Role of Pacific Ocean Climate Variability on Rainfall Variability in the Murrumbidgee Catchment, Australia

Reports show a necessity for measuring spatial water availability to manage the problem of water scarcity with rising per capita water demand. The ever-increasing demand for water is being enforced by mounting agricultural food demand to cope with an increasing population and industrial development. The nature of water cycle determines water availability and is influenced by climate variability. Irrespective of the scale, agriculture, which is accounted to
be the highest water consumer in the world, is obviously affected by climate variability. Characteristics of climate variability such as El-Niño southern oscillation, Indian Ocean Dipole, inter-tropical convergence zone, and other known circulations and attributes found in oceans, are connected with local and global precipitation. In particular, changing climate features in the surface of the Pacific Ocean (PO) drive a massive quantity of water vapour into the atmosphere, creating clouds that in turn deliver precipitation. This results in variation influencing agriculture around the world, not least within neighbouring Australia, in particular, the eastern and northern parts of Australia (ENA), and the Murrumbidgee Catchment (MC). Analysis revealed that PO climate variability is highly responsible for rainfall in the ENA, which covers about half of Australia and directly faces the west part of PO. These climate patterns are mostly based on the effect of PO surface diverse sea surface temperature (SST), turbulence and wind vector above the sea surface etc., which contribute to form water vapour and clouds, and other climate conditions in the atmosphere. When water vapour, in the form of total precipitable water (TPW) and cloud liquid water content
(CLWC), is subjected to particular conditions such as cloud top air temperature (CTAT) and cloud top air pressure (CTAP), rainfall occurs. Therefore, TPW, CLWC, CTAT and CTAP in the atmosphere over PO have been considered to carry PO Surface (POS) climate variability to the ENA.

The MC, which is about 1.1% of Australia in size, is situated in southern New South Wales (part of ENA), would not be considered separately in the context of the Australian climate that experiences Pacific and other oceanic influences. Water availability and the rainfall variability is heavily connected to the water cycle: vaporisation, cloud formation, condensation and precipitation. Precipitation, which is a major source of water availability for human requirements, varies spatially and temporally, within MC, ENA and Australia as
well as other parts of the world.

Agricultural production in the MC, which contributes a considerable amount to the Australian economy by producing 25% of fruits and vegetables and 50% of rice, mainly from irrigated agriculture, is encased in the issues related to water availability that oscillate between long-term droughts and floods. A consequence of the highly variable climate in the catchment is the damage to crops and yields as well as the discouragement of farmers and growers in the region. Streamlined research is focussing on how the POS climate variability
impacts rainfall variability across the MC and ENA via the atmosphere. Delivering the PO effect on the MC is a time delay exercise, which involves evaporation from the PO, atmospheric condensation in the west PO producing rainfalls to the ENA, then extraction of rainfalls for the MC from the ENA. This has been itemised into three model components, POS2ATMOS, ATMOS2ENA and ENA2MC. Sea surface temperature (SST), wind vector, surface air temperature and related energies of the PO surface add moisture to the air parcels in the atmosphere experiencing atmospheric circulations, making clouds and water vapours in the stratosphere. The model developed for this component has been considered
to be in two-months delay. The intermediate model developed for atmospheric TPW and CLWC under the conditions of CTAT and CTAP of the west PO to distribute rainfalls in the ENA, is considered as being one month lagged and finally, MC’s rainfall variability could be estimated from ENA rainfall with obviously no delay by a downscaling model. Therefore, the overall delay in delivering PO surface effect on both the ENA and MC has been three months
in total.

The three models were developed using artificial intelligence (AI) rather than physical properties because to undertake the latter would have been a massive project far beyond the scale of this research. Instead, AI’s artificial neural network, which has capabilities in generalisation, adaptation and non-linearity in modelling, has provided promising results in this research.

The relationship between November POS climate variability and January TPW, CLWC, CTAT and CTAP was found to be significant, having 0.99, 0.96, 0.82 and 0.95 correlation coefficients with high performance measures showing AI’s forecasting capability for atmospheric variability. The position of the POS has been the most influential variable, with SST, surface air temperature and the meridional wind component being the next level prominent inputs for this particular model.

January atmospheric variability in the ENA selected coastal region (ENA-SCR) was found to be forecasting February ENA rainfall as high, such as 0.92 correlation coefficients in this case, with promising performance measure. All participated variables were important in both POS2ATMOS and ATMOS2ENA models that could provide a facility to forecast ENA rainfall variability for the month of February from three months lagged (November month) POS variability. The most predominant inputs were the distance between ENA and ENA-SCR, and CTAP.

No direct relationship was found between rainfall of ENA and MC (unlike in two previous relationships between PO and atmosphere), and ENA-SCR and ENA rainfall. In the middle of the analysis, MC rainfall was found to be strongly linked to its landscape in terms of elevation and it was found to be true for all the months. It was also discovered that certain patterns have existed in ENA rainfall distributions. Those were quantified by introducing partial rainfall patterns, giving rise to the downscaling model.

As mentioned, MC rainfall distribution in February was found to be 99.6% correlated with its landscape (elevation), which became the natural weight matrix for downscaling MC rainfall. It was interesting to reveal that ENA partial rainfalls weighted by MC elevation were the dominating inputs downscaling MC total and partial rainfalls for all the months. Downscaling relationship for the month of February was at 0.97, 0.77, 0.95, 0.87 and 0.93 for MC total rainfall, four partial variables, respectively, indicating the forecasting significance of the model.

Combination of all the above models could deliver POS climate variability effect on MC’s February total rainfall variability with great success and there are twelve such combinations available for all months. Furthermore, these models have opened pathways to forecast three months' lagged extreme and moderate rainfall events in MC from the effect of POS. Therefore, this project facilitates the farmers and irrigators to optimise agricultural production and to improve their economic benefits by providing them prior knowledge about the partiality of rainfall conditions. From the combination of ENA partial rainfalls and MC elevation, the latter stipulates possible drought or moderate and high flood levels in advance, enabling facility for climate adaptation.

Ultimately, a single Bayesian model has also been proposed encapsulating the results of these three models as a future research project.

Eleven additional sets of polymorphic versions developed from the three models, predicting MC rainfall variability for the other eleven months, demonstrate promising results, as depicted in Appendices II and III. This completes forecasting MC rainfall variability from the POS effect in advance of three months, making it available for all the months of the year and
ultimately benefiting the farmers of the MC.