• Berbery, E., , and Rasmusson E. , 1999: Mississippi moisture budgets on regional scales. Mon. Wea. Rev., 127 , 26542673.

  • Berbery, E., , Rasmusson E. , , and Mitchell K. , 1996: Studies of North American continental-scale hydrology using Eta model forecast products. J. Geophys. Res., 101 , 73057319.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Betts, A., , Viterbo P. , , and Wood E. , 1998: Surface energy and water balance for the Arkansas–Red River basin from the ECMWF reanalysis. J. Climate, 11 , 28812897.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Draper, C., 2007: The atmospheric water balance over the Murray–Darling Basin. Bureau of Meteorology Research Centre Research Rep. 127, 137 pp. [Available online at http://www.bom.gov.au/bmrc/pubs/researchreports/RR127.pdf.].

  • Ellett, K. M., , Walker J. , , Rodell M. , , Chen J. , , and Western A. , 2005: GRACE gravity fields as a new measure for assessing large-scale hydrological models. Proc. MODSIM 2005 Int. Congress on Modelling and Simulation, Melbourne, Australia, Modelling and Simulation Society of Australia and New Zealand, 2911–2917. [Available online at http://www.mssanz.org.au/modsim05/papers/ellett.pdf.].

  • Ellett, K., , Walker J. , , Western A. , , and Rodell M. , 2006: A framework for assessing the potential of remote-sensed gravity to provide new insight on the hydrology of the Murray–Darling Basin. Aust. J. Water Resour., 10 , 125138.

    • Search Google Scholar
    • Export Citation
  • Gibson, J., , Kallberg S. , , Uppala A. , , Hernandez A. , , Nomura A. , , and Serrano E. , 1997: ERA description. ERA Project Rep. Ser. 1, ECMWF, 63 pp.

  • Glowacki, T., , Penna N. , , and Bourke W. , 2006: Validation of GPS-based estimates of integrated water vapour for the Australian region and identification of diurnal variability. Aust. Meteor. Mag., 55 , 131148.

    • Search Google Scholar
    • Export Citation
  • Hirschi, M., , Seneviratne S. , , and Schär C. , 2006: Seasonal variations in terrestrial water storage for major midlatitude river basins. J. Hydrometeor., 7 , 3960.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kanamaru, H., , and Salvucci G. , 2003: Adjustments for wind sampling errors in an estimate of the atmospheric water budget of the Mississippi River basin. J. Hydrometeor., 4 , 518529.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kanamitsu, M., , and Saha S. , 1996: Systematic tendency error in budget calculations. Mon. Wea. Rev., 124 , 11451160.

  • Kerr, Y., , Waldteufel P. , , Wigneron J-P. , , Martinuzzi J-M. , , Font J. , , and Berger M. , 2001: Soil moisture retrieval from space: The Soil Moisture and Ocean Salinity (SMOS) mission. IEEE Trans. Geosci. Remote Sens., 39 , 17291735.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lioubimtseva, E., 2004: Climate change in arid environments: Revisiting the past to understand the future. Prog. Phys. Geogr., 28 , 502530.

  • Maheshwari, B., , Walker K. , , and McMahon T. , 1995: Effects of regulation on the flow regime of the River Murray, Australia. Regul. Rivers: Resour. Manage., 10 , 1538.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McMahon, T., 1992: Global Runoff: Continental Comparisons of Annual Flows and Peak Discharges. Catena-Verlag, 166 pp.

  • Moise, A., , Colman R. , , and Zhang H. , 2005: Coupled model simulations of current Australian surface climate and its changes under greenhouse warming: An analysis of 18 CMIP2 models. Aust. Meteor. Mag., 54 , 291307.

    • Search Google Scholar
    • Export Citation
  • Nicholls, N., 2004: The changing nature of Australian droughts. Climatic Change, 63 , 323336.

  • Pescod, N., 1994: A four-parameter, three-layer model of soil moisture based on hydraulic properties of the soil in the absence of vegetation. Parametrisation of Physical Processes: Proc. Fifth BMRC Modelling Workshop, Melbourne, Australia, Bureau of Meteorology Research Centre, 101–106.

    • Search Google Scholar
    • Export Citation
  • Prasad, A., , and Khan S. , 2002: Murray–Darling Basin dialogue on water and climate. Murray Darling Basin Commission, 48 pp. [Available online at http://www.waterandclimate.org/dialogue/basin/Murray-Darling/documents/Murray-Darling%20Report.pdf.].

  • Puri, K., , Dietachmayer G. , , Mills G. A. , , Davidson N. , , Bowen R. , , and Logan L. , 1998: The new BMRC Limited Area Prediction System, LAPS. Aust. Meteor. Mag., 47 , 203223.

    • Search Google Scholar
    • Export Citation
  • Rasmusson, E., 1968: Atmospheric water vapor transport and the water balance of North America II. Large-scale water balance investigations. Mon. Wea. Rev., 96 , 720734.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Richter, H., , Western A. , , and Chiew F. , 2004: The effect of soil and vegetation parameters in the ECMWF land surface scheme. J. Hydrometeor., 5 , 11311146.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roads, J., , Chen S. , , Kanamitsu M. , , and Juang H. , 1998: Vertical structure of humidity and temperature budget residuals over the Mississippi River basin. J. Geophys. Res., 103 , 37413759.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roads, J., , Kanamitsu M. , , and Stewart R. , 2002: CSE water and energy budgets in the NCEP–DOE reanalysis II. J. Hydrometeor., 3 , 227248.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roads, J., and Coauthors, 2003: GCIP water and energy budget synthesis (WEBS). J. Geophys. Res., 108 .8609, doi:10.1029/2002JD002583.

  • Rouse, W., and Coauthors, 2003: Energy and water cycles in a high-latitude, north-flowing river system. Bull. Amer. Meteor. Soc., 84 , 7387.

  • Ruprecht, E., , and Kahl T. , 2003: Investigation of the atmospheric water budget of the BALTEX area using NCEP/NCAR reanalysis data. Tellus, 55A , 426437.

    • Search Google Scholar
    • Export Citation
  • Seaman, R. S., , Bourke W. , , Steinle P. J. , , Hart T. , , Embery G. , , Naughton M. , , and Rikus L. , 1995: Evolution of the Bureau of Meteorology’s Global Assimilation and Prediction System. Part 1: Analysis and initialisation. Aust. Meteor. Mag., 44 , 118.

    • Search Google Scholar
    • Export Citation
  • Seneviratne, S., , Viterbo P. , , Lüthi D. , , and Schär C. , 2004: Inferring changes in terrestrial water storage using ERA-40 reanalysis data: The Mississippi River basin. J. Climate, 17 , 20392057.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Turato, B., , Reale O. , , and Siccardi F. , 2004: Water vapor sources of the October 2000 Piedmont flood. J. Hydrometeor., 5 , 693712.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Viterbo, P., , and Beljaars A. , 1995: An improved land surface parameterization scheme in the ECMWF model and its validation. J. Climate, 8 , 27162748.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Weymouth, G., , Mills G. , , Jones D. , , Ebert E. , , and Manton M. , 1999: A continental-scale daily rainfall analysis system. Aust. Meteor. Mag., 48 , 169179.

