A gravity profile across the Kenya Rift at the equator was modeled with recent results of petrologic and seismic investigations as constraints. This profile is dominated by a 600‐km‐wide gravity low of 100 × 10−5 m/s2 (100 mGal). The model incorporates the important concept of modification of the lower crust by magmatic processes. Alkali basaltic magmas are believed to have modified the lower crust in southern Kenya during early rift‐related basaltic volcanism (23–14 Ma) on the basis of (1) synchronous voluminous eruption of alkali basalt in northern Kenya; (2) Kenya Rift International Seismic Project 1990 (KRISP 90) seismic lines which delineate an ∼10‐km‐thick layer of basal lower crust in southern Kenya which is absent or very thin to the north; (3) results from high pressure/temperature experiments on a late Miocene Plateau phonolite and related geochemical modeling which indicate a lower crustal origin for these voluminous lavas (14–11 Ma) by fusion of alkali basaltic material; and (4) Plateau phonolite distributions that are conspicuously limited to the southern Kenya Rift above this anomalous lower crust. The gravity model features a lens of mafic intrusives (3000 kg/m3; constrained by petrologic arguments) at the base of the crust. It extends from 25 to 34 km depth beneath the rift valley and pinches out at a distance of about 250 km on each side of the rift. Surrounding lower crust (2850 kg/m3) is bounded by upper crust (2700 kg/m3) at about 15 km depth. The gravitational effect of the positive density contrast in the lower crust due to the lens of mafic intrusives is offset by an underlying wedge of anomalously low density mantle (3150 kg/m3). This wedge is about 300 km wide at the Moho and is relatively steep sided, in agreement with KRISP teleseismic results. East of the rift, this anomalous mantle is bounded by normal upper mantle (3260 kg/m3) extending to 100 km depth. West of the rift, normal upper mantle extends to 90 km depth. Within the rift valley, shallow horst and graben are indicated by the gravity data.

Near‐liquidus melting relations have been determined for a mafic, plateau‐type, flood phonolite from the Kenya rift at 0.5, 0.7, 0.9, and 1.2 GPa, with H2O added through saturation, and at 0.7 GPa with H2O and CO2 added. Mixed‐volatile experiments at 0.7 GPa delineate a near‐liquidus multiple saturation of augite, andesine, phlogopite, oxides, and apatite at 1000°C, XCo2 = 0.42, with calcic amphibole melting above 975°C. The multiple saturation and phase assemblage are interpreted to indicate that plateau phonolites were in equilibrium with the residuum of a parental alkali basaltic composition at 0.7 GPa consisting of augite, andesine, titanomagnetite, and olivine (a product of incongruent melting of phlogopite and, possibly, amphibole). This evidence for lower crustal equilibration refutes suggestions that plateau phonolites are low‐pressure differentiates. Their enormous volumes (about 50,000 km3), restricted eruptive period (14–11 Ma), uniform major element compositions, and the paucity of associated mafic‐intermediate rocks also argue against a deep origin by fractional crystallization. A two‐stage process for the origin of the phonolites is consistent with the thermal evolution of the rift. The lower crust was pervasively injected by alkali basaltic magmas during the period of voluminous eruption of early to middle Miocene basalts. Rising isotherms during rift evolution caused subsequent partial melting of this predominantly basaltic lower crust in the late Miocene, generating the plateau phonolites.

Geochemical investigations support the petrogenesis of Kenya rift plateau‐type flood phonolites (14–11 Ma) by partial melting of an alkali basaltic material at lower crustal pressures. High‐pressure/high‐temperature experiments on a natural plateau phonolite (Hay and Wendlandt, this issue) document multiple saturation of augite, andesine, titanomagnetite, and phlogopite at 0.7 GPa, 1000°C, XCO2 = 0.42, with amphibole appearing at 975°C. A least squares solution to major element modeling, involving subtraction of the compositions of near‐liquidus augite, andesine, titanomagnetite, and olivine (Fo67; hypothesized product of incongruent melting of hydrous phases) from a Kenya Miocene alkali basalt composition, indicates that plateau phonolites can be derived by 15 wt % fusion of this hypothetical parental material (∑R2 = 0.07). Alkali basaltic magmas may have injected and/or underplated the lower crust in southern Kenya during prior rift‐related basaltic volcanism (23–14 Ma). Bulk Earth values of (87Sr/86Sr)i and εNd near zero for four plateau phonolite samples are consistent with a mantle‐derived parental composition. Three of these four samples reflect little, if any, postmelting modification; one sample may have evolved by fractional crystallization (high Rb/Sr, low Ba, Sr and Mg #). A fifth sample may show evidence of assimilation and fractional crystallization processes (elevated radiogenic Sr and Pb, large negative Eu anomaly, and low Ba, Sr, and Mg #). Much of the geochemical variation among plateau phonolite lavas, however, can be ascribed to melting of a predominantly alkali basaltic source with contributions from a lower crustal protolith. A mantle‐derived source is also supported by Sr‐Nd‐Pb isotope data for the phonolites, which indicate that the alkali basaltic source can be described in terms of high U/Pb (HIMU) and enriched mantle (EM1 and EM2) components.