Skip to main content
PMC home page
Proceedings of the National Academy of Sciences of the United States of America logoLink to Proceedings of the National Academy of Sciences of the United States of America
. 2005 May 12;102(21):7612–7617. doi: 10.1073/pnas.0500208102

Human modification of global water vapor flows from the land surface

Line J Gordon *,†, Will Steffen ‡, Bror F Jönsson §, Carl Folke *, Malin Falkenmark , Åse Johannessen *
PMCID: PMC1140421  PMID: 15890780

Abstract

It is well documented that human modification of the hydrological cycle has profoundly affected the flow of liquid water across the Earth's land surface. Alteration of water vapor flows through land-use changes has received comparatively less attention, despite compelling evidence that such alteration can influence the functioning of the Earth System. We show that deforestation is as large a driving force as irrigation in terms of changes in the hydrological cycle. Deforestation has decreased global vapor flows from land by 4% (3,000 km3/yr), a decrease that is quantitatively as large as the increased vapor flow caused by irrigation (2,600 km3/yr). Although the net change in global vapor flows is close to zero, the spatial distributions of deforestation and irrigation are different, leading to major regional transformations of vapor-flow patterns. We analyze these changes in the light of future land-use-change projections that suggest widespread deforestation in sub-Saharan Africa and intensification of agricultural production in the Asian monsoon region. Furthermore, significant modification of vapor flows in the lands around the Indian Ocean basin will increase the risk for changes in the behavior of the Asian monsoon system. This analysis suggests that the need to increase food production in one region may affect the capability to increase food production in another. At the scale of the Earth as a whole, our results emphasize the need for climate models to take land-use change, in both land cover and irrigation, into account.

Keywords: deforestation, irrigation, land-use changes, climate change, evapotranspiration

Meeting water requirements for future food production without compromising the water needs of other human life-support systems is a major challenge for future freshwater management (1, 2). Food production involves the transformation of liquid water (from irrigation or precipitation) to water vapor (evapotranspiration), and recent estimations show that annual global vapor flows from future food production are likely to double within the next 25–50 yr (3, 4). It is well known that irrigation is, by far, the largest water user in terms of liquid water withdrawal from rivers and aquifers (5, 6) and that human modification of the hydrological cycle has profoundly affected the flow of liquid water across the Earth's land surface (7–10). Alteration of water vapor flows through land-use and land-cover changes has received less attention, compared to liquid flows, although there is compelling evidence that such alterations can influence the functioning of the Earth System (11–15).

Several examples from around the world show that changes in vapor flows due to land-cover change and irrigation have affected the functioning of the terrestrial biosphere and influenced climate at local and regional scales. For example, irrigated croplands have increased vapor flows from a converted steppe in Colorado by 120% (16), contributing to higher precipitation, lower temperature, and an increase in thunderstorm activity (17, 18). Large-scale deforestation for crop- and grasslands in Australia has caused a 10% decrease in vapor flows on that continent, with consequent widespread and irreversible dryland salinity that has reduced crop productivity in some regions and made farming impossible in others (13). Simulations of changed vapor flows due to land-cover change in East Asia have affected the behavior of the East Asian Monsoon, including a weakening of the summer monsoon low-pressure system and an increase in irregular northerly flow (12, 19). Modeled vegetation changes for agricultural expansion in West Africa have shown potentially dramatic impacts on rainfall in the African monsoon circulation, especially through deforestation along the southern coast, because this deforestation decreases the vapor flow into the region (20).

These examples, however, address only local to regional changes. No spatially explicit analysis exists of the extent of vapor-flow changes at a global scale. Here, we present a spatially explicit global estimate of changes in vapor flows as a consequence of both irrigation and deforestation. We elucidate which regions have experienced substantial changes in vapor flows as a consequence of human actions and the scale and direction of those changes. We analyze these changes and address the challenges facing future freshwater management to increase vapor flows for food production without compromising the capacity of the Earth System to support human welfare.

Methods

By comparing vapor flows from potential vegetation with flows from the actual vegetation of today, we quantify the alteration of global vapor flows due to deforestation and irrigation. Potential vegetation is here defined as the vegetation most likely to have been present in the absence of human activities, assuming a climate similar to the present one as the main determinant of biome distribution. We focused on deforestation because it has, in general, a higher impact on vapor flows than do other land-cover changes (21).