    • Search Google Scholar
    • Export Citation
  • Yatagai, A., 2003: Evaluation of hydrological balance and its variability in arid and semi-arid regions of Eurasia from ECMWF 15 year reanalysis. Hydrol. Processes, 17 , 28712884.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zangvil, A., , Portis D. , , and Lamb P. , 2004: Investigation of the large-scale atmospheric moisture field over the midwestern United States in relation to summer precipitation. Part II: Recycling of local evapotranspiration and association with soil moisture and crop yields. J. Climate, 17 , 32833301.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • View in gallery

    The Murray–Darling Basin (dashed outline) overlaid on mean annual (1961–90) observed rainfall (mm), from the Australian Bureau of Meteorology NCC data.

  • View in gallery

    Comparison of daily precipitation from analyzed rain gauge data and LAPS forecasts over the Murray–Darling Basin.

  • View in gallery

    Comparison of monthly precipitation (mm day−1) from analyzed rain gauge data and LAPS forecasts over the Murray–Darling Basin.

  • View in gallery

    Monthly water balance terms (mm day−1) from LAPS forecasts over the Murray–Darling Basin.

  • View in gallery

    Daily water balance terms (mm day−1) from LAPS forecasts over the Murray–Darling Basin, summer (December–February) 2003/04.

  • View in gallery

    Daily water balance terms (mm day−1) from LAPS forecasts over the Murray–Darling Basin, winter (June–August) 2003.

  • View in gallery

    Daily precipitation, and moisture flux divergence lagged one day, from LAPS forecasts over the Murray–Darling Basin (2000–04).

  • View in gallery

    Daily moisture flux vectors from LAPS forecasts over eastern Australia on 22 Feb 2004. The Murray–Darling Basin is outlined, and gray (black) vectors indicate net moisture flux divergence (convergence).

  • View in gallery

    Daily moisture flux vectors from LAPS forecasts over eastern Australia on (a) 10 Jan, (b) 13 Jan, and (c) 16 Jan 2004. The Murray–Darling Basin is outlined, and gray (black) vectors indicate net moisture flux divergence (convergence).

  • View in gallery

    Mean monthly (1966–96) precipitation, evaporation, and precipitation minus evaporation (mm day−1) from NRA over the Murray–Darling Basin. NRA was provided by the National Oceanic and Atmospheric Administration (NOAA) Climate Diagnostics Center (CDC; http://www.cdc.noaa.gov).

  • View in gallery

    Three-hourly forecast precipitable water (PW) and root-zone soil moisture (SM) over the Murray–Darling Basin (mm), from consecutive 1200:00 UTC LAPS model runs for January 2003.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 86 86 17
PDF Downloads 67 67 11

The Atmospheric Water Balance over the Semiarid Murray–Darling River Basin

View More View Less
  • 1 Centre for Australian Weather and Climate Research, Australian Bureau of Meteorology, Melbourne, Australia
© Get Permissions
Full access

Abstract

The atmospheric water balance over the semiarid Murray–Darling River basin in southeast Australia is analyzed based on a consecutive series of 3- to 24-h NWP forecasts from the Australian Bureau of Meteorology’s Limited Area Prediction System (LAPS). Investigation of the LAPS atmospheric water balance, including comparison of the forecast precipitation to analyzed rain gauge observations, indicates that the LAPS forecasts capture the general qualitative features of the water balance. The key features of the atmospheric water balance over the Murray–Darling Basin are small atmospheric moisture flux divergence (at daily to annual time scales) and extended periods during which the atmospheric water balance terms are largely inactive, with the exception of evaporation, which is consistent and very large in summer. These features present unique challenges for NWP modeling. For example, the small moisture fluxes in the basin can easily be obscured by the systematic errors inherent in all NWP models. For the LAPS model forecasts, there is an unrealistically large evaporation excess over precipitation (associated with a positive bias in evaporation) and unexpected behavior in the moisture flux divergence. Two global reanalysis products (the NCEP Reanalysis I and the 40-yr ECMWF Re-Analysis) also both describe (physically unrealistic) long-term negative surface water budgets over the Murray–Darling Basin, suggesting that the surface water budget cannot be sensibly diagnosed based on output from current NWP models. Despite this shortcoming, numerical models are in general the most appropriate tool for examining the atmospheric water balance over the Murray–Darling Basin, as the atmospheric sounding network in Australia has extremely low coverage.

Corresponding author address: Clara Draper, Centre for Australian Weather and Climate Research, GPO Box 1289, Melbourne 3001, Australia. Email: c.draper@bom.gov.au

Abstract

The atmospheric water balance over the semiarid Murray–Darling River basin in southeast Australia is analyzed based on a consecutive series of 3- to 24-h NWP forecasts from the Australian Bureau of Meteorology’s Limited Area Prediction System (LAPS). Investigation of the LAPS atmospheric water balance, including comparison of the forecast precipitation to analyzed rain gauge observations, indicates that the LAPS forecasts capture the general qualitative features of the water balance. The key features of the atmospheric water balance over the Murray–Darling Basin are small atmospheric moisture flux divergence (at daily to annual time scales) and extended periods during which the atmospheric water balance terms are largely inactive, with the exception of evaporation, which is consistent and very large in summer. These features present unique challenges for NWP modeling. For example, the small moisture fluxes in the basin can easily be obscured by the systematic errors inherent in all NWP models. For the LAPS model forecasts, there is an unrealistically large evaporation excess over precipitation (associated with a positive bias in evaporation) and unexpected behavior in the moisture flux divergence. Two global reanalysis products (the NCEP Reanalysis I and the 40-yr ECMWF Re-Analysis) also both describe (physically unrealistic) long-term negative surface water budgets over the Murray–Darling Basin, suggesting that the surface water budget cannot be sensibly diagnosed based on output from current NWP models. Despite this shortcoming, numerical models are in general the most appropriate tool for examining the atmospheric water balance over the Murray–Darling Basin, as the atmospheric sounding network in Australia has extremely low coverage.

Corresponding author address: Clara Draper, Centre for Australian Weather and Climate Research, GPO Box 1289, Melbourne 3001, Australia. Email: c.draper@bom.gov.au

1. Introduction

Arid regions are thought to be particularly sensitive to global climate change and yet the response of the world’s arid regions to future climate changes is not well understood (Lioubimtseva 2004). In the semiarid Murray–Darling River basin in southeast Australia, observations over the last half-century already suggest that rising regional temperatures have increased the severity of droughts (Nicholls 2004). However, the impacts of increasing CO2 and/or global climate change on the Murray–Darling Basin’s hydrology are uncertain. For example, different climate models simulate very different precipitation responses in Australia because of increased atmospheric CO2 concentrations, with differences in the net direction of precipitation change over the Murray–Darling Basin (Moise et al. 2005). As with many arid and semiarid regions, the Murray–Darling Basin has extremely strained water resources, making it susceptible to even small changes in the regional hydrology. Adverse changes to the basin’s hydroclimate have the potential to dramatically affect the Australian economy, as the basin produces over 40% of the national net agricultural earnings. There is a clear need to better understand the region’s hydroclimate, so that appropriate strategies can be developed to mitigate any adverse future changes.

This study will provide a better understanding of the Murray–Darling Basin’s hydroclimate, through examining the basin’s atmospheric water balance. The work presented here is mostly confined to the area-averaged water balance over the basin, and is in part a scoping study for the Global Energy and Water Cycle Experiment (GEWEX) Murray–Darling Basin Regional Hydroclimate Project (RHP) water budget study. The GEWEX program is aimed at better understanding the hydrological cycle and energy fluxes of the atmosphere, with the ultimate goal of predicting global and regional climate change. Under the auspices of GEWEX the atmospheric water balances over many regions around the globe have been investigated, typically in affiliation with the RHPs (formerly catchment-scale experiments). This work provides an interesting contrast with these investigations, since most of these were focused on more humid regions (e.g., see Roads et al. 2003; Rouse et al. 2003; Ruprecht and Kahl 2003), and this is the first study focused on the Australian region.

