to remote sensing from satellites inated by tropical maritime (mT) air for most of the (Mockler 1995; Chaboureau et al. 1998). The radiosonde year, network has long been the primary in situ observing 2) the Guinea–savannah zone (GS; 6.68–7.88N, 3.98– system for monitoring atmospheric water vapor. Ra- 7.58E), where mT air dominates for about 7 months diosondes provide vertical profiles of meteorological and tropical continental (cT) air dominates for the parameters such as pressure, temperature, relative hu- remaining 5 months, midity, and, occasionally, wind information. The use of 3) the midland zone (ML; 8.08–10.88N, 3.98–9.88E), radiosondes is limited by their high operational costs, which is predominantly highland, where the cT air decreasing sensor performance during the harmattan mass dominates but where the topography effec- period when a dry easterly wind is blowing from the tively extends the length of the humid period because Sahara desert across the West African region during the of localized convection and orographic effects, and dry season, and their poor coverage over oceans and in 4) the Sahelian zone (SA; 11.08–13.08N, 3.78–13.38E), the Southern Hemisphere (Li et al. 2003). Radiosondes where the cT air mass predominates and the mT air are usually expected to measure PWV with an un- mass invades for between 3 and 5 months at most certainty of a few kilograms per meter squared, which (Olaniran and Sumner 1989). is considered to be the accuracy standard of PWV b. CM-SAF water vapor products (CM-SAF PWV) for meteorologists (Niell et al. 2001). These reasons, among others, have made spaceborne monitoring the CM-SAF provides global fields of daily mean verti- only effective way to assess water vapor distribution cally integrated water vapor and layered vertically in- on a global basis (Reuter 2005; Adeyemi 2008). Vari- tegrated water vapor in five layers (surface–850 hPa, ous missions have been deployed to monitor water 850–700 hPa, 700–500 hPa, 500–300 hPa, and 300– vapor amount [e.g., Television and Infrared Observa- 200 hPa). For this study, data have been extracted from tion Satellite Operational Vertical Sounder (TOVS), the global products for 20 radiosonde stations in Nigeria Advanced TOVS (ATOVS), Advanced Microwave (see Fig. 1) for the period 2004–08. Sounding Unit (AMSU-A and AMSU-B), and Special The CM-SAF water vapor products are generated Sensor Microwave Image (SSM/I)] (Chaboureau et al. through measurements from NOAA and EUMETSAT 1998; Randel et al. 1996). polar-orbiting platforms (Schulz et al. 2009). The ATOVS The objective of this paper is to study the variations suite of instruments [High Resolution Infrared Radiation of PWV over Nigeria as have been retrieved from Sounder (HIRS), AMSU-A/B, and Microwave Humidity ATOVS on board National Oceanic and Atmospheric Sounder (MHS)] on NOAA and Meteorological Oper- Administration (NOAA) and European Organisa- ational (MetOp) satellites represent infrared spectrom- tion for the Exploitation of Meteorological Satellites eters and microwave radiometers where the combination (EUMETSAT) satellites. These data will then be com- of all three instruments contains enough information pared with data from radiosondes (RS) that were to infer atmospheric profiles of temperature and specific taken over three stations spanning the major climatic humidity. regions in Nigeria. This is of importance in Nigeria, Temperature and water vapor profile retrieval mostly because the maintenance of the radiosonde network employ the one-dimensional variational technique that for national service is costly. Thus, if the radiosondes uses the variational principle to solve the retrieval can be replaced with reliable satellite-derived infor- problem. CM-SAF applies the scheme developed by mation, it would be a real benefit to the Nigerian Li et al. (2000) to TOVS/ATOVS observations. The government, because only a few stations delivering a so-called International ATOVS Processing Package (IAPP) higher quality of data might be retained for quality- scheme performs an inversion of the radiances to re- monitoring purposes. trieve simultaneously the temperature and humidity

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC OCTOBER 2012 A D E Y E M I A N D J O E R G 1857