We developed a geographic information system (GIS) model of the terrestrial surface of the Earth with a spatial resolution of 0.5° × 0.5°. Each grid cell contains information on land cover, climate (temperature and precipitation), and percentage irrigated area. For land cover, the 5-min data set for potential vegetation (22) was regridded, by using interpolation, to 0.5° and the same land mask as the data set on actual vegetation (23). See Table 1 for the total area of the 17 vegetation types in the actual and potential vegetation datasets. Mean monthly precipitation and temperature were derived from a 30-yr (1961–1990) data set (24). The same climatic data were applied for estimating water vapor flows for both actual vegetation and potential vegetation. A digital global map of irrigated areas (25) was used to estimate the percentage of each grid that is equipped for irrigation.

Table 1. Data used for the estimation of annual vapor flows from actual and potential vegetation.

Area, 103 km2
Potential vegetation
Actual vegetation
kveg
Vegetation type Difference, % Summer Winter
Intensive cropland 0 15,872* 0.30 0.60
Marginal cropland used for grazing 0 0.15 0.30
Tropical evergreen forest/woodland 19,332 16,278 -16 0.50 0.50
Tropical deciduous forest/woodland 6,153 3,062 -50 0.30 0.60
Temperate broadleaf evergreen forest/woodland 1,452 356 -75 0.50 0.50
Temperate needleleaf evergreen forest/woodland 3,591 2,134 -41 0.30 0.60
Temperate deciduous forest/woodland 6,035 2,095 -65 0.30 0.60
Boreal evergreen forest/woodland 6,608 6,356 -4 0.30 0.60
Boreal deciduous forest/woodland 3,109 2,165 -30 0.30 0.60
Evergreen deciduous mixed forest/woodland 16,260 14,222 -13 0.30 0.60
Savanna† 19,562 19,562 0.25 0.40
Grassland steppe 14,393 14,393 0.15 0.30
Dense shrubland 6,901 6,901 0.10 0.20
Open shrubland 11,748 11,748 0.15 0.25
Tundra 8,321 8,321 0.02 0.95
Hot desert 15,596 15,596 0.02 0.02
Polar desert/rock/ice 3,605 3,605 0.95 0.95
Irrigated areas‡ 0.50 0.50
Total 142,657 142,667
Open in a new tab

The data include vegetation types, areas of different vegetation types, difference in area between actual and potential vegetation, and the monthly vegetation-specific coefficient (kveg).

*

Area includes both intensive croplands and marginal croplands used for grazing.

†

kveg of savannas was originally 0.15 for summer and winter months.

‡

Irrigated areas are defined as a percentage of each grid cell.

Vapor-Flow Estimates

The monthly vapor flow for each vegetation type was approximated as

graphic file with name M1.gif

where VF is the vapor flow, kveg is the vegetation-specific coefficient (see below), and PET is the potential evapotranspiration. The PET was estimated with a modified Thornthwaite equation (26):

graphic file with name M2.gif

This equation was chosen because it requires only one climatic input variable (Celsius temperature T) and because it has been suggested that Thornthwaite-based methods are as good as more refined methods when dealing with monthly data as compared with more detailed daily data (27).

A “cutoff” for precipitation was made, ensuring that the mean monthly vapor flow for each grid cell could not be larger than the mean monthly precipitation. To avoid the cutoff in irrigated lands, we estimated vapor flows from these lands separately:

graphic file with name M3.gif

where kveg is the vegetation coefficient of the vegetation within that cell, kveg irri is the vegetation coefficient for irrigated crops, and irri is the percentage of irrigated area within that cell.

The coefficient kveg is vegetation-specific and is related to parameters that influence water uptake and transpiration, such as depth of roots, leaf area, length of growing season, and soil moisture. The kveg have been developed for summer and winter values of 25 different vegetation types used in the terrain subprogram of Pennsylvania State University–National Center for Atmospheric Research (PSU/NCAR) Mesoscale Model (MM5) (28); see Table 1. Monthly kveg were constructed by combining the winter and summer values with weighted averaging. To account for the austral seasons, the summer and winter values of the Southern Hemisphere were switched at the equator. For “marginal cropland/areas used for grazing” we used the same value as for natural grasslands.

