A 1-year longδ18O record of water vapor in Niamey (Niger) reveals insightful atmospheric processes at different timescales



[1] We present a 1-year long representativeδ18O record of water vapor (δ18Ov) in Niamey (Niger) using the Wavelength Scanned-Cavity Ring Down Spectroscopy (WS-CRDS). We explore how local and regional atmospheric processes influenceδ18Ov variability from seasonal to diurnal scale. At seasonal scale, δ18Ovexhibits a W-shape, associated with the increase of regional convective activity during the monsoon and the intensification of large scale subsidence North of Niamey during the dry season. During the monsoon season,δ18Ovrecords a broad range of intra-seasonal modes in the 25–40-day and 15–25-day bands that could be related to the well-known modes of the West African Monsoon (WAM). Strongδ18Ovmodulations are also seen at the synoptic scale (5–9 days) during winter, driven by tropical-extra-tropical teleconnections through the propagation of a baroclinic wave train-like structure and intrusion of air originating from higher altitude and latitude.δ18Ovalso reveals a significant diurnal cycle, which reflects mixing process between the boundary layer and the free atmosphere during the dry season, and records the propagation of density currents associated with meso-scale convective systems during the monsoon season.

1. Introduction

[2] Although crucial progress have been made in understanding atmospheric processes over West Africa [Lafore et al., 2010], several questions prevail. Quantification of the relative effect of large-scale circulation, convection processes such as rain evaporation, and continental recycling on the atmospheric water cycle budget is not perfectly known from diurnal to seasonal scales [Risi et al., 2010a]. Teleconnections between the WAM and extra-tropical modes of intra-seasonal variability are not well understood [Janicot et al., 2010; Chauvin et al., 2010].

[3] The isotopic composition of Niger precipitation (δ18Op) has been recorded during the 2006 Monsoon season [Risi et al., 2008a, 2010b]. Those latter studies showed the potential of water stable isotopes to further examine convective processes and organization. However, to further disentangle the various atmospheric controls on δ18Op, to better document the evolution of moistening processes by rain evaporation, to better examine mesoscale subsidence in convective systems, to better know the origin of water vapor or to better evaluate convective processes in climate models, the isotopic composition of water vapor (δ18Ov) is needed.

[4] We extend here the previous observations through an original continuous 1-year long representativeδ18Ov (accuracy of ±0.25‰) [Tremoy et al., 2011] recorded in Niamey (Niger), at the Institut des RadioIsotopes (IRI, 13.31°N 2.06°E, 218 m.a.s.l) using a Picarro laser instrument (L1102-i model) [Gupta et al., 2009] from 2 July 2010 to 12 May 2011. δDv was also recorded but we focus here on δ18Ov. Precipitation was also collected on an event-based resolution and analyzed with an accuracy of ±0.05‰ forδ18Op. Our paper explores the δ18Ov-δ18Op relationship, discusses the potential of δ18Ovat the seasonal, intra-seasonal and diurnal timescales and examines the comparison between our observations and a nudged LMZ-iso simulation [Risi et al., 2010c].

2. Representativeness of Surface Data

[5] We present here near surface measurements (∼8 m above the ground) that may lead to important limitations. However, we argue that near-surfaceδ18Ov can capture atmospheric processes. First, Risi et al. [2008b] clearly showed that water vapor in the subcloud layer is highly influenced by processes such as reevaporation in convective downdrafts. Second, by comparing observed δ18Ov and δ18Op from July to October 2010 (Figure 1 and see next section), we show that surface water vapor variations are not disconnected from the water vapor from which precipitation forms in altitude (r2 = 0.66, n = 39). Third, in the LMDZ-iso simulation described byRisi et al. [2010c], the correlation between δ18Ov in Niamey grid point at 1000 hPa and the same values at 850 hPa is of 0.86, and decreases to 0.62 (0.33) at 750 hPa (650 hPa) over the whole period of simulation (January 2010–April 2011). In the same simulation, the correlation between δ18Ovin Niamey grid point and adjacent grid points is higher than 0.6 all over the year. The LMDZ-iso simulation thus suggests that the very local surface near observation might be useful as a proxy for boundary layer processes.

Figure 1.

