2.1.1.2 Statistical analysis and interpolation


All geographic data specific to the Gloucester subregion were obtained from state or national datasets. This means no statistical analysis or interpolation was performed to generate any of the geographic datasets. However, to characterise errors of the input climate data used for the water balance modelling (reported in companion product 2.5 for the Gloucester subregion (Herron et al., 2018)), some subregion-specific spatial analysis was performed. This is outlined in the following sections.

In addition to the work documented in Section 2.1.1.2.1 there is a separate ‘climate sensitivity’ activity underway to determine how sensitive the hydrological models in the Gloucester subregion are to climate inputs. Currently, only one future climate is input to the prospective hydrological modelling conducted in the Gloucester subregion. Obviously it is not known what the future climate in the Gloucester subregion will be out to ~2100, yet the output from GCMs can be used to generate 100 or 1000 stochastic/ensemble future climate sequences reflecting variability over different time scales based on characteristics of observed instrumental records that can be used as input to the prospective hydrological modelling conducted in the Gloucester subregion. Doing this will allow the relative impacts of the most likely future coal development and a range of possible future climates to be compared.

2.1.1.2.1 Retrospective climate error analysis

In addition to generating daily and monthly grids of meteorological variables (P, Tmax and Tmin), the Bureau of Meteorology (Jones et al., 2009) also generates daily and monthly RMSE grids of the same variables. These daily and monthly RMSE grids are a combined measure of the observational error and geostatistical error; the latter being a function of the interpolation algorithm, density of isolated station observations and degree of spatial-autocorrelation of the process(es) driving the spatial variance captured in the data being interpolated.

To characterise errors of the input climate data the long-term (from January 1980 to December 2009) monthly mean values for P, Tmax and Tmin were calculated. Also calculated were the long-term monthly RMSE mean values for the same variables for the same time period. Relative error, expressed as a percent, was calculated by dividing monthly RMSE mean grid by the monthly mean grids (i.e. RMSE grid/mean grid for each meteorological variable).

Precipitation error analysis

The spatially-averaged long-term monthly mean P for the Gloucester subregion is 87 mm/month, and the associated P RMSE subregion mean is 40 mm/month, see Figure 3a and Figure 3b, respectively. This results in a relative error of 46% in the input P grids (Figure 3c). The relative high error is due, in part, to P being a highly spatially variable process (it has low spatial autocorrelation) especially in areas such as those proximal to the Gloucester subregion that have high amounts of relative relief. Additionally, in the relative error grid the influence of spatial density of meteorological stations is clearly seen. In the north-east of the Karuah river basin (near the location of Myall Lakes National Park) and in the north-western portion of the Manning river basin (proximal to the Barrington Tops National Park) the higher relative P errors are found due to the low density of observations in these regions.

Figure 3

Figure 3 Spatial variation of precipitation from 1980 to 2009

(a) monthly mean precipitation, (b) monthly mean root mean square error (RMSE) precipitation and (c) monthly mean precipitation relative error for the Gloucester subregion and proximal surface water basins.

Data: CSIRO Land and Water (Dataset 1)

Air temperature error analysis

For air temperatures, a meteorological field that has higher spatial autocorrelation than P, regional distribution is governed by topography and distance from the ocean. The Tmax spatially-averaged long-term monthly mean is 22.8 °C for the Gloucester subregion (Figure 4a). The associated RMSE is approximately 0.47 °C (Figure 4b), which leads to a relative error of approximately 2% for Tmax (Figure 4c). For Tmin in the Gloucester subregion, there are similar spatial patterns, with the spatially-averaged long-term monthly mean being 11.4 °C (Figure 5a) and the associated RMSE being approximately 0.67 °C (Figure 5b), which leads to a relative error of 6% for Tmin (Figure 5c).

Figure 4

Figure 4 Spatial variation of maximum air temperature (Tmax) from 1980 to 2009

(a) monthly mean Tmax, (b) monthly mean root mean square error (RMSE) Tmax and (c) monthly mean Tmax relative error for the Gloucester subregion and proximal surface water basins.

Data: CSIRO Land and Water (Dataset 1)

Figure 5

Figure 5 Spatial variation of minimum air temperature (Tmin) from 1980 to 2009

(a) monthly mean Tmin, (b) monthly mean root mean square error (RMSE) Tmin and (c) monthly mean Tmin relative error for the Gloucester subregion and proximal surface water basins.

Data: CSIRO Land and Water (Dataset 1)

Last updated:
22 January 2019
Thumbnail of the Gloucester subregion

Product Finalisation date

2018