The kveg for savannas was recalibrated after a first run of the model. We performed a qualitative comparison between our estimate of vapor flows from actual land cover and an estimate of present vapor flows in the global-water-use and hydrology model watergap 2.1 developed by Döll and colleagues (29), which showed our data to be much lower from savannas than what was found in watergap 2.1. In the final run of the model, we assumed a kveg for savannas corresponding to the average value of kveg for grassland and forest (0.25 for summer and 0.45 for winter months) (Table 1). This assumption seems valid because tree cover determines the distribution of savannas in the vegetation classification, placing them geographically between grasslands (<10% tree cover) and forests (>80% tree cover).

Results

The total global vapor flows for potential vegetation was estimated at 67,000 km3/yr. Fig. 1 shows the distribution of vapor flows, clearly illustrating the major contribution of the humid tropical regions to the total global flow and underscoring the importance of these regions for this aspect of Earth System functioning. The highest vapor flows are found in the Southeast Asian archipelago, which, when coupled with the high evaporation rates from the adjacent warm surface waters of the western Pacific Ocean and eastern Indian Ocean, shows why this region is so important for the functioning of the Earth System.

Fig. 1.

Fig. 1.

Open in a new tab

Spatial distribution of annual water vapor flows (mm/yr) from potential vegetation, illustrating water vapor flows before human impacts. The figure illustrates the importance of vapor flows from the humid tropics. The total global vapor flow amounts to 67,000 km3/yr.

The areas where the conversion of forests to another land-cover type has led to significant changes in vapor flows are shown in Fig. 2. Aggregated globally, the net decrease of vapor flows due to deforestation is ≈3,000 km3/yr, which is 4% of the total global vapor flow from the land surface. However, the global total does not reveal significant regional differences. The main areas that have experienced decreases are Southeast Asia (particularly the Philippine Islands, Sumatra, and Borneo), southwestern and central Africa, the arc of deforested areas around the edge of the Amazon basin, northern South America, and the temperate forest regions of the eastern United States and central and western Europe (Fig. 2).

Fig. 2.

Fig. 2.

Open in a new tab

Spatial distribution of changes in vapor flows due to deforestation (mm/yr), based on the change in vapor flows between potential vegetation and actual vegetation in deforested areas. The total decrease in vapor flows is ≈3,000 km3/yr.

The increases in vapor flows owing to irrigation are shown in Fig. 3. Aggregated globally, the net increase through irrigation was found to be ≈2,600 km3/yr, which is consistent with previous global estimates that range from 1,800 to 2,500 km3/yr (5, 6, 30–31). The only estimate based on the same data set of irrigation as our study, however, is more conservative (1,100 km3/yr) (32). The authors of ref. 32 suggest a method of modeling global irrigation requirements that separates rice from other cereals and builds on length of growing periods for different crops and in different regions. This is a more sophisticated method than ours. However, because our result is consistent with other estimates, we believe it to be appropriate for this first-cut global estimation.

Fig. 3.

Fig. 3.

Open in a new tab

Spatial distribution of changes in vapor flows due to irrigation (mm/yr), defined as the change in vapor flows when irrigation only is added to actual vegetation. The total increase in vapor flows amounts to 2,600 km3/yr.

With irrigation, there is, again, a large regional variation, the most prominent regions of irrigation clearly correlated with areas of intensive food production: southeastern China and, especially, the Indo-Gangetic Plain of the Indian subcontinent. Significant vapor flows from irrigation also occur in central Asia, the central and western United States, southern Europe, and north-central Chile.

Discussion

Our results show that deforestation is as large a driving force as irrigation in the human alteration of global vapor flows. Thus, at the global aggregate level, there has not been a substantial net change in vapor flows. The decrease in vapor flows of 3,000 km3/yr from deforestation is almost compensated for by increased vapor flows from irrigation of 2,600 km3/yr. The combined effects of deforestation and irrigation have led to substantial redistribution of vapor flows at the global scale, as shown in Fig. 4.

Fig. 4.

Fig. 4.

Open in a new tab

Spatial distribution of net changes in vapor flows between potential vegetation and actual deforested and irrigated vegetation in mm/yr. The aggregated global change as compared with the potential vegetation is small (400 km3/yr), but the map illustrates the large spatial redistribution of water vapor flows from the land surface at the global scale.

Three different types of regional vapor-flow change can be seen: (i) regions with net loss of vapor flows from deforestation and no compensation from irrigation (e.g., around the Amazon Basin); (ii) regions with increased vapor flows due to irrigation (e.g., the western United States, where many of these irrigated regions are in semiarid zones and carried no dense forests originally); and (iii) regions where loss of vapor flow from deforestation is compensated for by increases from irrigation (e.g., northern China).