Temporal evolution from June 2010 to May 2011 of (a) Specific humidity q (g/kg) along with event-based rainfall amount P (mm). From 23 April to 26 October 2010, 66 rain events were collected (total amount of 523.8 mm, precipitation varied from 0.1 mm to 40.3 mm); (b)δ18Ov(‰): black, red and blue lines are hourly averages, 24-h and 15-days running averages respectively. Intense power cuts (from 15 to 28 July 2010) and maintenance period (from 5 to 18 Nov 2010), where a new automated calibration system was set up, are responsible for the 2 major gaps. Event-based (δ18Op − 10) (‰) are in green dots; and (c) δ18Ov(‰) in our observations (red line) and in LMDZ-iso (green dotted line).

3. Is Water Vapor in Isotopic Equilibrium With Precipitation?

[6] In this section, we explore to what extent δ18Ov and δ18Op are far from the theoretically isotopic equilibrium (δ18Oveq) (please see legend of Figure S1 in the auxiliary material for the δ18Oveq calculation from δ18Op). We find that δ18Ov-δ18Oveq is mostly negative, specifically for the 2 events in April where δ18Ov-δ18Oveq is of −6.4 and −4.9‰ suggesting high evaporative effects. The robust feature (mostly independent of condensation temperature) is the positive linear evolution of δ18Ov-δ18Oveq as a function of relative humidity (r2 = 0.54) showing that as expected, isotopic exchanges are closer to the isotopic equilibrium for high relative humidity (see Figure S1).

4. The δ18Ov Modes of Variability

4.1. A W-Shape Seasonalδ18Ov Variation

[7] As expected, q shows a seasonal cycle with the highest (lowest) values (15–20 g/kg) (<10 g/kg) occurring from July to September 2010 (from November 2010 to March 2011) (Figure 1a). By contrast, δ18Ovexhibits a distinct W-shape seasonal variability (Figure 1b) with a seasonal amplitude of around 6‰. A first δ18Ov minimum is observed within the monsoon season, from August to September (mAug–Sept = −15.45 ± 1.78‰) and captures two minimum values in mid-August and mid-September while a second minimum is exhibited within the dry season, in January (mJanuary = −15.45 ± 1.94‰). The δ18Ov depletion during the summer monsoon reflects the increasing convective activity at the regional scale as expected from the “amount effect” and its consequence on the subsequent vapor [Dansgaard, 1964]. Interestingly, this first isotopic minimum is not observed by SCIAMACHY near-surfaceδDvspace-borne measurements [Frankenberg et al., 2009]. The second isotopic depletion (observed by SCIAMACHY near-surfaceδDvalthough more depleted) can be related to large-scale subsidence originating from the descending Hadley cell, stronger from January to March and transporting depleted air masses from high altitude down to the surface as suggested byFrankenberg et al. [2009].

[8] The monsoon retreat (October to November 2010) is characterized by an enrichment of the vapor as the convective activity decreases. Averaged δ18Ov over Oct.–Nov. is −13.39 ± 1.34‰. The moistening period (April–May) is characterized by a progressive increase of δ18Ov, which could reflect change in moisture source: the monsoon flow propagates northward and the water vapor from the Equatorial band is more enriched than at the latitude of Niamey [Frankenberg et al., 2009, Figure 1a]. Annual δ18Ov maximum is reached right before the monsoon onset (mMay = −9.59 ± 1.28‰; from 1 to 12 May).

4.2. Intra-seasonal Timescales

[9] To determine the dominant intra-seasonal modes ofδ18Ov variability, we performed a wavelet analysis [Torrence and Compo, 1998] on the hourly interpolated δ18Ov data (see Figure S2).

[10] During summer monsoon, a significant signal is seen in the 25–40-day band. Other significant modes (at the 90% confidence level) are detected in the 15–25-day band (only at the beginning of the monsoon in July), and in the 6–12-day band at the beginning (1 to 10 July) and at the end of the monsoon (20 September to 10 October). Fluctuations in the 1–5-day band are also recorded, but with a weaker amplitude. These modes are known to modulate convection in West Africa [Janicot et al., 2010] and may reflect the property of δ18Ov to integrate convective activity (the stronger the convection, the lower the δ18Ov) both in time and space. Indeed, the correlation between daily δ18Ovtime-series and OLR (Outgoing Longwave Radiation, from the National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites [Liebmann and Smith, 1996]) averaged over the 9 previous days from July to September 2010 is maximum (r > 0.6) Southwest and East of Niamey, suggesting a strong influence of integrated convective activity along the southerly monsoon flow and the westwards propagation of convective systems on δ18Ov (see Figure S3).