FIG. 1. The selected stations in Nigeria and their climatic zones. profiles, as well as the surface temperature and cloud- c. Radiosonde water vapor products (RS-PWV) top pressure and amount. It employs an iterative The radiosonde data from three stations [ method that finds the maximum probability solution to (68289N, 38289E), Minna (98379N, 68309E), and a nonlinear retrieval/analysis problem. It can operate (12829N, 88309E)] were used for the in situ computation on cloud-free or, in some cases, cloudy radiances. The of PWV in this paper. These data were obtained from main satellite data source for the retrieval process de- the archives of the Department of Meteorological Ser- pends on the cloudiness of a scene and the underlying vices, Federal Ministry of Aviation, Oshodi, Lagos, Ni- surface. Retrievals over oceans rely on all sensors geria. Since the data collected in West Africa have many whereas retrievals over land surfaces are mainly based large spatial gaps, we have combined all available data on cloud-free HIRS measurements. The retrieval relies into monthly climatologies. This was done by finding the on an a priori background that is given by the 6-h monthly average of all available data for each of the forecast from the German numerical weather pre- stations. Note that these selected stations adequately diction model (Majewski et al. 2002). represent the regions they belong to in Nigeria. The Global daily fields are constructed using a specific PWV calculations at the three stations have been com- kriging algorithm (Schulz et al. 2009). The daily fields puted for the surface–850 hPa (W ), 850–700 hPa (W ), are accompanied by an uncertainty estimate that reflects a b 700–500 hPa (W ), 500–400 hPa (W ), and 400–300 hPa the retrieval uncertainty and the sampling error that is, c d (W ). Mean values of these quantities at the different in particular, important in tropical areas over land. e levels were then calculated from (Adeyemi 2008) Validation activities are routinely performed by the ð CM-SAF, but validation results are only presented in p N globally averaged numbers. On the global scale, radio- 5210 upper boundary 5210 å D Wa,b,c,d,e qy dp qy pi , (1) g g 5 sondes and satellite-derived water vapor agree reason- plower boundary i 1 2 ably well, with systematic differences of 0.5 kg m 2 and 22 21 root-mean-square difference of approximately 4 kg m . where qy is the specific humidity (g kg ) for the layer In this study, the CM-SAF product is tested over an area considered, W is in kilograms per meter squared, and Δp where the retrieval of water vapor is particularly difficult is the pressure gradient. because of the use of only HIRS observations with Observations from higher than 300 hPa were not used limited spectral coverage. because in many instances the radiosonde did not reach

