Our petrological knowledge of the upper mantle composition comes from several sources:
(1) Nodules brought up in volcanic pipes;
(2) Large sections of mantle found in obducted ophiolites;
(3) Slices of mantle thrust up in mountain belts such as the Alps (Ivrea-Verbano Zone)
and the Caledonides (especially Norway); and
(4) Modelling "backwards" from erupted basalt compositions.
(1) Olivine rich nodules are quite common in erupted alkali basalts worldwide, and are almost all of spinel lherzolite composition (olivine, orthopyroxene, Cr-diopside, spinel). In kimberlite (diamond) pipes there is a greater diversity in that both garnet peridotite and eclogite xenoliths occur (the former dominant). Some nodules contain diopside-rich, phlogopite-rich or amphibole-rich veins. It is thought that these nodules are representative of the sub- continental mantle. This material seems to be rather refractory (could not yield much basalt on melting), but at the same time can be quite enriched in incompatible trace elements such as Sr, Ba, K, Rb and the light rare- earths.
(2) The peridotitic material in ophiolite complexes (obducted ocean floor) is mainly HARZBURGITE (ol+opyx) or DUNITE (ol), often cut by pyroxenite (opyx) veins and sometimes having chromite segregations (podiform chromites). There is a consensus that this mantle is the refractory residue left after basalt extraction at mid-ocean ridges. But are the pyroxenite veins the result of silica-rich fluids coming off the subduction zone? (see in Fig. 5a all the free quartz present in eclogite in subduction zones) - note that many ophiolites are thought to be fragments of "back-arc" spreading centres.
(3) In the Ivrea Zone and the Lanzo-Seisia Zone of the Italian Alps, the peridotite slices are overlain by layered gabbros. They are mainly LHERZOLITE (ol-opyx-cpyx -spinel, or -hornblende or -phlogopite). Most carry veins composed of orthopyroxene, orthopyroxene-spinel, diopside-orthopyroxene, phlogopite-diopside, or hornblende. Different segments different veins, suggesting a complex make-up of the sub-continental mantle. Some of the rock types are similar to those brought up in volcanic breccia pipes.
(4) Considerable progress has been made in understanding the compositions of basaltic rocks in recent years, and interpreting them in terms of melting models. So it is possible to "model back" to the primary mantle from which the basalt was derived, and estimate its composition. We now know that there are several distinct types of mantle, that have been kept separated for many hundreds of millions of years. They have distinct trace element and isotopic characteristics. (You may come across references to them as DMM, HIMU, EM1, EM2 and PREMA), but there is uncertainty as to where they are located.