Moisture budget analysis of SST-driven decadal Sahel precipitation variability in the twentieth century

Abstract

It is well known that the Sahel region of Africa is impacted by decadal scale variability in precipitation, driven by global sea surface temperatures. This work demonstrates that the National Center for Atmospheric Research’s Community Atmosphere Model, version 4 is capable of reproducing relationships between Sahelian precipitation variability and Indian and Atlantic Ocean sea surface temperature variations on such timescales. Further analysis then constructs a moisture budget breakdown using model output and shows that the change in precipitation minus evaporation in the region is dominated by column integrated moisture convergence due to the mean flow, with the convergence of mass in the atmospheric column mainly responsible. It is concluded that the oceanic forcing of atmospheric mass convergence and divergence to a first order explains the moisture balance patterns in the region. In particular, the anomalous circulation patterns, including net moisture divergence by the mean and transient flows combined with negative moisture advection, together explain the drying of the Sahel during the second half of the twentieth century. Diagnosis of moisture budget and circulation components within the main rainbelt and along the monsoon margins show that changes to the mass convergence are related to the magnitude of precipitation that falls in the region, while the advection of dry air is associated with the maximum latitudinal extent of precipitation.

Appendix

Figure 3 from the manuscript clearly shows the overestimation of rainfall variability in the NCEP/NCAR Reanalysis when compared to the observations or CAM4 modeled precipitation. It is further shown here that this discrepancy is due to the overestimation of the variance in the precipitation field prior to 1968 (Fig. 13a), and that it is much more realistic afterwards (Fig. 13b).

Fig. 13

JAS precipitation root mean square anomaly (mm/day) in a NCEP/NCAR 1949–1967 and b 1968–2008

Given the documented discrepancies in the variability of the NCEP/NCAR Reanalysis data (Poccard et al. 2000; Kintner et al. 2004), an alternative method to utilize the moisture budget of this dataset would be to compute the precipitation as a summation of the column integrated moisture flux and the evaporation, as explained in the text. However, a quick glance at Fig. 14 clearly shows that virtually all moisture budget variables through time (1901–2008, unfiltered) seem to show the same discrepancies as the previous studies have indicated. These include surface evaporation (in W/m²), total column precipitable water (in kq/m²), specific humidity (at 925, 600, and 300 mb and unitless), and zonal and meridional wind velocity (at 925, 600, and 300 mb in m/s). In virtually all cases, not just limited to precipitation and wind fields as documented by previous works, but also with evaporation, specific humidity, and total column precipitable water, the troubling drop off in magnitude and variability is apparent (note the differences in scale among the various panels). All variables reveal similar patterns to the NCEP/NCAR Reanalysis generated spurious precipitation field (Fig. 14a). There is a huge dropoff in the magnitude of all the study variables around 1968, with concurrent decreases in all moisture budget terms’ variability. This brief exploratory analysis, along with the results of previous works, leads us to conclude that studying regional variations in Sahel rainfall previous to 1968 cannot be done accurately using the NCEP/NCAR Reanalysis fields, even if the precipitation is computed as a residual of other variables’ due to their apparent shortcomings as well. This finding further encourages us to use the SST-driven model to understand the twentieth century moisture budget and its variability.

Fig. 14

NCEP/NCAR Reanalysis variables averaged over the Sahel domain for JAS 1901–2008. a Precipitation, including UEA observations (red line), b evaporation, c precipitable water, d–f specific humidity at 925, 600, and 300 mb, g–i zonal wind at 925, 600, and 150 mb, and j–l meridional wind at 925, 600, and 150 mb. Scales differ among all plots