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC 1858 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 51 that height. For this analysis, the five layers are com- station of Kano. UPW for all the zones has low and bined into three atmospheric layers: low-level water almost uniform values throughout the year. vapor (LPW) 5 Wa, midlevel water vapor (MPW) 5 The observations described above are in agreement Wb 1 Wc, and upper-level water vapor (UPW) 5 Wd 1 with the findings of Adedokun (1986), Balogun and We. The CM-SAF data are combined in the same way to Adedokun (1985), Balogun (1981), Olaniran and Sumner make the products comparable. The uppermost layer of (1989), Adeyemi and Aro (2004), and Adeyemi (2008) the CM-SAF data was not used. using radiosonde data taken over the regions. The in- creasing precipitable water vapor observed between January and May is linked to the movement of the rain- 3. Results and discussion producing zone known as the intertropical discontinuity (ITD) from the coast inland, and the decrease in water a. Distribution of total and layered PWV over Nigeria vapor occurring during September–December is due Figures 2a–d and 3a–d show the distribution of total to its retreat. The movement of the ITD is in close and layered precipitable water and the annual cycles association with the northward movement of the in- over the four climatic zones making up Nigeria, re- tertropical convergence zone (ITCZ) during boreal spectively. From Figs. 2a–d, TPW displays an increase in summer. The intervening minimum that is prominent water vapor from January to April–May in all four cli- in August in the CO and GS zones and less prominent matic zones. This increase is small in the CO and GS at the ML zone is well related to the period of dryness zones while in the ML and SA zones the increase is of 2–3 weeks occurring at the middle of the rainy season abrupt. The increase continues until June, after which in the coastal part of West Africa. This phenomenon, the values slightly decrease in the CO and GS zones. known as the intramonsoonal period, is believed to be April and May in these zones are associated with the first a consequence of several factors such as coastal up- peak of the rainy season. The ML and SA zones, be- welling and the northern advance of the subtropical high tween May and June, are characterized by high values of pressure systems of the southern Atlantic Ocean. It is TPW. These values are mostly higher than those in the also caused by the circulation aloft, which becomes CO and GS zones. Between July and September, the CO divergent and subsident as a result of the frequent oc- and the GS zones experience lower precipitable water, currence of inversions and isothermals in the upper with a minimum in August. In both the ML and SA atmosphere along the coast when the weather zone E1 zones, the opposite is the case. Here, high values ob- makes its appearance a short way inland from the coast served between May and June persist until the end of (Adedokun 1978, 1986; Balogun and Adedokun 1985; September. The August minimum observed at the CO Balogun 1981; Olaniran and Sumner 1989). and GS zones does not occur. The single maximum lasting between May and Between October and December, the CO and GS September observed at the ML and SA zones is also zones are characterized by an increasing water vapor a confirmation of the findings of past researchers (e.g., content that peaks in October and November for the Adeyemi and Adeyemi 2005; Adedokun 1986; Willoughby respective zones. This is associated with the second peak et al. 2002) over the regions. This single maximum of the rainy season in these zones. The ML and SA is due to the fact that during this period the ITD would zones, on the other hand, maintain a single peak lasting have reached its most northerly limit, and the region between May and September. would be under intense rainfall brought up by the Figures 3a–d, which show the mean annual cycle of ITD. layered precipitable water over Nigeria, present some At all of the zones, with the exception of the CO zone noteworthy characteristics at the different zones. In the during the dry season, MPW is seen to be higher than CO and GS zones, LPW and MPW display structures LPW throughout the year. This observation, whereby similar to their TPW. UPW, on the other hand, displays LPW is lower than MPW at all of the zones except the almost constant and low values throughout the year. CO zone during the dry season, may be explained us- These values increase only slightly during the rainy ing a dynamic effect such as the austauch phenomenon season of May–October. At the ML station of Minna the annual cycles of the layers also display similar structures to their TPW, with the exception of the MPW structure 1 West African weather, as demarcated by the movement of the for which the August minimum that was completely ITD (ITCZ), has been classified into zones depending on the lo- cation of a place with respect to the ITD (Olaniran and Sumner absent in the TPW showed up, although it is not as 1989). Its northward and southward movements bring about the conspicuous as it is in the CO and GS zones. LPW and onset and retreat of rain, respectively. These weather zones are MPW show similar annual cycles to their TPW at the SA known as A, B, C, D, and E, with E located over the oceans.

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC OCTOBER 2012 A D E Y E M I A N D J O E R G 1859

FIG. 2. Seasonal distributions of TPW over the (a) CO, (b) GS, (c) ML, and (d) SA zones of Nigeria as derived from the CM-SAF ATOVS water vapor product.

(i.e., lifting of the boundary layer). The persistent comparison with retrieved data from radiosondes, as heating of the lowest layers of the atmosphere through mentioned in section 2. High correlations (see Table 1) surface heating causes surface air to become hot. It ranging from 0.55 [coefficient of determination (CD) 5 expands and becomes less dense and then rises. Water 32%] to 0.93 (CD 5 88.4%) have been observed at all of vapor is then transported upward, resulting in its de- the levels of PWV and at all of the available stations pletion at the lowest layer of the atmosphere and except at the midlevel in Lagos and the upper level in making MPW larger than LPW. The dry-season re- Kano where the regressions have been found not to be versal at the CO zone is most likely because there is significant. The CM-SAF PWV at all of the levels ob- enough evaporation over the ocean to replace the served has been found to be significantly larger than uplifted water vapor. RS-PWV with very high standard deviation (STD) at the lower and middle levels of the atmosphere, whereas b. Comparisons between CM-SAF PWV and at the upper level the STD was low. This was observed radiosondes (RS-PWV) at all of the stations. The scattergram (Fig. 5) also in- Comparisons have been made between the two sour- dicates that CM-SAF PWV was larger than RS-PWV ces of water vapor estimates at the seasonal scale. It is with scale factors of 0.1572 6 0.07 in Lagos, 0.8766 6 important to note that the CM-SAF PWV values were 0.105 at Minna, and 2.416 6 0.696 in Kano at the low averaged over monthly intervals to correspond to level; 0.938 6 0.149 at Minna and 1.231 6 0.1411 at Kano available radiosonde data. at the midlevel; and 1.257 6 0.467 at Lagos and 1.255 6 Figure 4 and Table 1 show PWV from CM-SAF water 0.500 at Minna at the upper level (note that the re- vapor products averaged over the period of 2004–08 in gressions that are not significant are not considered).