The total continental vapor flow for actual vegetation of 66,600 km3/yr corresponds well with previous global estimates of water vapor flow that range from 61,000 to 72,500 km3/yr (2, 33, 34). In Table 2, we compare our biome-specific vapor flow (mm/yr) with estimates from a review of field data on annual vapor flows in various biomes (2). The comparison reflects a high level of agreement between the field data and our estimate.

Table 2.

Comparison between the estimate of vapor flows from various biomes in the model of this article and field data from Rockström et al. (2)

Model data
Field data
Biome Vegetation type Vapor flow, mm/yr Vegetation type Vapor flow, mm/yr
Forests/woodlands
    Tropical Evergreen forest/woodland 1,146 Wet forest 1,245
Deciduous forest/woodland 795 Forest, dry/deciduous/seasonal 792
    Temperate Needleleaf evergreen forest/woodland 510 Predominantly coniferous forest 487
Broadleaf evergreen forest/woodland 592 Predominantly deciduous forest 729
Deciduous forest/woodland 553
    Boreal Evergreen forest/woodland 339 Taiga 401
Deciduous forest/woodland 293
Evergreen/deciduous mixed forest/woodland 313
Savanna* Savanna 556 Savanna/woodland, dry 882
Wet savanna 1,267
Woodland/woody savanna 416
Grasslands/shrublands Grassland/steppe 258 Cool grassland 410
Marginal cropland used for grazing 431
Dense shrubland 280 Dry shrubland 270
Open shrubland 196
Open in a new tab
*

Savanna in this estimate is based on both tropical and temperate savannas/woodlands, whereas the estimate by Rockström et al. (2) includes only tropical savannas.

The global pattern of vapor flows from the land surface to the atmosphere will change further over the next 50 yr to meet the food requirements of a population of eight to nine billion and to reduce today's malnourishment. The need to nearly double vapor flows associated with food production (3, 4) implies significant regional changes in vapor flows in several parts of the world, with likely global-scale impacts on the functioning of the Earth System.

Much interest has focused on the Amazon Basin. Currently, about 15% of the Brazilian Amazon has been deforested and converted to agricultural uses, with concomitant decreases in vapor flows (Fig. 2). Model projections suggest that there is a critical threshold beyond which the reduction in rainfall due to reduced vapor flows leads to dieback of the forest, thus forming a positive feedback loop that hastens the conversion of rainforest to savanna or grassland (35). Conversion of the Amazon rainforest would obviously have large impacts on regional vapor flows with possible global consequences for climate, but projections of future land-cover change in the Amazon Basin are not consistent, with some projections showing that much of the rainforest will still remain by the end of this century (36).

More likely and more immediate, however, are changes in the pattern of vapor flows around the Indian Ocean basin. South Asia and sub-Saharan Africa are two regions with an acute need to increase food production (37). Irrigation is widespread in South Asia, and vapor flows have already increased substantially (Fig. 3). Given the looming water scarcity in the region, future increases in food availability will likely come from increased production efficiency and from food trade (i.e., virtual water trade). In sub-Saharan Africa, on the other hand, increased food production will most likely come from further expansion of agricultural land through deforestation (36), causing a net decrease in vapor flows. Coupled with the substantial changes in vapor flows that have already occurred in Asia (Fig. 2), the deforestation of sub-Saharan Africa would create a ring of highly modified terrestrial vapor flows around the Indian Ocean basin, a key component of the Asian monsoon system. Given the apparent sensitivity of the Asian monsoon system to land-cover change (19), further modification of vapor flows in the lands around the Indian Ocean basin will increase the risk for changes in the behavior of the Asian monsoon system, raising the possibility that the need to increase food production in one region (sub-Saharan Africa) may affect the capability to increase food production in another (South Asia).

In terms of the functioning of Earth System as a whole, our results also raise important questions related to the relative role of land-use and climate change in altering vapor flows. For example, we know that the vapor flows from the oceans amount to ≈80–85% of the total vapor flow from the Earth's surface. It has been estimated that climatic changes in the last four decades have changed vapor flows over the oceans by 0.2–0.5 mm/day, with both increases and decreases depending on location and season (38). No consistent estimate exists of climatic-change effects on vapor flows from terrestrial areas. Some estimates indicate a decrease in pan evaporation in some regions despite a temperature increase over the past 50 yr (39), whereas others show increased evaporation from many land areas over the past 30 yr due to increased temperature (40). The relative magnitude of changes in water vapor flow due to land-use change compared with those due to climatic change thus remains unknown and needs to be explored in further work.