While the atmospheric water balance equation is conceptually simple, the selection of appropriate data for its study is more complicated. Traditionally, water balances have been investigated using radiosonde data (e.g., Rasmusson 1968), which are still used in regions with dense radiosonde networks (e.g., Kanamaru and Salvucci 2003; Zangvil et al. 2004; both of which concerned North American regions). However, recent studies are more often based on NWP-derived analyses and reanalyses (e.g., Roads et al. 2003; Ruprecht and Kahl 2003; Turato et al. 2004). Compared to the direct use of radiosonde data, NWP-derived analyses have the advantages of 1) lower expected errors, due to the incorporation of multiple information sources and minimization of known errors during data assimilation; and 2) increased information at higher spatial and temporal resolution, due to the dynamic continuity of NWP models. Alternatively, Berbery et al. (1996) estimated the atmospheric water balance from a series of consecutive model forecasts. The increased accuracy of analyses, relative to the raw data, is retained after the application of a (up to 24 h) forecast integration (Kanamitsu and Saha 1996), and Berbery et al. (1996) obtained results similar to those derived from the (dense) observation network in the Mississippi River basin.

Despite their considerable advantages, analyses and forecasts do not offer a perfect representation of the atmosphere. NWP systems inevitably contain errors in their initial states and in their parameterization and physics. As a result, NWP-derived balances (from either analyses or forecasts) do not close, and the magnitude of the balance residual is often similar to that of the leading terms (e.g., Roads et al. 2003; Ruprecht and Kahl 2003). Atmospheric water balance residuals are generated when increments are added to the model humidity to constrain it from drifting away from observations toward its own internal climate (determined by its imperfect physics and parameterization). For a thorough review of balance residuals see Kanamitsu and Saha (1996) and Roads et al. (1998). A water balance based on forecast output, rather than analyses, will likely contain a greater residual, since the additional forecast integration allows the model to drift further from the observed climate, requiring a greater correction.

In the Australian context, an investigation of the atmospheric water balance must be based on NWP-derived output, as the extremely poor radiosonde coverage is insufficient for making meaningful estimates of regional quantities: there are just four radiosonde sites in the Murray–Darling Basin, with one flight per day at each site. Accordingly, this study is based on forecast output from the Australian Government Bureau of Meteorology’s operational NWP system, the Limited Area Prediction System (LAPS; Puri et al. 1998). LAPS is preferred over a global reanalysis, as it is produced locally and has greater spatial and temporal resolution.

In summary, this study will investigate the atmospheric water balance from 2000 to 2004 over the Murray–Darling Basin, based on a series of consecutive 3- to 24-h LAPS forecasts. This work is primarily directed toward establishing, for the first time, the basic characteristics of the atmospheric water balance over the Murray–Darling Basin, and to inform future research priorities in this area. There has been relatively little work relating to the atmospheric water cycle of arid and semiarid regions, and this work will provide an interesting contrast with studies of more humid regions. Additionally, this study will determine the utility of the LAPS model as a tool for examining the atmospheric water cycle within the GEWEX Murray–Darling Basin RHP. This work also has broader relevance to the NWP community, since it will provide an alternative method of assessing the model’s treatment of moisture and help identify areas for future model improvements.

2. Study area and methods

a. The Murray–Darling Basin

The Murray–Darling Basin covers an area of 1 061 469 km2 in the southeast of Australia, and contains 20 major rivers, including Australia’s three longest—the Murray, Darling, and Murrumbidgee Rivers. Note that the basin satisfies the critical basin size of 105 km2 for NWP-based atmospheric water balance studies suggested by Hirschi et al. (2006). The Murray–Darling Basin is shown in Fig. 1, overlaid on the mean annual Australian rainfall. Most of the basin is an extensive low-lying plain with average annual rainfall between 100 and 300 mm, although there is a small elevated region in the Great Dividing Range on the basin’s southern and eastern boundary that receives up to 1000 mm, giving an annual average basinwide rainfall of 503.5 mm. As a result of the strong precipitation gradient across the basin, three river catchments in the Great Dividing Range generate nearly 50% of the basin’s mean annual runoff from just 11% of its area (Prasad and Khan 2002). In total over 90% of the basin is arid or semiarid (Maheshwari et al. 1995), and Prasad and Khan (2002) have estimated that 86% of the basin (presumably the low-lying plains) generates virtually no runoff except during floods.

In addition to being dry, the basin’s climate is extremely variable. In particular, temperature and precipitation in southeast Australia are strongly influenced by the multiyear ENSO cycle, contributing to Australia having one of the most variable runoff regimes in the world (McMahon 1992). The basin’s agricultural industry has prospered despite the dry and variable climate, largely due to extensive surface water regulation: an estimated 80% of the basin’s available surface water is used in the basin. However, the agricultural industry is still susceptible to climate fluctuations, and frequently suffers from both floods and droughts.

During the period of this study conditions were unusually dry in the Murray–Darling Basin. In particular, one of Australia’s most severe droughts occurred in 2002, following the development of El Niño conditions in the latter half of that year (Nicholls 2004). During the official 11-month dry period from March 2002 to January 2003, National Climate Centre (NCC) data indicate that rainfall was in the lowest decile or the lowest on record across virtually all of the Murray–Darling Basin. At the same time, evaporation was enhanced by above-average temperatures, with the Murray–Darling Basin experiencing what was then the highest mean daily maximum temperature on record in 2002 (Nicholls 2004). While rainfall was close to average in 2003, there was no period of extended above-average rainfall to fully remove the moisture deficiencies that developed in late 2002. As a result, surface drought conditions persisted through the rest of the study period, causing enormous agricultural losses and resulting in the Australian government spending $1 billion (Australian dollars) on drought relief assistance by May 2005. Other impacts of the drought included unusually severe forest fires, dust storms, and water shortages. For example, during the 2002–03 summer, 3 million hectares were burnt in forest fires, and seven towns required road deliveries to maintain the town water supply (Draper 2007). Given the unusually dry conditions across this period, the results and conclusions presented here are most relevant to below-average precipitation (and surface moisture) conditions in the basin.

b. The Limited Area Prediction System

LAPS is a regional NWP model and data assimilation system, used operationally at the Australian Bureau of Meteorology since July 1996. The last major model upgrade to LAPS was in late 1999, and there were no major model changes during this study. LAPS is run operationally twice a day, at 1200 UTC and 0000 UTC (10 a.m. and 10 p.m. Australian eastern standard time), and forecasts are archived at 3-hourly intervals. The model runs on a 0.375° latitude–longitude grid, and has 29 vertical (sigma) levels, up to ∼50 hPa. The data assimilation scheme is a three-dimensional multivariate statistical interpolation objective analysis system, with all fields analyzed on the same latitude–longitude–sigma grid as used by the forecast model (Seaman et al. 1995).