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC 1860 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 51

FIG. 3. Seasonal distribution of layered precipitable water vapor averaged over 4 yr at the (a) CO, (b) GS, and (c) ML and (d) the SA zones of Nigeria as derived from the CM-SAF ATOVS water vapor product.

Figure 5 shows clearly the scatterplots of the differ- increasing tendency is noticeable, with a scale factor of ences in PWV (i.e., CM-SAF PWV 2 RS-PWV) at the about 15%. At the upper level, a linear tendency has different levels and for the stations for which radiosonde also been established with R2 5 0.76. In the case of Kano data were available. Figure 5 (top panel) shows that the (Fig. 5, bottom panel), an increasing tendency has been difference in PWV at higher values of CM-SAF PWV at established between the differences in PWV and CM- the lower level in Lagos was low whereas at the midlevel SAF PWV with high correlation ranging from 0.73 to it was high and at the upper level an increasing tendency 0.96. This shows that at this station the differences in- with a high coefficient of determination (R2 5 0.74) was crease with the amount of CM-SAF PWV. noticeable. That is, at this level, the difference in PWV c. Dry- and wet-season comparisons increases as CM-SAF PWV increases. Figure 5 (middle panel) shows that differences in PWV at Minna are Previous studies over West Africa (Adeyemi and Aro found within the same range for all values of CM-SAF 2004; Balogun and Adedokun 1985; Olaniran and PWV at the lower level whereas at the midlevel an Sumner 1989) have confirmed that there are seasonal

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC OCTOBER 2012 A D E Y E M I A N D J O E R G 1861

FIG. 4. Comparison between layered CM-SAF- PWV and RS-PWV over Nigeria. Only the stations for which radiosonde data are available have been considered.

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC 1862 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 51

TABLE 1. Values of best-fit parameters A and B in the regression equation of CM-SAF PWV at different levels on RS-PWV, i.e., CM- SAF PWV 5 A(RS-PWV) 1 B. Here, A and B are parameter estimates, SE is standard error, and p value is the probability value. In the remark column, S stands for significance and NS stands for not significant.

Station A SE B SE STD CD (%) p value Remark Low level Lagos 0.1572 0.0721 20.37 1.271 4.888 32.2 0.05 S Minna 0.8766 0.1048 10.03 1.337 6.242 87.5 0.00 S Kano 2.416 0.6960 25.037 5.284 5.530 54.6 0.01 S

Midlevel Lagos 0.2227 0.1660 20.87 2.006 7.132 15.3 0.21 NS Minna 0.9379 0.1487 10.32 2.138 6.498 79.9 0.00 S Kano 1.231 0.1411 6.282 1.711 7.153 88.4 0.00 S

Upper level Lagos 1.257 0.4665 2.267 0.1844 1.236 42.1 0.02 S Minna 1.255 0.5004 2.021 0.2765 1.151 38.6 0.03 S Kano 0.013 0.6410 2.739 0.3348 1.173 0.004 0.98 NS