[11] From December to March, δ18Ov shows strong and significant periodicities at synoptic scales, with a major periodicity around 5–9 days, for example around 16 January (Figure 2a). Figure 2bshows a sequence from 13 to 17 January of the 500 hPa vertical velocity (to inspect the mid-tropospheric subsidence), the 300 hPa stream function (upper tropospheric circulation) and surface winds anomalies (relative to the 3 to 20 January based period). We observe a well-developed baroclinic wave train-like structure consisting of alternating positive and negative vertical velocity anomaly centers following an arch trajectory from the mid-latitude North Atlantic to North Africa across southern Mideast Africa. Prior to theδ18Ov minimum, subsidence, associated with anticyclonic circulation, is maximum north of Niamey, and southwards surface wind speed increases. As the wave train propagates eastward, δ18Ov decreases. We also observe a decrease in surface temperature (which exceeds −3°C) that could be related to the southward moisture transport (positive vertical velocity) of cold air from higher latitude at around 30°N (see Figure S4). On 17 January, δ18Ov increases as subsidence weakens (the vertical velocity anomaly becomes mostly negative over North Africa at the Niamey longitude). Our observations suggest that surface δ18Ovin Niamey is strongly modulated at synoptic time-scale by tropical-extra-tropical teleconnections.

Figure 2.

(a) Temporal evolution from 3 to 20 January 2011 (horizontal dotted lines are at 00 UTC) of δ18Ov (‰). (b) Sequence from 13 to 17 January of daily anomalies (relative to the 3 to 20 January base period) in vertical velocity at 500 hPa (shaded, Pa s−1), 300 hPa stream function (contour, interval is 3.106 m2.s−1), and 925 hPa wind (red vectors, a 10 m s−1vector wind is indicated at the bottom right of the figure. Positive (red) vertical velocity indicate region of subsidence, positive stream function (solid contour) indicate upper-tropospheric anticyclonic circulation. Niamey is located at 13.31°N, 2.06°E (black dot). Black arrows link the time-series in Figure 2a and the maps in Figure 2b referring to the same days. Green arrows indicate the eastward displacement of the vertical velocity anomalies.

4.3. Variability at the Convective Event Scale

[12] During the monsoon season, no systematic diurnal cycle can be seen: the strongest δ18Ovvariations occur with precipitation events and no diurnal cycle is seen when accounting for non-rainy days only. However, robust features are seen considering the rainy days. In the following, we only consider rain events where P > 5 mm. 16% are associated with an increase inδ18Ov from the beginning (sometimes before) to the end of the rain event, with a maximum increase of +2.9‰ observed on 17 August 2010. These variations could correspond to strong rain evaporation events and weak subsidence in the unsaturated downdraft (see discussion below). On the contrary, for 84% of rain events, δ18Ov exhibits a sharp decrease (from ∼−2 to −5‰), most of the time at the beginning of the rain events (Figure 3b), in phase with a surface temperature drop (∼−3 to −10°C). However, δ18Ov and temperature drops are observed more than 30 min before rain onset in 19% of the depleting events and up to 2 hours before for less that 1% of all events. This δ18Ovfeature (including the possible delay) could reflect the propagation of density currents (cold pools) initiated by rain evaporation in unsaturated downdrafts, and propagating at the bottom of the system through a rear-to-front dry flow induced by meso-scale subsidence in the stratiform zone of the squall line [Houze et al., 1989]. Indeed, the strong droplet reevaporation in the stratiform zone induces an isotopic decrease (increase) of δ18Ov, provided that evaporation rate is low (high) and that relative humidity is high (low) [Stewart, 1975; Bony et al., 2008]. Thus, along a squall line, surface δ18Ov may essentially result from the relative contribution of the dynamics, which bring depleted water vapor from higher altitude by previous condensation process, and the rain evaporation, which tend to enrich or to deplete the water vapor, depending on the relative humidity of the environment and the fraction of remaining droplets in the downdraft [Risi et al., 2010b]. We clearly show in Figure 3c that the δ18Ov evolution in the course of precipitation event, and also before and after it, is not explained by a basic Rayleigh distillation. We may explain the small increase in q at the end of the precipitation event by reevaporation of droplets leading to remoistening the environment.

Figure 3.

