The extent and consequences of global land-cover and land-use change are in
creasingly apparent. One consequence not so apparent is the altered structu
re of plants belowground. This paper examines such belowground changes, emp
hasizing the interaction of altered root distributions with other factors a
nd their treatment in models. Shifts of woody and herbaceous vegetation wit
h deforestation, afforestation, and woody plant encroachment typically alte
r the depth and distribution of plant rests, influencing soil nutrients, th
e water balance, and net primary productivity (NPP). For example, our analy
sis of global soil data sets shows that the major plant nutrients C, N, P,
and K are more shallowly distributed than are Ca, Mg, and Na, but patterns
for each element vary with the dominant vegetation type. After controlling
for climate, soil C and N are distributed more deeply in arid shrublands th
an in arid grasslands, and subhumid forests have shallower nutrient distrib
utions than do subhumid grasslands. Consequently, changes in vegetation may
influence the distribution of soil carbon and nutrients over time (perhaps
decades to centuries). Shifts in the water balance are typically much more
rapid. Catchment studies indicate that the water yield decreases 25-40 mm
for each 10% increase in tree cover, and increases in transpiration of wate
r taken up by deep roots may account for as much as 50% of observed respons
es. Because models are increasingly important for predicting the consequenc
es of vegetation change, we discuss the treatment of belowground processes
and how different treatments affect model outputs. Whether models are param
eterized by biome or plant life form (or neither), use single or multiple s
oil layers, or include N and water limitation will all affect predicted out
comes. Acknowledging and understanding such differences should help constra
in predictions of vegetation change.