2 differences in the quality of relative humidity measure- less than 10 kg m 2. This seems to be unrealistically small ments by radiosondes. To study the effect of the sea- when compared with European Centre for Medium- sonal differences, comparisons between CM-SAF PWV Range Weather Forecasts reanalysis results. A possible and radiosonde PWV were also made by separating dry- cause for the larger scale-factor variation at Lagos may and wet-season cases (see Table 2). During the dry be that, during the dry season, radiosonde sensors are season, linear relationships between CM-SAF PWV and heated up, resulting in lower relative humidity measure- RS-PWV were only found at Minna at both the lower ments than expected (Smout et al. 2001). and upper levels (see Table 2) whereas for other stations The satellite-derived product is also prone to problems— at all of the levels and for Minna at the midlevel the in particular, in areas such as Nigeria. The ATOVS ob- relationships are not significant. During the wet season, serving system and the associated retrieval system (Li no relationship was found between CM-SAF PWV and et al. 2000) use different sensor combinations over land RS-PWV at all levels and for all stations except at Kano and ocean and are restricted to cloud-free HIRS ob- at the midlevel and Lagos at the lower level, where servations (Schulz et al. 2009). The use of only infrared linear relationships of very high correlation have been (IR) measurements, employing the differential absorp- observed. By comparison, the scale factors for the RS- tion in the atmospheric window channels (11.11 and PWV relative to CM-SAF PWV measurements changed 12.47 mm) and using directly the absorption features in abruptly from a low value ranging from 0.08 to the water vapor channels (6.52 and 7.33 mm), limits the 2 1.5 kg m 2 at Lagos during the dry season to a high information content regarding the profile of the lower 2 value ranging from 1.5 to 9.2 kg m 2 at the same station and upper layers. The midlevel is almost left without during the wet season. information. The retrieval is then driven toward the background used in the iterative retrieval scheme of Li d. Discussion et al. (2000). The depth of this information gap becomes There are many potential causes for these differences larger the more humid the atmosphere is, because the that are almost impossible to discern from the available peaks of the water vapor channels are high up in the data. It is evident that the availability and quality of atmosphere and provide little information to the mid- radiosonde data in Nigeria have significant shortcom- level estimate. This effect might be observed in Fig. 4 for ings. In carrying out good quality assurance of the data, the Kano station where the difference in midlevel water series of averaging and reaveraging processes were vapor clearly increases during the wet season. carried out to filter the radiosonde data. The few avail- It is also known that changes in the IR surface emis- able radiosondes are characterized by large temporal sivity can strongly influence the retrieval of the layer- gaps. An example of this can be seen in Fig. 4 at the averaged water vapor contents. Bennartz et al. (2008) Lagos station where the low- and midlevel water vapor have shown that retrievals of total column water vapor contents dramatically dropped off in August. Also the using IR window channels have a sensitivity of 0.6 and 2 low-level water vapor at the Kano station shows a very 2kgm 2 % emissivity change in the two HIRS window small annual cycle with absolute water vapor values of channels, respectively. Using available IR emissivity

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC OCTOBER 2012 A D E Y E M I A N D J O E R G 1863

FIG. 5. Scatterplots of differences in PWV (diff in PWV) 5 CM-SAF PWV minus RS-PWV at the different levels for (top) Lagos, (middle) Minna, and (bottom) Kano. The least squares regression line was also shown.

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC 1864 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 51

TABLE 2. As in Table 1, but for the dry and wet seasons.

Station Season A SE B SE CD (%) p value Remark Low level Lagos Dry 1.463 1.097 217.44 24.81 37.2 0.2744 NS Wet 2.762 1.871 246.67 43.77 30.4 0.1999 NS Minna Dry 0.974 0.219 28.287 3.501 86.8 0.0213 S Wet 20.469 0.975 26.57 23.24 4.4 0.6596 NS Kano Dry 0.229 0.692 4.413 4.348 3.5 0.7625 NS Wet 0.208 0.134 4.817 2.391 32.4 0.1827 NS

Midlevel Lagos Dry 0.077 0.757 7.855 15.98 0.29 0.9315 NS Wet 9.169 4.571 2217.5 114.8 44.6 0.1011 NS Minna Dry 0.328 0.591 3.291 10.94 9.3 0.6173 NS Wet 0.432 0.487 5.421 13.01 13.6 0.4161 NS Kano Dry 0.287 0.867 2.868 12.14 3.5 0.7622 NS Wet 0.758 0.176 24.225 4.356 78.8 0.0076 S