The land surface scheme in LAPS is an adaptation of the scheme developed by Viterbo and Beljaars (1995), and used in the 15-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-15; Gibson et al. 1997). The initialization of the LAPS land surface differs from that of Viterbo and Beljaars (1995) in that it is cold started. The soil moisture in LAPS is initialized 12 h before the free forecast with a first guess generated from the antecedent observed precipitation and a climatological evaporation, following Pescod (1994). However, the soil moisture “nudging” process of Viterbo and Beljaars (1995) has been retained; the LAPS first-guess is forecast forward 6 h, and soil moisture increments are added according to low-level humidity errors in the 6-h forecasts (with this process repeated once before the free forecast is made).

c. The atmospheric water balance

The atmospheric water balance over the Murray–Darling Basin has been calculated according to
i1525-7541-9-3-521-e1
where ∂W/∂t is the change in precipitable water (W) over time, E is evaporation (positive upward), P is precipitation, · Q is the vertically integrated horizontal moisture flux divergence, and R is the water balance residual; 〈·〉 indicates the area average over the Murray–Darling Basin. Equation (1) is based on the derivation of Rasmusson (1968), with the addition of an explicit residual term (R) in response to the expectation that NWP-derived water balances do not close. To reduce interpolation errors, the water balance terms have been calculated on the same latitude–longitude–sigma level grid used by LAPS, as advised by Berbery and Rasmusson (1999).

The water balance time series has been calculated using the 3- to 24-h LAPS forecasts at the maximum available 3-hourly resolution. The maximum resolution was used to better resolve the diurnal cycle, since all of the water balance terms (except precipitation) are calculated from instantaneous model quantities, and a preliminary analysis indicated a loss of accuracy at coarser temporal resolution. The water balance was first calculated separately for the 0000 and 1200 UTC model runs, before the average of the two was calculated for each 3-h time step. A daily water balance time series has then been produced by aggregating the 3-hourly values (with each day commencing at 0000 UTC).

d. Precipitation data

Observations of water balance variables over the Murray–Darling Basin are extremely limited, making verification of the LAPS forecast water balance difficult. Indeed, if there were extensive observations of the water balance terms then this study would likely be based on those data. Precipitation is the only water balance variable that is comprehensively observed across the Murray–Darling Basin, and the LAPS forecast Murray–Darling Basin precipitation has been compared to the Bureau of Meteorology’s daily real-time rain gauge analysis (Weymouth et al. 1999). The analysis projects precipitation data from approximately 200 rain gauges in the Murray–Darling Basin onto a 0.25° grid, with each daily analysis beginning at 2300:00 UTC (1 h earlier than the daily LAPS values). There are no known uncertainties in the rain gauge analyses in the Murray–Darling Basin and they are expected to provide an accurate representation of rainfall across the region.

e. The terrestrial water balance

While this work is focused on atmospheric moisture, the terrestrial component of the moisture cycle is relevant, since it is intimately linked to the atmospheric component via precipitation and evaporation. The terrestrial water balance equation is
i1525-7541-9-3-521-e2
where S is the terrestrial water storage and R0 is the surface runoff. The surface water transfer into the Murray–Darling Basin is negligible, and the basin is large enough that interbasin subsurface water transfer can be assumed to be zero. Both of these transfers have been neglected in Eq. (2).

As with the atmospheric water cycle, there is very little basin-scale terrestrial moisture data available for Australia. The Murray–Darling Basin surface discharge (an integrator of the runoff, R0) can be estimated from the river flow at Blanchtown, 274 km upstream from the mouth of the Murray River (the basin’s only surface discharge point). The net discharge in 2000–04 was just 6.3 × 10−3 mm day−1, while the long-term (1967–2004) average is 1.8 × 10−2 mm day−1. There is no coordinated in situ network to monitor the basin’s terrestrial moisture storage (S), and it is not well understood. While the basin’s terrestrial moisture storage or storage change over time are unknown, early results from the Gravity Recovery and Climate Experiment (GRACE) satellites indicate that the total amplitude of intra-annual variability in S for the Murray–Darling Basin in 2002 and 2003 was approximately 60 mm (Ellett et al. 2005), which distributed evenly through the year equates to 0.3 mm day−1. This is two orders of magnitude larger than the average daily discharge over the same period, indicating that at the time scales of this study ∂S/∂t cannot be neglected in Eq. (2), and R0 cannot be usefully compared to PE.

3. Precipitation verification

The LAPS forecast precipitation for the Murray–Darling Basin has been compared to the equivalent analyzed rain gauge data. The archive of the analyzed Murray–Darling Basin average rainfall is incomplete, particularly prior to October 2000, restricting comparison between the LAPS precipitation and analyzed rain gauge data to the period after this date. The LAPS forecast Murray–Darling Basin daily precipitation shows broad agreement with the analyzed values. The relationship between the daily forecast and analyzed precipitation is close to linear (Fig. 2), with a correlation of 0.90. The LAPS forecast precipitation (1.30 mm day−1) is slightly (<10%) higher than the analyzed average (1.16 mm day−1). However, the mean absolute error (0.52 mm day−1) is relatively large, being nearly 50% of the analyzed average rainfall, producing a wide spread in the forecasts to either side of the analyses in Fig. 2.

Comparison of the monthly precipitation time series reveals some differences (Fig. 3). While the forecasts have captured the main temporal variations of the analyses (the occurrence of high/low rainfall months is generally captured although not necessarily with correct magnitude), there are several periods during which LAPS is consistently biased. For example, the rainfall from September 2001 to January 2002 was overpredicted by 45% (LAPS = 1.66 mm day−1 and analyses = 1.14 mm day−1), and the extremely low rainfall in the spring [September–November (SON)] of 2002 was overpredicted by 100% (LAPS = 1.18 mm day−1 and analyses = 0.56 mm day−1). Precipitation was consistently underpredicted by 20% during the (austral) winter of 2003 (LAPS = 1.07 mm day−1 and analyses = 1.37 mm day−1) and the extremely high rainfall in November 2000 was dramatically overpredicted, with LAPS predicting 4.55 mm day−1, compared to 3.12 mm day−1 in the analyses (which is not surprising given the difficulty of accurately forecasting highly localized intense rainfall events).

4. Water balance results

The seasonal cycle and interannual variability of the LAPS forecast atmospheric water balance are illustrated in the monthly atmospheric water balance plots in Fig. 4. Examples of the daily water balance are provided here for the (austral) summer of 2003/04 (Fig. 5) and the winter of 2004 (Fig. 6). Both of these seasons were fairly typical of the 5-yr study period, with the exception of January 2004, during which precipitation was unusually high (3.18 mm day−1, or more than double the average precipitation forecast over the study period). In particular, intense rainfall in the middle of the month resulted in widespread flooding in the north of the basin. The spatial distribution of the daily LAPS forecast atmospheric moisture fluxes has been investigated over the same two seasons, focusing on significant precipitation events, to identify flow patterns bringing moisture into the basin.

Over long time periods (monthly or longer), precipitation and evaporation are the leading terms in the atmospheric water balance derived from the LAPS forecasts, with 5-yr averages of 1.40 and 1.88 mm day−1, respectively. The 5-yr average moisture flux divergence (2.5 × 10−2 mm day−1) is two orders of magnitude smaller than the leading terms, consistent with expectations for a semiarid river basin. The 5-yr average ∂W/∂t is also small (0.30 mm day−1), as is expected of any regional water balance. The 5-yr average water balance residual (−0.421 mm day−1) is smaller than the leading terms, but greater than ∂W/∂t and · Q. In contrast to the longer-term aggregates, on any individual day, · Q, ∂W/∂t, and sometimes precipitation are typically the greatest terms. Below, the individual LAPS forecast atmospheric water balance terms are described and referenced against expectations.