(a) Evolution of 5-min average q from 15 to 19 August 2011. (b) Same as Figure 3a but forδ18Ov. Blue dotted vertical lines correspond to the passage of meso-scale convective systems. The evolution ofδ18Ov for these events is representative of 84% of the sampled convective systems. (c) Relationship between δ18Οv and q during 15 August 2011 (10UTC to 18UTC) where a meso-scale convective system brought 10.8 mm precipitation amount between 1130UTC and 1430UTC (blue dots). Cyan solid line accounts for a Rayleigh distillation under isotopic equilibrium conditions. We assume a condensation temperature of 0°C and an initial q (δ18Ov) of 19 g/kg (−13.8‰). (d) Same as Figure 3a but for 26 to 29 April 2011. (e) Same as Figure3b but for 26 to 29 April 2011. (f) Relationships between 5-min averageδ18Ovand 1/q over 12-h periods (am and pm) from 26 to 29 April 2011.

4.4. Diurnal Cycle During the Dry Season

[13] Outside of the monsoon season, a strong, repeatable diurnal signal is detected from November 2010 to May 2011. Specific humidity and δ18Ov vary in phase (decreasing around 06–08 UTC to 12–14 UTC and increasing during the night) (Figures 3d and 3e). Maximum amplitudes are 10‰ (15 g/kg) peak to peak for δ18Ov (q) and occur in April–May. A strong linear relationship between δ18Ovand 1/q at a 12-hour scale is exhibited over this period (seeFigure 3f) and reflects a mixing between two sources [Noone et al., 2011], which are potentially the water vapor advected within the boundary layer during nighttime and the drier and more depleted water vapor from the free atmosphere entrained at the top of the boundary layer during daytime. Between 26 and 29 April 2011, where diurnal amplitudes are maximum, δ18Ov of the boundary layer (upper tropospheric) source can be deduced from Figure 3f to be −8.62 ± 0.66‰ (−17.15 ± 0.79‰), n = 6. Thus, outside of the monsoon, δ18Ov combined with q may be used to quantify the relative contribution of horizontal transport and convective mixing.

5. Comparison of Our Observations With LMDZ-iso

[14] Comparing our δ18Ovobservations to the nudged LMDZ-iso simulation [Risi et al., 2010c] over 2010–2011 in the Niamey grid point, the model fails to reproduce the W-shape seasonal cycle and its amplitude (Figure 1c). The first isotopic depletion of low-level vapor observed from July to September 2010 is not as strong as in our observations (δ18Ov is overestimated by 2.2‰) whereas on the opposite the model underestimates δ18Ovduring the dry season (by 4.6‰ in January). The model does not capture the intra-seasonal variability ofδ18Ov during the monsoon season (σobs = 1.76‰ and σLMDZ = 0.66‰, calculated from daily averaged) (Figure 1c). However, from December 2010 to May 2011, LMDZ-iso overestimatesδ18Ov variability (σobs = 2.16‰ and σLMDZ = 3.20‰). The LMDZ-iso hourly outputs are not available and prevent us from discussing the squall line scale.

6. Conclusion and Perspectives

[15] At the seasonal scale, δ18Ovexhibits a “W-shape”, different from the seasonality of the water vapor concentration. Our observations confirm the role of convective activity in depleting the low-level water vapor during the monsoon, and the depleting effect of large-scale subsidence north of 10°N during the dry season. At the intraseasonal scale, summerδ18Ov records main modes likely associated to convection. Winter δ18Ovis modulated by tropical/extra-tropical teleconnection through the propagation of a baroclinic wave train-like structure and intrusion of air originating from higher latitude. Winter diurnal cycle ofδ18Ov and q reflect boundary layer mixing process between the lower and upper atmosphere. During the monsoon season, the strongest diurnal variations are associated with rain events and may record the propagation of density currents. We show the potential of simultaneous q and δ18Ovmeasurements to investigate the initiation of convection and to quantify convective processes such as meso-scale subsidence and rain evaporation within convective systems. Additional constraints will be brought by deuterium excess (d = δD − 8δ18O) on this aspect. Finally, we show that our observations are poorly reproduced by LMDZ-iso model and think that such in-situ time series can be very helpful for validating space-based remote sensing observations.


[16] We warmly thank J. R. Lawrence and M. Schneider for their very constructive and encouraging reviews. We thank G. Brissebrat, L. Fleury for their guidance to collect on-site meteorological data as well as J-P. Lafore, F. Guichard, F. Couvreux, D. Bouniol and R. Roehrig for fruitful discussions. We also thank V. Masson-Delmotte for her fruitful comments and Tahirou Bana Hachimou for on-site technical help. This work was funded by IRD and the INSU-LEFE-EVE YOUPI proposal.

[17] The Editor thanks James Lawrence and an anonymous reviewer for assisting with the evaluation of this paper.