Our estimate illustrates the need to increase the understanding of direct human alteration of water vapor flows from the land surface on atmospheric circulation by including the impact of land-use change in global climate models. In particular, the results show that the effects of irrigation on water vapor flows are equally as important as deforestation in describing the climatic effects of human modification of the land surface.

Acknowledgments

We thank J. Rockström for several years of collaboration and conceptual development of research in this area; J. Barron, P. Fox, L. Deutsch, J. Norberg, and Ö. Bodin for comments on the manuscript; two anonymous reviewers for providing useful comments; and P. Döll for providing estimates of water vapor flows from a model run by WaterGAP (University of Kassel, Kassel, Germany). L.J.G.'s work was funded by the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS) and the Swedish International Development Cooperation Agency (SIDA).

Author contributions: L.J.G., W.S., and B.F.J. designed research; L.J.G., B.F.J., and A.J. performed research; B.F.J. and L.J.G. contributed new reagents/analytic tools; L.J.G., W.S., B.F.J., C.F., and M.F. analyzed data; and L.J.G., W.S., C.F., and M.F. wrote the paper.

This paper was submitted directly (Track II) to the PNAS office.

References

  • 1.Postel, S. L. (2003) Nat. Resour. Forum 27, 89–98. [Google Scholar]
  • 2.Rockström, J., Gordon, L., Folke, C., Falkenmark, M. & Engwall, M. L. (1999) Conserv. Ecol. 3. Available at www.consecol.org/vol3/iss2/art5.
  • 3.Postel, S. L. (1998) Bioscience 48, 629–637. [Google Scholar]
  • 4.Rockström, J. (2003) Philos. Trans. Roy. Soc. London B, 358 1997–2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Postel, S. L., Daily, G. C. & Ehrlich, P. R. (1996) Science 271, 785–788. [Google Scholar]
  • 6.Gleick, P. H. (1993) Water in Crisis: A Guide to the World's Fresh Water Resources (Oxford Univ. Press, New York).
  • 7.Dynesius, M. & Nilsson, C. (1994) Science 266, 753–762. [DOI] [PubMed] [Google Scholar]
  • 8.Vörösmarty, C. J., Sharma, K. P., Fekete, B. M., Copeland, A. H., Holden, J., Marble, J. & Lough, J. A. (1997) Ambio 26, 210–219. [Google Scholar]
  • 9.Rosenberg, D. M., McCully, P. & Pringle, C. (2000) Bioscience 50, 746–751. [Google Scholar]
  • 10.Smakhtin, V., Revenga, C. & Döll, P. (2004) Taking into Account Environmental Water Requirements in Global-Scale Water Resources Assessments (Comprehensive Assessment, Colombo, Sri Lanka), Comprehensive Assessment Research Report 2.
  • 11.Kabat, P., Claussen, M., Dirmeyer, P. A., Gash, J. H. C., Bravo de Guenni, L., Meybeck, M., Pielke, R. A., Sr., Vörösmarty, C. J., Hutjes, R. W. A. & Lütkemeier, S. (2004) Vegetation, Water, Humans and Climate: A New Perspective on an Interactive System (Springer, Berlin).
  • 12.Fu, C. (2003) Glob. Planet. Change 37, 219–229. [Google Scholar]
  • 13.Gordon, L., Dunlop, M. & Foran, B. (2003) Philos. Trans. Roy. Soc. London B 358, 1973–1984. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Nemani, R. R., Running, S. W., Pielke, R. A. & Chase, T. N. (1996) J. Geophys. Res. 101, 7157–7162. [Google Scholar]
  • 15.Marland, G., Pielke, R. A., Sr., Apps, M., Avissar, R., Betts, R. A., Davis, K. J., Frumhoff, P. C., Jackson, S. T., Joyce, L., Kauppi, P., et al. (2003) Climate Policy 3, 149–157. [Google Scholar]