a. Precipitation

Precipitation over the Murray–Darling Basin has two distinct regimes, centered around summer and winter, reflecting the typical synoptic conditions in each season. During summer, precipitation maxima typically occur in the basin’s north and are generally associated with heavy convective storms (although significant rainfall may also occur in the south). From Fig. 5 the summer precipitation regime typically results in higher precipitation, individual precipitation events that are more intense and persist for longer, and greater variability between months. In contrast, during winter when precipitation is mostly associated with the passage of frontal systems over the south and southeast of the basin, monthly precipitation is typically lower and less variable (Fig. 6). While nearly half (45%; Prasad and Khan 2002) of the basin’s runoff is generated in the catchments in the Great Dividing Range (in the basin’s southeast), the sporadic summer rains in the basin’s north generate only a slightly lesser contribution (32%; Prasad and Khan 2002).

b. Surface water budget and evaporation

The surface water budget (precipitation minus evaporation) is of particular interest as it determines changes in surface moisture storage [together with surface runoff; see Eq. (2)]. The LAPS forecast surface water budget is dominated by the temporal dynamics of the more variable precipitation. During summer the forecast evaporation is consistently very high (∼3 mm day−1), implying large surface moisture losses each day. In the absence of significant precipitation events to replenish some of this moisture, large negative monthly surface water budgets (of up to −2 mm day−1) are generated during the summer (and also during the adjacent warm months). When significant precipitation does occur during the warmer months, the surface water budget becomes positive (and up to +1 mm day−1). While precipitation is generally reduced in winter, evaporation is reduced by a greater amount, generating small positive surface moisture budgets (up to +0.5 mm day−1). The seasonal cycle described by the LAPS forecast surface water budget, of a negative budget in the summer months and a positive budget in winter, is qualitatively consistent with the observed surface conditions. For example, groundwater and soil moisture records from within the basin show declining moisture in spring and summer, followed by a rebound in winter (see Richter et al. 2004).

In each year of the study, the large negative surface water budgets in the warmer months outweigh the smaller positive budgets in winter, leading to a negative annual surface water budget. The net surface moisture budget across the five years is very large and negative, with an average of −0.48 mm day−1, equivalent to 40% of the observed precipitation over this period (1.3 mm day−1). Over a long enough time frame that the atmospheric and terrestrial moisture storages can be assumed stationary, the surface water budget must be small and positive to balance the surface runoff (∼0.018 mm day−1). However, the time period over which the terrestrial moisture storage is stationary is not well known, particularly for Australia with its extremely variable climate. Given that rainfall was well below average and the basin experienced drought conditions throughout the study period, it is possible that the basin did experience continuous annual surface moisture loss over the 5-yr period. The Murray–Darling Basin has certainly experienced sustained surface drying in the past, since previous to surface water regulation the River Murray would dry to a series of pools during periods of low rainfall.

While it is possible that the basin’s surface did provide net water to the atmosphere over the 5 yr of this study, the LAPS forecast PE was extremely large and negative, particularly during spring and summer. Recall that the expected average magnitude of storage moisture change in the Murray–Darling Basin during 2002 and 2003 is ∼0.3 day−1 (and that monthly R0 is negligible). Yet from Fig. 4, in eight of the twelve spring and summer months during the same period, the LAPS forecast monthly PE was more negative than −1 mm day−1, with a maximum magnitude of −2.4 mm day−1 in March 2003. In contrast, the forecast PE during the winter months was within the expected magnitude, rarely being greater than 0.5 mm day−1.

In additional to the very large surface moisture losses implied by the LAPS forecast PE during the warmer months, the annual PE forecasts do not agree well with the climate record. In 2002, while the drought conditions did indicate a significant loss of surface moisture, the forecast PE of −0.6 mm day−1 is extremely large, being equivalent to 220 mm year−1, and nearly equal to the observed rainfall in that year (240 mm). Yet the LAPS forecast PE in both 2001 and 2003 was similarly large and negative (−0.63 and −0.61 mm day−1, respectively), when the conditions in those years did not suggest significant surface moisture losses; in 2001 rainfall was slightly below average and in 2003 it was close to average (1.19 and 1.39 mm day−1, respectively). Additionally, in 2000, La Niña conditions resulted in below-average temperatures and above-average precipitation across the basin (most of the Murray–Darling Basin receiving annual precipitation above the 80th percentile of NCC records), suggesting an above-average (hence positive) surface water budget, and yet the LAPS forecast budget was again negative (although much smaller than in other years; −0.13 mm day−1).

In summary, the LAPS forecast net PE is extremely large and negative, due largely to a negative bias in the forecast PE in spring and summer. Recall from section 3 that the net forecast precipitation across the study is biased high and Fig. 3 shows a frequent high precipitation bias in summer and spring. The negative PE bias must then be generated by a positive bias in the evaporation forecasts, with the greatest bias occurring in spring and summer, when evaporation forecasts are extremely large.

c. Moisture flux divergence

The average annual moisture flux divergence forecast by LAPS is small (0.026 mm day−1), as was expected for the semiarid Murray–Darling Basin. There is a tendency for precipitation episodes to follow one day after large convergence events, and for sustained precipitation events to be supported by continued moisture flux convergence. This is evident in the scatterplot of the LAPS forecast · Q and precipitation lagged by one day (Fig. 7): the points are clustered in the upper left quadrant, corresponding to strong convergence one day before heavy precipitation. Most days with large precipitation (and every day over 9 mm day−1) were preceded by moisture flux convergence the previous day, with the larger precipitation days generally being preceded by larger moisture flux convergence (although the converse is not evident). Scatterplots with no time lag, or with longer lags, have no obvious patterns (not shown).

Consideration of the moisture flux vectors for the two seasons included in Figs. 6 and 5 shows that moisture is generally carried into the basin with the predominant westerly flow. For example, on 22 February 2004 (Fig. 8), moisture entered the basin from the west, leading to precipitation across the central Murray–Darling Basin on 23 and 24 February. A different flow pattern occurred from 10–17 January 2004, when moisture flowed into the basin directly from the Gulf of Carpentaria to the northwest, resulting in heavy and sustained rain in the basin’s north. On 10 January (Fig. 9a) a monsoon low over the Gulf of Carpentaria was circulating moist tropical maritime air over Cape York. By January 13 (Fig. 9b) a trough had developed over inland Australia, associated with the passage of several cold fronts to the south. This trough extended up to the northwest to connect with the monsoon low, cutting off the easterly flow south of the low. By 16 January (Fig. 9c), the low pressure system and associated circulation of moist air had been drawn south, channeling the moist airflow into the north of the basin.

Annually, moisture flux convergence (− · Q) can be equated to PE (by neglecting ∂W/∂t) to estimate the surface water budget. As with the forecast of negative PE, the 5-yr net moisture divergence in this study implies a net surface moisture loss. At 0.026 mm day−1 the moisture flux divergence is closer to the expected magnitude, being equivalent to 2% of the observed precipitation over the study period. Additionally, from Fig. 4 the monthly magnitude of − · Q − ∂W/∂t (the latter must be included at monthly time scales) matches the expected magnitude of monthly terrestrial moisture storage change, generally being less than 0.5 mm day−1. While the monthly moisture flux divergence time series is noisy, there is a weak seasonal cycle in the five-year average for each month, with small negative · Q (convergence) in winter months and small positive · Q (divergence) in other months, consistent with the expected direction of the surface water budget.

