Climatic factors reflected by magnetic susceptibility of the Chinese loess-paleosol sequences affected both depositional sites and source areas. Such factors include the wetness and temperature of the soil that control the rate of formation and the preservation of pedogenic magnetite [ Maher and Thompson, 1992], as well as the wind velocity and aridity in the source regions for the dust that control variations in dust flux [e.g., An et al., 1991a; Zhang et al., 1994].
Several recent studies have attempted to extract records of precipitation over the past 130,000 years from the Chinese loess-paleosol sequences, with the underlying assumption that the present concentration of pedogenic magnetite/maghemite is directly related to precipitation [ An et al., 1991b; Banerjee et al., 1993; Heller et al., 1993; Maher et al., 1994b]. Maher et al. [1994b] calculated a function that describes the positive correlation between modern rainfall and magnetic susceptibility in nine modern, spatially distributed soil types. The function was used to express ancient rainfall as a difference with modern rainfall, by calculating differences between the magnetic susceptibility of a loess or paleosol layer and that of a much older loess layer (L9, about 800,000 years old), which showed little evidence of weathering. Among the results were estimates that during the last interglacial stage, currently dry western sites received as much as 60% more annual rainfall than at present and some currently humid sites in the east and south received as much as 30% more [ Maher et al., 1994b]. Glacial episodes were characterized by less rainfall than at present. Moreover, higher rainfall than at present is indicated across the loess plateau in the early part of the Holocene, about 9,000 years ago, consistent with evidence from studies of lake-level and pollen, and similar to an interpretation by An et al. [1991b]. Even a currently dry, westerly site near Baicaoyuan, west of the northwest-southeast trending mountain range Liupan-shan, was apparently wetter between 5,000 and 10,000 years ago, on the basis of the amount of SP grains and magnetic susceptibility [ Banerjee et al., 1993; Maher et al., 1994b]. Conclusions drawn from these studies included the following: (1) the current easterly summer monsoon in China came from a more southerly direction during this time to flow parallel to the mountain range [ Banerjee et al., 1993; see also An et al., 1993], or (2) the east Asian monsoon intensified during the early Holocene to overwhelm the rain shadow effect of the intervening mountain range [ Maher et al., 1994b].
The analysis by Heller et al. [1993] compared magnetic
susceptibility and 
Be isotopes, and it used a ``calibration''
of the present rainfall at a site to the pedogenic magnetic
susceptibility of the early Holocene paleosol (S0) at the site.
From this analysis, Heller et al. [1993] estimated that
precipitation was virtually absent during glacial periods, except
for two brief periods at about 25,000 and 55,000 years ago, when
precipitation was about double the average for the early
Holocene. At one southern site the rainfall estimate for the
last interglacial is less than current annual rainfall.
The studies cited above mark important attempts to derive paleoclimate information from terrestrial deposits. The approaches, however, are based on some untested assumptions, and the conclusions from these studies thus must be considered only as preliminary attempts to unravel a quantitative record of climate change in Chinese loess. As Maher et al. [1994b] pointed out, the major uncertainties in their analysis are introduced by (1) the untested accuracy of the modern rainfall-susceptibility function, in which the annual rainfalls at the calibration sites have a smaller range than that of the estimates for past rainfall, (2) the assumption that the modern Chinese soils have matured to a magnetic steady state over a few thousand years and thus also that the ancient soils and loess had also achieved maximum enhancement of susceptibility, and (3) the assumption that the original susceptibility of each parent loess was close to that of L9. Along these lines, the evidence for changes in precipitation during the Holocene cloud the results of Heller et al. [1993] that depend on a correlation between modern rainfall and the Holocene paleosol [ Maher et al., 1994b].