Upper level Lagos Dry 0.118 0.170 20.030 0.409 13.9 0.5372 NS Wet 1.403 0.322 23.701 0.948 79.2 0.0073 S Minna Dry 20.402 0.2366 1.238 0.531 49.0 0.1882 S Wet 0.341 0.296 20.378 0.881 20.9 0.3019 NS Kano Dry 20.356 0.632 1.318 1.567 9.6 0.6124 NS Wet 20.262 0.176 1.315 0.518 30.8 0.1957 NS databases such as Seemann et al. (2008), Bennartz et al. A last possible cause for the discrepancies observed (2008) have shown that the variability of surface emis- mostly during the wet season between CM-SAF PWV sivity in the window channels is approximately 1%, and RS-PWV is associated with the comparison method leading to uncertainties on the order of the sensitivity in used here. The radiosonde PWV is the vertical inte- the window channels. The Li et al. (2000) retrieval as gration of specific humidity along its flight trajectory, for employed in CM-SAF uses a constant surface emissivity which the horizontal drift of the radiosonde is signifi- (0.99) for the window channels. Thus, already small cant. The collocation to an average of vertically integrated deviations from this value can lead to large changes in satellite-derived instantaneous water vapor profile based the low-level water vapor estimates. The Seemann et al. on the starting location of the radiosonde can lead to very (2008) emissivity database shows that the emissivity in large differences—in particular, for the upper-layer water the IR channels varies with changing vegetation cover, vapor estimates (Li et al. 2003). which is not considered in the satellite retrieval scheme. Bennartz et al. (2008) also showed that iterative re- 4. Conclusions trievals over land may introduce large artificial daily variations in retrieved column water vapor that were Satellite retrievals of total precipitable water over negatively correlated with surface temperature. This is Nigeria, as obtained from the archives of the CM-SAF, caused by the background and its error covariance in Deutscher Wetterdienst, Germany, have been found to the retrieval that do not match the local conditions. In increase from January to April–May at all of the four the presented comparison the satellite observations are climatic zones in Nigeria. This increase is a small one in taken under very different surface temperature condi- the case of the CO and GS zones whereas at the ML and tions. During the day, the morning NOAA satellite over- SA zones the increase is abrupt. Two maxima are dis- pass is looking at relatively cold surfaces, whereas the cernible in the mean annual cycles at all vertical levels at afternoon overpass is looking at a very hot surface, par- both the CO and GS zones, whereas at the ML and SA ticularly in the dry season. This of course can lead to biases, zones a single maximum was observed. Comparisons especially in the low-level water vapor estimate over arid between the PWV values derived from CM-SAF and surfaces where the surface temperature variation during radiosonde data were also performed. High correlations the day can be larger than 40 K. Since the comparison with (see Table 2) ranging from 0.55 (CD 5 32%) to 0.93 the radiosondes is only possible at seasonal scales, no more (CD 5 88.4%) have been observed at all of the levels of insight into this problem could be gained. PWV and at all of the available stations except at the