While it has a more accurate scale than the forecast PE, the · Q for each individual year does not reflect the expected surface water budget signal. For example, in 2001 and 2002 the LAPS forecast · Q is an order of magnitude larger than in the other years of the study (0.26 and −0.26 mm day−1, respectively). The forecast of a net moisture convergence in 2002 is inconsistent with the development of drought conditions in that year. In contrast to 2001, while a net divergence is not unlikely, precipitation was only slightly below average and there is no evidence of an unusually strong divergence, as was forecast. For the remaining years of the study the LAPS forecast annual · Q is in agreement with expectations [convergence in 2000 (−0.024 mm day−1) and divergence in each of 2003 (0.095 mm day−1) and 2004 (0.065 mm day−1)].

d. Rate of change of precipitable water

The monthly rate of change of precipitable water is very small, and dW/dt is erratic as it is the difference between two instantaneous values. In contrast, the average monthly precipitable water has a smooth seasonal cycle with maxima in the warmer months and a seasonal range of 15–22 mm. At the daily time scale, ∂W/∂t closely mirrors · Q, with convergence corresponding to increasing precipitable water.

e. Water balance residual

The residual is a significant term in the LAPS forecast atmospheric water balance over the Murray–Darling Basin. The 5-yr average residual (−0.421 mm day−1) is approximately one quarter of the leading terms and is greater than the two remaining terms (∂W/∂t and · Q). This is fairly typical of NWP-based atmospheric water balance studies. For example, Kanamitsu and Saha (1996) estimate that the ratio of the residual to the leading terms is typically 10%–20%, and for the RHPs investigated by Roads et al. (2003) using National Centers for Environmental Prediction–Department of Energy (NCEP–DOE) Reanalyses II, this ratio averaged 22%.

The 5-yr-average LAPS water balance residual for each calendar month is negative. The largest negative residuals consistently occur in summer and spring, with the monthly minimum being −1.72 mm day−1 in December 2002. Recall that the water balance has been calculated from a series of consecutive (and internally consistent) 24-h forecasts. When each new forecast is used in the water balance calculation, if the precipitable water in the newly initialized forecast is lower (higher) than that in the previous day’s 24-h forecast then the introduction of the new forecast effectively removes (adds) moisture from the water balance, generating a negative (positive) residual increment. The persistently negative residual in this study then signals a systematic tendency for the atmospheric humidity to be increased in the 24-h LAPS forecasts. The large negative residuals during spring and summer are consistent with the positive bias in the forecast evaporation at these times (although this is not necessarily the only cause).

5. Discussion

From the results presented in sections 3 and 4, the LAPS 3- to 24-h forecasts are able to capture the broad features of the area-averaged atmospheric water balance over the Murray–Darling Basin. LAPS exhibits reasonable skill in forecasting the Murray–Darling Basin area-average precipitation, although there is a significant spread in the results to either side of the observations, and several periods of sustained bias. Also of concern is that the LAPS forecast evaporation over the study period has a substantial positive bias during spring and summer, and the water balance residual is systematically large and negative, indicating a (likely related) tendency toward increasing atmospheric moisture. Despite these shortcomings, the water cycle described by the LAPS forecasts qualitatively conforms to expectations of a semiarid climate.

The basin’s water cycle consists (simplistically) of a background state during which there is little activity in all of the water balance terms except for evaporation, which remains consistent (and very high during summer), driving steady surface moisture losses. While the LAPS forecast evaporation was persistently high, evaporation in the basin is moisture limited and should decrease as the soil moisture approaches the wilting point. The periods of extended surface drying are punctuated by short relatively active events, consisting of moisture flux convergence (and increasing precipitable water), supporting precipitation that leads to rapid surface moisture gain. Over time, convergence is balanced by divergence, so that the net long-term transport ( · Q) of moisture into the basin is low.

The net moisture flux divergence in the Murray–Darling Basin is much smaller across all time scales than is typically observed in more humid regions (both in absolute terms, and relative to the leading terms). In particular the small annual net · Q in the Murray–Darling Basin was generated by consistently small monthly values (rarely exceeding 0.5 mm day−1), rather than an offset between periods of strong convergence and divergence, as often occurs in regions with small annual · Q [e.g., the monsoon influenced the Coupling Tropical Atmosphere and Hydrological Cycle experiment region in western equatorial Africa, and the Huaihe River basin in China, as reported by Roads et al. (2002)]. This result does not generalize to other regions with dry climates, as the arid zones in Asia considered by Yatagai (2003) had prominent seasonal cycles in · Q with maxima between ±0.8–2.5 mm day−1.

Despite the small magnitude of · Q, its importance to the basin’s hydrological cycle should not be underestimated. Recall that the majority of instances of large daily precipitation followed one day after a large moisture flux convergence (Fig. 7). These moisture flux/precipitation events will frequently be highly localized, and so are unlikely to be detected by the current radiosonde network (of once daily soundings at four locations across the basin). This reinforces the earlier decision that the current atmospheric moisture data network is inadequate for examining the water balance. In the near future, the LAPS model could benefit from the assimilation of GPS-derived precipitable water (Glowacki et al. 2006), which has the particular advantage of greatly increased temporal resolution (up to 30 s). However, even with the additional GPS-derived moisture data, the spatial coverage of atmospheric moisture soundings across Australia will remain sparse, necessitating the continued use of numerical models (such as LAPS) to investigate regional moisture quantities.

For the daily moisture flux vector maps investigated in this study, moisture generally entered the basin with the prevailing westerly flow, with the exception of the direct northwesterly flow of moisture from the Gulf of Carpentaria that led to intense precipitation in mid-January 2004. Only two 3-month periods were considered here, and a much longer time period would need to be investigated to determine whether this result could be generalized. Additionally, a trace analysis would be useful for establishing the source regions of moisture flowing into the basin. Of interest are the relative proportions of moisture sourced from the moist tropical air mass to the northwest and from the colder Southern Ocean air mass to the west and southwest, and also how often moisture from the South Pacific penetrates over the Great Dividing Range and into the basin.

The surface water budget is one of the most important aspects of the water cycle, since it quantifies the flux of moisture between the surface and the atmosphere. While the LAPS model is able to describe the broad features of the water balance as described earlier, it cannot forecast physically realistic surface water budgets. Recall from section 2 that over the 5 yr of this study, the LAPS forecast PE implies surface moisture losses that are far larger than previous estimates, particularly during summer and spring. Since the surface water budget is the small difference between two comparatively large and parameterized terms, it is not surprising that it cannot be estimated from P and E forecasts. In response to the greater uncertainty of the parameterized moisture processes in NWP models, the “aerological method” of using the vertically integrated atmospheric terms in Eq. (1) can be used to estimate the surface water budget (e.g., Kanamaru and Salvucci 2003; Zangvil et al. 2004). However, from section 3 this method still does not yield the expected results. While the LAPS forecast · Q has a more reasonable scale than the forecast PE, the annual signal in the forecast · Q appears to be incorrect.

This inability to predict accurate surface moisture budgets over the Murray–Darling Basin is not unique to LAPS, and both the NCEP Reanalysis I (NRA) and 40-yr ECMWF Re-Analysis (ERA-40) also forecast physically unrealistic budgets. Over the much longer (multidecadal) time scale of the reanalyses, the budget should be positive (and ultimately balance the long-term observed surface discharge of 0.02 mm/day), yet both ERA-40 and NRA describe negative long-term budgets. Figure 10 shows the 30-yr (1966–96) average monthly surface water budget over the Murray–Darling Basin based on NRA. The seasonal cycle of the monthly averages is similar to that predicted by LAPS; precipitation exceeds evaporation only in May, resulting in a negative long-term annual surface water budget of −0.22 mm day−1. For ERA-40, the long-term (1958–2001) surface water budget is net negative over the Murray–Darling Basin, based on either PE (Hirschi et al. 2006) or the aerological approach (Seneviratne et al. 2004).