  • 16.Baron, J. S., Hartman, M. D., Kittel, T. G. F., Band, L. E., Ojima, D. S. & Lammers, R. B. (1998) Ecol. Appl. 8, 1037–1051. [Google Scholar]
  • 17.Chase, T. N., Pielke, R. A., Kittel, T. G. F., Baron, J. S. & Stohlgren, T. J. (1999) J. Geophys. Res. Atmos. 104, 16673–16690. [Google Scholar]
  • 18.Pielke, R. A., Lee, T. J., Copeland, J. H., Eastman, J. L., Ziegler, C. L. & Finley, C. A. (1997) Ecol. Appl. 7, 3–21. [Google Scholar]
  • 19.Fu, C. & Yuan, H. (1994) Chin. Sci. Bull. 45, 1199–1203. [Google Scholar]
  • 20.Zheng, X. & Eltahir, E. A. B. (1998) J. Climate 11, 2078–2096. [Google Scholar]
  • 21.Bosch, J. M. & Hewlett, J. D. (1982) J. Hydrol. 55, 3–23. [Google Scholar]
  • 22.Ramankutty, N. & Foley, J. A. (1999) Glob. Biogeochem. Cycles 13, 997–1027. [Google Scholar]
  • 23.Goldewijk, K. K. (2001) Glob. Biogeochem. Cycles 15, 417–433. [Google Scholar]
  • 24.Willmott, C. J. & Matsuura, K. (1998) Global Air Temperature and Precipitation: Regridded Monthly and Annual Climatologies (Center for Climatic Research, University of Delaware, Newark), Version 2.01.
  • 25.Döll, P. & Siebert, S. (1999) A Digital Global Map of Irrigated Areas (Center for Environmental Systems Research, Kassel, Germany), Report A9901.
  • 26.Holland, H. D. (1978) The Chemistry of the Atmosphere and Oceans (Wiley, New York).
  • 27.Federer, C. A., Vörösmarty, C. & Fekete, B. (1996) Water Resour. Res. 32, 2315–2321. [Google Scholar]
  • 28.Dudhia, J., Gill, D., Guo, Y.-R., Hansen, D., Manning, K. & Wang, W. (1999) PSU/NCAR Mesoscale Modeling System Tutorial Class Notes and Users' Guide: MM5 Modeling System Version 2 (Mesoscale and Microscale Meteorology Division, National Center for Atmospheric Research, Boulder, CO).
  • 29.Döll, P., Kaspar, F. & Lehner, B. (2003) J. Hydrol. 270, 105–134. [Google Scholar]
  • 30.L′vovich, M. I. & White, G. F. (1990) in The Earth as Transformed by Human Action: Global and Regional Changes in the Biosphere Over the Past 300 Years, eds. Turner, B. L., Clark, W. C. & Kates, W. C. (Cambridge Univ. Press, Cambridge, U.K.).
  • 31.Shiklomanov, I. A. (2000) Water Int. 25, 11–32. [Google Scholar]
  • 32.Döll, P. & Siebert, S. (2002) Water Resour. Res. 38, 10.1029/2001WR000355. [DOI]
  • 33.Baumgartner, A. & Reichel, E. (1975) The World Water Balance (Elsevier, New York).
  • 34.L'vovich, M. I. (1979) World Water Resources and Their Future (LithoCrafters Chelsea, U.K.).
  • 35.Oyama, M. D. & Nobre, C. A. (2003) Geophys. Res. Lett. 30, 2199–2203. [Google Scholar]
  • 36.de Vries, H. J. M., Bollen, J., Bouwman, L., den Elzen, M. G. J., Janssen, M. & Kreilemann, E. (2000) Technol. Forecast Soc. 63, 137–174. [Google Scholar]
  • 37.Food and Agriculture Organization (2002) The State of Food and Agriculture. The State of Agriculture and Global Public Goods Ten Years After the Earth Summit (Food and Agriculture Organization of the United Nations, Rome).
  • 38.Intergovernmental Panel on Climate Change (IPCC) (1995) Climate Change 1995: The Science of Climate Change. Contribution of the Working Group 1 to the Second Assessment of the IPCC, eds. Houghton, J. T., Meira Filho, L. G., Callender, B. A., Harris, N., Kattenberg, A. & Maskell, K. (Cambridge Univ. Press, Cambridge, U.K.).
  • 39.Ohmura, A. & Wild, M. (2002) Science 298, 1345–1346. [DOI] [PubMed] [Google Scholar]
  • 40.Dai, A., Trenberth, K. E. & Qian, T. (2004) J. Hydrometeorol. 5, 1117–1130. [Google Scholar]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences

RESOURCES