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC OCTOBER 2012 A D E Y E M I A N D J O E R G 1865 midlevel in Lagos and at the upper level in Kano, where ——, and O. E. Adeyemi, 2005: Variation of water vapour content the regressions are not significant. This shows that the over Nigeria. Zuma J. Pure Appl. Phys. , Nig., 7, 219–224. CM-SAF data are a better alternative in the regions Ajayi, O. S., and B. Adeyemi, 2009: Empirical modelling of surface refractivity over Akure using ground electrical conductivity. where there is good agreement between these data and Nig. J. Pure Appl. Sci., 5, 24–27. the radiosonde data in Nigeria. The CM-SAF data will Balogun, E. E., 1981: Seasonal and spatial variations in thunder- therefore provide the much-needed data in this part of storm activity over Nigeria. Weather, 36 (7), 192–197. the world where there is general lack of upper-air data. ——, and J. A. Adedokun, 1985: On the variations in precipitable The CM-SAF PWV at all of the levels observed has been water over some West African stations during the special ob- servation period of WAMEX. Mon. Wea. Rev., 114, 772–776. found to be greater than RS-PWV, with very high STD Bennartz, R., A. Walther, and M. Stengel, 2008: Potential im- at the lower and middle levels of the atmosphere and provements of retrieval methods for total and vertically low STD at the upper level. This observation was found resolved water vapor content over land surfaces from SEVIRI at all of the stations and may be likened to the presence (DVAP-III and DVAP-IV): Scientific and technical report— of a dry bias in the radiosonde humidity measurements final version 2.0. CMSAF Working Contract Rep., 47 pp. [Avail- able online at http://www.cmsaf.eu/bvbw/generator/CMSAF/ at both the upper and the lower levels of the atmo- Content/Publication/vs__pdf/WV__Report__SEVIRI-WV__ sphere (Ha¨berli 2006). CM-SAF PWV was larger than Bennartz__080229,templateId5raw,property5publicationFile. RS-PWV with scale factors of 0.1572 6 0.07 in Lagos, pdf/WV_Report_SEVIRI-WV_Bennartz_080229.pdf.] 0.8766 6 0.105 at Minna, and 2.416 6 0.696 at Kano at Chaboureau, J.-P., A. Che´din, and N. A. Scott, 1998: Remote the low level; 0.938 6 0.149 in Minna and 1.231 6 sensing of the vertical distribution of atmospheric water vapor from the TOVS observations: Method and validation. 0.1411 in Kano at the mid level; and 1.257 6 0.467 at 6 J. Geophys. Res., 103, 8743–8752. Lagos and 1.255 0.500 at Minna at the upper level. It Del Genio, A. D., A. A. Lacis, and R. A. Reudy, 1991: Simulations was generally observed also that at Lagos the differ- of the effect of a warmer climate on atmospheric humidity. ences between measurements of CM-SAF PWV and Nature, 351, 382–385. RS-PWV have an increasing tendency at the upper level Duvel, J. Ph., and F. M. Breon, 1991: The clear-sky greenhouse only, at Minna an increasing tendency is noticed at both effect sensitivity to a sea surface temperature change. J. Climate, 4, 1162–1169. the middle and the upper levels, and at Kano all of the Ha¨berli, C., 2006: Assessment, correction and impact of the dry levels show increasing tendency. This implies that the bias in radiosonde humidity data during the MAP SOP. Quart. differences in PWV measurements at these stations and J. Roy. Meteor. Soc., 132, 2827–2852. at the levels considered increase as PWV increases. Sea- Hanssen, R. F., A. J. Feijt, and R. Klees, 2001: Comparison of sonal effects on data acquisition revealed that, in some precipitable water vapor observations by spaceborne radar interferometry and Meteosat 6.7-mm radiometry. J. Atmos. cases, discrepancies exist between the PWV estimates Oceanic Technol., 18, 756–764. retrieved from both sources during the dry- and the Li, J., W. W. Wolf, W. P. Menzel, W. J. Zhang, H. L. Huang, and wet-season periods. T. H. Achtor, 2000: Global soundings of the atmosphere from ATOVS measurements: The algorithm and validation. J. Appl. Acknowledgments. The authors express their grati- Meteor., 39, 1248–1268. tude to the officials and management of the Department Li, Z., J.-P. Muller, and P. Cross, 2003: Comparison of precipitable water vapor derived from radiosonde, GPS, and Moderate- of Climate Monitoring at Deutscher Wetterdienst, Resolution Imaging Spectroradiometer measurements. J. Geo- Offenbach, Germany, for providing the data. The efforts phys. Res., 108, 4651, doi:10.1029/2003JD003372. of Markus Jonas, Dr. Marc Schroeder, and Nathalie Majewski, D., and Coauthors, 2002: The operational global Courcoux of the same department in making sure that icosahedral–hexagonal gridpoint model GME: Description the data were available in ASCII format during the visit and high-resolution tests. Mon. Wea. Rev., 130, 319–338. Mockler, S. B., 1995: Water vapor in the climate system. Amer. of the corresponding author to CM-SAF are also greatly Geophys. Union Special Rep. [Available online at http://www. appreciated. eso.org/gen-fac/pubs/astclim/espas/pwv/mockler.html.] Niell, A. E., A. J. Coster, F. S. Solheim, V. B. Mendes, P. C. Toor, REFERENCES R. B. Langley, and C. A. Upham, 2001: Comparison of mea- surements of atmospheric wet delay by radiosonde, water Adedokun, J. A., 1978: West African precipitation and dominant vapor radiometer, GPS, and VLBI. J. Atmos. Oceanic Tech- atmospheric mechanisms. Meteor. Atmos. Phys., 27, 289–310. nol., 18, 830–850. ——, 1986: On a relationship for estimating precipitable water Olaniran, O. J., and G. N. Sumner, 1989: A study of climatic vari- vapour aloft from surface humidity over West Africa. ability in Nigeria based on the onset, retreat, and length of the J. Climatol., 6, 161–172. rainy season. Int. J. Climatol., 9, 253–269. Adeyemi, B., 2008: Empirical modeling of layered integrated water Randel, D. L., T. J. Greenwald, T. H. V. Haar, G. L. Stephens, vapor using surface mixing ratio in Nigeria. J. Appl. Meteor. M. A. Ringerud, and C. Combs, 1996: A new global water Climatol., 48, 369–380. vapor dataset. Bull. Amer. Meteor. Soc., 77, 1233–1254. ——, and T. O. Aro, 2004: Variation in surface water vapour density Raval, A., and V. Ramanathan, 1989: Observational determination over four Nigerian stations. Nig. J. Pure Appl. Phys., 3, 38–44. of the greenhouse effect. Nature, 342, 758–761.