These results indicate that the surface water budget over the semiarid Murray–Darling Basin cannot be confidently estimated using output from current NWP models, even though this approach has been successful for other regions, such as the Mississippi River basin (Seneviratne et al. 2004; Hirschi et al. 2006). In the Murray–Darling Basin the flux of moisture is consistently low, so that for an NWP system to forecast (or analyze) an accurate surface water budget, its errors need be even lower. Even a small (absolute) systematic error in the relevant atmospheric terms will obscure the true budget signal, as occurred for LAPS, ERA-40, and NRA. Future advances in estimating surface moisture dynamics are likely to come from approaches that more directly monitor the surface state. For example, early results indicate that the Gravity Recovery and Climate Experiment can provide a statistically significant measure of terrestrial moisture storage change across the Murray–Darling Basin (Ellett et al. 2006).

While the inability of NWP models to forecast accurate surface water budgets reduces their utility for studying the water cycle over the Murray–Darling Basin, this is not directly relevant to the broader NWP community, since neither · Q nor evaporation are routinely diagnosed from NWP forecasts. However, these findings do have relevance to NWP since they indicate the presence of systematic errors in the models, which will affect forecasts of other variables. For example, the positive evaporation bias detected in this study will generate spurious low-level humidity, which in a convective environment could influence the diagnosis of boundary layer stability, and hence forecasts of deep convection. The model shortcomings identified in this study are then useful for identifying areas of future model improvement.

In this study there is a positive bias in the LAPS forecast evaporation during spring and summer, and the water balance residual is consistently negative, indicating a tendency toward increased humidity in 24-h forecasts. The excess humidity is supplied by the model surface, as demonstrated by Fig. 11, which shows the diurnal cycle of the precipitable water and surface moisture for January 2003 from the 1200 UTC LAPS forecasts. There was very little precipitation through this month, and in general the surface moisture is decreased and the precipitable water is increased during each forecast, due to evaporative transfer of moisture from the surface to the atmosphere. When each new forecast is introduced into the time series the surface moisture is abruptly increased and the atmospheric moisture is abruptly decreased (the latter generates the negative water balance residual).

In the absence of data for either evaporation or soil moisture the cause of the above behavior cannot be attributed to either a positive bias in the forecast evaporation or to errors in the initial conditions. The soil moisture initialization in LAPS is currently being investigated as a potential source of the evaporation bias. Soil moisture is cold started in LAPS, and the subsequent moisture nudging has been shown to be sensitive to errors in other model components when used in the ERA-15 model (Betts et al. 1998). One option being considered is to assimilate remotely sensed surface moisture data into the Australian NWP models. While a more realistic soil moisture representation may not necessarily generate instant improvements in the surface flux forecasts due to errors in the model parameterization (Richter et al. 2004), realistic surface moisture fields (that do not respond to model errors) are expected to make the diagnosis of model errors simpler and to aid the selection of more appropriate soil parameters.

6. Conclusions

The characteristics of the Murray–Darling Basin’s atmospheric water balance highlight the aridity of the region. The water balance is typified by very small atmospheric moisture flux convergence (at daily to annual time scales), and a background state with little atmospheric activity that is punctuated by episodes of moisture flux convergence leading to precipitation. In between precipitation events, evaporation remains consistent and can be very high in summer, leading to the steady depletion of surface moisture, at least until the wilting point is approached. Australian precipitation is unusually variable, and in years with little rainfall, the surface moisture will be depleted to this lower limit, resulting in strained surface water supplies.

While the LAPS model simulates the water balance with enough accuracy to be able to describe the general characteristics described above, there are some clear limitations in its applicability as a tool for examining the water cycle. Currently, the surface water budget over the Murray–Darling Basin cannot be unambiguously quantified using atmospheric forecasts or analyses. NWP models contain systematic errors (as evidence by their inability to close the atmospheric water balance) and those errors can easily obscure the true (relatively small) surface water signal of the Murray–Darling Basin. For the LAPS model there was a substantial high evaporation bias during the summer and spring during this study, and research is underway to amend this. Novel remote sensing approaches to monitoring surface moisture, such as GRACE and the planned Soil Moisture Ocean Salinity Mission (SMOS; Kerr et al. 2001), will be useful for both improving Australian NWP models and for better understanding the Murray–Darling Basin surface water budget.

Despite the limitations identified above, numerical models are in general the most appropriate tool for quantifying regional atmospheric moisture over Australia. Numerical models, such as LAPS, can provide detail at much greater spatial and temporal resolution than is available from current atmospheric sounding data over Australia (or will be available in the foreseeable future). This resolution is particularly important in the semiarid Murray–Darling Basin, where important moisture fluxes are small scale and episodic.

Acknowledgments

This work was funded by the Cooperative Research Centre for Catchment Hydrology and the Australian Government Bureau of Meteorology Research Centre (BMRC). We thank Beth Ebert (BMRC) for supplying the Murray–Darling Basin precipitation analyses, Alan Wain (BMRC) for assistance generating figures, and Kevin Ellett (U.S. Geological Survey) for advice regarding terrestrial moisture storage in the Murray–Darling Basin.

REFERENCES

  • Berbery, E., , and Rasmusson E. , 1999: Mississippi moisture budgets on regional scales. Mon. Wea. Rev., 127 , 26542673.

  • Berbery, E., , Rasmusson E. , , and Mitchell K. , 1996: Studies of North American continental-scale hydrology using Eta model forecast products. J. Geophys. Res., 101 , 73057319.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Betts, A., , Viterbo P. , , and Wood E. , 1998: Surface energy and water balance for the Arkansas–Red River basin from the ECMWF reanalysis. J. Climate, 11 , 28812897.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Draper, C., 2007: The atmospheric water balance over the Murray–Darling Basin. Bureau of Meteorology Research Centre Research Rep. 127, 137 pp. [Available online at http://www.bom.gov.au/bmrc/pubs/researchreports/RR127.pdf.].

  • Ellett, K. M., , Walker J. , , Rodell M. , , Chen J. , , and Western A. , 2005: GRACE gravity fields as a new measure for assessing large-scale hydrological models. Proc. MODSIM 2005 Int. Congress on Modelling and Simulation, Melbourne, Australia, Modelling and Simulation Society of Australia and New Zealand, 2911–2917. [Available online at http://www.mssanz.org.au/modsim05/papers/ellett.pdf.].

  • Ellett, K., , Walker J. , , Western A. , , and Rodell M. , 2006: A framework for assessing the potential of remote-sensed gravity to provide new insight on the hydrology of the Murray–Darling Basin. Aust. J. Water Resour., 10 , 125138.

    • Search Google Scholar
    • Export Citation
  • Gibson, J., , Kallberg S. , , Uppala A. , , Hernandez A. , , Nomura A. , , and Serrano E. , 1997: ERA description. ERA Project Rep. Ser. 1, ECMWF, 63 pp.

  • Glowacki, T., , Penna N. , , and Bourke W. , 2006: Validation of GPS-based estimates of integrated water vapour for the Australian region and identification of diurnal variability. Aust. Meteor. Mag., 55 , 131148.

    • Search Google Scholar
    • Export Citation
  • Hirschi, M., , Seneviratne S. , , and Schär C. , 2006: Seasonal variations in terrestrial water storage for major midlatitude river basins. J. Hydrometeor., 7 , 3960.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kanamaru, H., , and Salvucci G. , 2003: Adjustments for wind sampling errors in an estimate of the atmospheric water budget of the Mississippi River basin. J. Hydrometeor., 4 , 518529.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Kanamitsu, M., , and Saha S. , 1996: Systematic tendency error in budget calculations. Mon. Wea. Rev., 124 , 11451160.

