Climate Change 2001:
Working Group I: The Scientific Basis
Other reports in this collection

11.2.5 Surface and Ground Water Storage and Permafrost

An important contribution to present day sea level rise could result from changes in the amount of water stored in the ground, on the surface in lakes and reservoirs, and by modifications to surface characteristics affecting runoff or evapotranspiration rates. Changing practices in the use of land and water could make these terms larger in future. However, very little quantitative information is available. For some of the components of the terrestrial water budget, Gornitz et al. (1997), updated by Gornitz (2000), give net results which differ substantially from those of Sahagian (2000) and Vörösmarty and Sahagian (2000), and also from those of Sahagian et al. (1994) used by Warrick et al. (1996). The largest positive contribution to sea level probably comes from ground water mining, which means the extraction of ground water from storage in aquifers in excess of the rate of natural recharge. Gornitz et al. (1997) estimate that ground water is mined at a rate that has been increasing in time, currently equivalent to 0.2 to 1.0 mm/yr of sea level, but they assume that much of this infiltrates back into aquifers so the contribution to sea level rise is only 0.1 to 0.4 mm/yr. Sahagian (2000) considers fewer aquifers; consequently he obtains a smaller total of 0.17 mm/yr from mining, but assumes that all of this water eventually reaches the ocean through the atmosphere or runoff. If Sahagian's assumption were applied to the inventory of Gornitz et al. it would imply a sea level contribution of 0.2 to 1.0 mm/yr.

Volumes of many of the world's large lakes have been reduced in recent decades through increased irrigation and other water use. Sahagian et al. (1994) and Sahagian (2000) estimate that the reduced volumes of the Caspian and Aral Seas (and associated ground water) contribute 0.03 and 0.18 mm/year to sea level rise, on the assumption that the extracted water reaches the world ocean by evapotranspiration. Recent in situ records and satellite altimetry data indicate that substantial fluctuations in the level of the Caspian Sea can occur on decadal time-scales (Cazenave et al., 1997) which suggests that short records may not give a good indication of the long-term average. The reduction of lake volumes in China may contribute a further 0.005 mm/yr (Shi and Zhou, 1992). Assuming there are no other large sources, we take 0.2 mm/yr as the upper limit of the present contribution to sea level from lakes. Gornitz et al. (1997) do not include a term from lake volume changes, because they assume the water extracted for irrigation largely enters the ground water rather than the world ocean, so we take zero as the lower limit.

Figure 11.6:
The raised 125,000 year old coral terraces of the Huon Peninsula of Papua New Guinea up to 400 m above present sea level (Chappell et al., 1996b).

Gornitz et al. (1997) estimate there is 13.6 mm of sea level equivalent impounded in reservoirs. Most of this capacity was created, at roughly a constant rate, from 1950 to 1990. This rate of storage represents a reduction in sea level of 0.34 mm/yr. They assume that annually 5 ± 0.5% of the water impounded seeps into deep aquifers, giving a 1990 rate of seepage of 0.61 to 0.75 mm/yr, and a total volume of 15 mm sea level equivalent. We consider that this represents an upper bound, because it is likely that the rate of seepage from any reservoir will decrease with time as the surrounding water table rises, as assumed by Sahagian (2000). On the basis of a typical porosity and area affected, he estimates that the volume trapped as ground water surrounding reservoirs is 1.2 times the volume impounded in reservoirs. His estimate of the storage in reservoirs is 14 to 28 mm sea level equivalent; hence the ground water storage is an additional 17 to 34 mm sea level equivalent. Lack of global inventories means that these estimates of storage may well be too small because of the many small reservoirs not taken into account (rice paddies, ponds, etc., provided they impound water above the water table) (Vörösmarty and Sahagian, 2000). The total stored could be up to 50% larger (Sahagian, 2000).