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC 1866 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 51

Reuter, M., 2005: Identification of cloudy and clear sky areas in MSG surface emissivity database for application to clear-sky sound- SEVIRI images by analyzing spectral and temporal information. ing retrievals from multispectral satellite radiance measure- Ph.D. thesis, Freie Universitat Berlin, Berlin, Germany, 132 pp. ments. J. Appl. Meteor. Climatol., 47, 108–123. [Available online http://www.diss.fu-berlin.de/2005/194.] Shine, K. P., and A. Sinha, 1991: Sensitivity of the earth’s climate to Rind, D., E. W. Chiou, W. Chu, J. Larsen, S. Oltmans, J. Lerner, height-dependent changes in water vapour mixing ratio. Na- M. P. McCormick, and L. McMaster, 1991: Positive water ture, 354, 382–384. vapour feedback in climate models confirmed by satellite data. Smout, R., J. Elms, D. Lyth, and J. Nash, 2001: Met Office Nature, 349, 500–503. RS90 humidity sensor evaluations. New technology in up- Rocken, C., T. V. Hove, and R. Ware, 1997: Near real-time GPS per-air observations report. Met Office, Berkshire, United sensing of atmospheric water vapor. Geophys. Res. Lett., 24, Kingdom, 4 pp. 3221–3224. Stephens, G. L., and T. J. Greenwald, 1991: The earth’s radiation Schulz, J., and Coauthors, 2009: Operational climate monitoring budget and its relation to atmospheric hydrology. 1. Obser- from space: The EUMETSAT Satellite Application Facility vations of the clear sky greenhouse effect. J. Geophys. Res., 96, on Climate Monitoring (CM-SAF). Atmos. Chem. Phys., 15 311–15 324. 9, 1687–1709. Willoughby, A. A., T. O. Aro, and I. E. Owolabi, 2002: Seasonal Seemann, S. W., E. E. Borbas, R. O. Knuteson, G. R. Stephenson, variations of radio refractivity gradients in Nigeria. J. Atmos. and H.-L. Huang, 2008: Development of a global infrared land Sol. Terr. Phys., 64, 417–425.

Unauthenticated | Downloaded 09/29/21 01:26 PM UTC