  • Kerr, Y., , Waldteufel P. , , Wigneron J-P. , , Martinuzzi J-M. , , Font J. , , and Berger M. , 2001: Soil moisture retrieval from space: The Soil Moisture and Ocean Salinity (SMOS) mission. IEEE Trans. Geosci. Remote Sens., 39 , 17291735.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lioubimtseva, E., 2004: Climate change in arid environments: Revisiting the past to understand the future. Prog. Phys. Geogr., 28 , 502530.

  • Maheshwari, B., , Walker K. , , and McMahon T. , 1995: Effects of regulation on the flow regime of the River Murray, Australia. Regul. Rivers: Resour. Manage., 10 , 1538.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McMahon, T., 1992: Global Runoff: Continental Comparisons of Annual Flows and Peak Discharges. Catena-Verlag, 166 pp.

  • Moise, A., , Colman R. , , and Zhang H. , 2005: Coupled model simulations of current Australian surface climate and its changes under greenhouse warming: An analysis of 18 CMIP2 models. Aust. Meteor. Mag., 54 , 291307.

    • Search Google Scholar
    • Export Citation
  • Nicholls, N., 2004: The changing nature of Australian droughts. Climatic Change, 63 , 323336.

  • Pescod, N., 1994: A four-parameter, three-layer model of soil moisture based on hydraulic properties of the soil in the absence of vegetation. Parametrisation of Physical Processes: Proc. Fifth BMRC Modelling Workshop, Melbourne, Australia, Bureau of Meteorology Research Centre, 101–106.

    • Search Google Scholar
    • Export Citation
  • Prasad, A., , and Khan S. , 2002: Murray–Darling Basin dialogue on water and climate. Murray Darling Basin Commission, 48 pp. [Available online at http://www.waterandclimate.org/dialogue/basin/Murray-Darling/documents/Murray-Darling%20Report.pdf.].

  • Puri, K., , Dietachmayer G. , , Mills G. A. , , Davidson N. , , Bowen R. , , and Logan L. , 1998: The new BMRC Limited Area Prediction System, LAPS. Aust. Meteor. Mag., 47 , 203223.

    • Search Google Scholar
    • Export Citation
  • Rasmusson, E., 1968: Atmospheric water vapor transport and the water balance of North America II. Large-scale water balance investigations. Mon. Wea. Rev., 96 , 720734.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Richter, H., , Western A. , , and Chiew F. , 2004: The effect of soil and vegetation parameters in the ECMWF land surface scheme. J. Hydrometeor., 5 , 11311146.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roads, J., , Chen S. , , Kanamitsu M. , , and Juang H. , 1998: Vertical structure of humidity and temperature budget residuals over the Mississippi River basin. J. Geophys. Res., 103 , 37413759.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roads, J., , Kanamitsu M. , , and Stewart R. , 2002: CSE water and energy budgets in the NCEP–DOE reanalysis II. J. Hydrometeor., 3 , 227248.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Roads, J., and Coauthors, 2003: GCIP water and energy budget synthesis (WEBS). J. Geophys. Res., 108 .8609, doi:10.1029/2002JD002583.

  • Rouse, W., and Coauthors, 2003: Energy and water cycles in a high-latitude, north-flowing river system. Bull. Amer. Meteor. Soc., 84 , 7387.

  • Ruprecht, E., , and Kahl T. , 2003: Investigation of the atmospheric water budget of the BALTEX area using NCEP/NCAR reanalysis data. Tellus, 55A , 426437.

    • Search Google Scholar
    • Export Citation
  • Seaman, R. S., , Bourke W. , , Steinle P. J. , , Hart T. , , Embery G. , , Naughton M. , , and Rikus L. , 1995: Evolution of the Bureau of Meteorology’s Global Assimilation and Prediction System. Part 1: Analysis and initialisation. Aust. Meteor. Mag., 44 , 118.

    • Search Google Scholar
    • Export Citation
  • Seneviratne, S., , Viterbo P. , , Lüthi D. , , and Schär C. , 2004: Inferring changes in terrestrial water storage using ERA-40 reanalysis data: The Mississippi River basin. J. Climate, 17 , 20392057.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Turato, B., , Reale O. , , and Siccardi F. , 2004: Water vapor sources of the October 2000 Piedmont flood. J. Hydrometeor., 5 , 693712.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Viterbo, P., , and Beljaars A. , 1995: An improved land surface parameterization scheme in the ECMWF model and its validation. J. Climate, 8 , 27162748.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Weymouth, G., , Mills G. , , Jones D. , , Ebert E. , , and Manton M. , 1999: A continental-scale daily rainfall analysis system. Aust. Meteor. Mag., 48 , 169179.

    • Search Google Scholar
    • Export Citation
  • Yatagai, A., 2003: Evaluation of hydrological balance and its variability in arid and semi-arid regions of Eurasia from ECMWF 15 year reanalysis. Hydrol. Processes, 17 , 28712884.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Zangvil, A., , Portis D. , , and Lamb P. , 2004: Investigation of the large-scale atmospheric moisture field over the midwestern United States in relation to summer precipitation. Part II: Recycling of local evapotranspiration and association with soil moisture and crop yields. J. Climate, 17 , 32833301.

    • Crossref
    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

The Murray–Darling Basin (dashed outline) overlaid on mean annual (1961–90) observed rainfall (mm), from the Australian Bureau of Meteorology NCC data.

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Fig. 2.
Fig. 2.

Comparison of daily precipitation from analyzed rain gauge data and LAPS forecasts over the Murray–Darling Basin.

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Fig. 3.
Fig. 3.

Comparison of monthly precipitation (mm day−1) from analyzed rain gauge data and LAPS forecasts over the Murray–Darling Basin.

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Fig. 4.
Fig. 4.

Monthly water balance terms (mm day−1) from LAPS forecasts over the Murray–Darling Basin.

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Fig. 5.
Fig. 5.

Daily water balance terms (mm day−1) from LAPS forecasts over the Murray–Darling Basin, summer (December–February) 2003/04.

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Fig. 6.
Fig. 6.

Daily water balance terms (mm day−1) from LAPS forecasts over the Murray–Darling Basin, winter (June–August) 2003.

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Fig. 7.
Fig. 7.

Daily precipitation, and moisture flux divergence lagged one day, from LAPS forecasts over the Murray–Darling Basin (2000–04).

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Fig. 8.
Fig. 8.

Daily moisture flux vectors from LAPS forecasts over eastern Australia on 22 Feb 2004. The Murray–Darling Basin is outlined, and gray (black) vectors indicate net moisture flux divergence (convergence).

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Fig. 9.
Fig. 9.

Daily moisture flux vectors from LAPS forecasts over eastern Australia on (a) 10 Jan, (b) 13 Jan, and (c) 16 Jan 2004. The Murray–Darling Basin is outlined, and gray (black) vectors indicate net moisture flux divergence (convergence).

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Fig. 10.
Fig. 10.

Mean monthly (1966–96) precipitation, evaporation, and precipitation minus evaporation (mm day−1) from NRA over the Murray–Darling Basin. NRA was provided by the National Oceanic and Atmospheric Administration (NOAA) Climate Diagnostics Center (CDC; http://www.cdc.noaa.gov).

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Fig. 11.
Fig. 11.

Three-hourly forecast precipitable water (PW) and root-zone soil moisture (SM) over the Murray–Darling Basin (mm), from consecutive 1200:00 UTC LAPS model runs for January 2003.

Citation: Journal of Hydrometeorology 9, 3; 10.1175/2007JHM889.1

Save