Gornitz et al. (1997) estimate that evapotranspiration of water from irrigated land leads to an increase in atmospheric water content and hence a fall in sea level of 0.14 to 0.15 mm/yr. We consider this to be an overestimate, because it implies a 20th century increase in global tropospheric water content which substantially exceeds observations (Section They further suggest that irrigation water derived from surface sources may infiltrate into aquifers, removing 0.40 to 0.48 mm/yr of sea level equivalent, based on the same assumption as for seepage from reservoirs. Urbanisation leads to reduced infiltration and increased surface runoff, which Gornitz et al. (1997) estimate may contribute 0.35 to 0.41 mm/yr of sea level rise. We consider these two terms to be upper bounds because, as with infiltration from reservoirs, a new steady state will be achieved after a period of years, with no further change in storage.

Estimates of the water contributed by deforestation are 0.1 mm/yr (Gornitz et al., 1997) and 0.14 mm/yr (Sahagian, 2000) of sea level rise. Water released by oxidation of fossil fuels, vegetation and wetlands loss is negligible (Gornitz et al., 1997).

Gornitz et al. (1997) estimate the total contribution to the 1990 rate of sea level change as -1.2 to -0.5 mm/yr. Integrating their estimates over 1910 to 1990 gives between -32 and -11 mm of sea level rise. In contrast, the estimate of Vörösmarty and Sahagian (2000) for the rate of sea level change from terrestrial storage is 0.06 mm/yr, equivalent to 5.4 mm over 80 years. The estimate of Sahagian et al. (1994), quoted by Warrick et al. (1996), was 12 mm during the 20th century. These discrepancies emphasise again the unsatisfactory knowledge of these contributions to sea level change.

Table 11.8 shows the ranges we have adopted for the various terms, based on the foregoing discussion. We integrate these terms over 1910 to 1990. (We use the time profiles of Gornitz et al. (1997) except that the infiltration from reservoirs is based on the approach of Sahagian (2000), and the rate of withdrawal from lakes is assumed constant over the last five decades.) This gives a range of -83 to +30 mm of sea level equivalent, or -1.1 to +0.4 mm/yr averaged over the period. However note that the rate of each of the terms increases during the 20th century.

This discussion suggests three important conclusions: (i) the effect of changes in terrestrial water storage on sea level may be considerable; (ii) the net effect on sea level could be of either sign, and (iii) the rate has increased over the last few decades (in the assessment of Gornitz et al. (1997) from near zero at the start of the century to -0.8 mm/yr in 1990).

Estimates of ice volume in northern hemisphere permafrost range from 1.1 to 3.7x1013 m3 (Zhang et al., 1999), equivalent to 0.03 to 0.10 m of global-average sea level. It occupies 25% of land area in the northern hemisphere. The major effects of global warming in presently unglaciated cold regions will be changes in the area of permafrost and a thickening of the active layer (the layer of seasonally thawed ground above permafrost). Both of these factors result in conversion of ground ice to liquid water, and hence in principle could contribute to the sea level change. Anisimov and Nelson (1997) estimated that a 10 to 20% reduction of area could occur by 2050 under a moderate climate-change scenario. In the absence of information about the vertical distribution of the ice, we make the assumption that the volume change is proportional to the area change. By 2100, the upper limit for the conversion of permafrost to soil water is thus about 50% of the total, or 50 mm sea level equivalent.

A thickening active layer will result in additional water storage capacity in the soil and thawing of ground ice will not necessarily make water available for runoff. What water is released could be mainly captured in ponds, thermokarst lakes, and marshes, rather than running off. Since the soil moisture in permafrost regions in the warm period is already very high, evaporation would not necessarily increase. We know of no quantitative estimates for these storage terms. We assume that the fraction which runs off lies within 0 and 50% of the available water. Hence we estimate the contribution of permafrost to sea level 1990 to 2100 as 0 to 25 mm (0 to 0.23 mm/yr). For the 20th century, during which the temperature change has been about five times less than assumed by Anisimov and Nelson for the next hundred years, our estimate is 0 to 5 mm (0 to 0.05 mm/yr).

Other reports in this collection