Citation

Geoscience Australia provides most of its products for free under a Creative Commons Attribution 3.0 Australia Licence. We only require that you reference the use of our data or information using the following citation:
Jiang, W. & Li, J., 2014. The Effects of Spatial Reference Systems on the Predictive Accuracy of Spatial Interpolation Methods. Record  2014/001. Geoscience Australia, Canberra. http://dx.doi.org/10.11636/Record.2014.001

Abstract

Geoscience Australia has been deriving raster sediment datasets for the continental Australian Exclusive Economic Zone (AEEZ) using existing marine samples collected by Geoscience Australia and external organisations. Since seabed sediment data are collected at sparsely and unevenly distributed locations, spatial interpolation methods become essential tools for generating spatially continuous information. Previous studies have examined a number of factors that affect the performance of spatial interpolation methods. These factors include sample density, data variation, sampling design, spatial distribution of samples, data quality, correlation of primary and secondary variables, and interaction among some of these factors. Apart from these factors, a spatial reference system used to define sample locations is potentially another factor and is worth investigating. In this study, we aim to examine the degree to which spatial reference systems can affect the predictive accuracy of spatial interpolation methods in predicting marine environmental variables in the continental AEEZ. Firstly, we reviewed spatial reference systems including geographic coordinate systems and projected coordinate systems/map projections, with particular attention paid to map projection classification, distortion and selection schemes; secondly, we selected eight systems that are suitable for the spatial prediction of marine environmental data in the continental AEEZ. These systems include two geographic coordinate systems (WGS84 and GDA94) and six map projections (Lambert Equal-area Azimuthal, Equidistant Azimuthal, Stereographic Conformal Azimuthal, Albers Equal-Area Conic, Equidistant Conic and Lambert Conformal Conic); thirdly, we applied two most commonly used spatial interpolation methods, i.e. inverse distance squared (IDS) and ordinary kriging (OK) to a marine dataset projected using the eight systems. The accuracy of the methods was assessed using leave-one-out cross validation in terms of their predictive errors and, visualization of prediction maps. The difference in the predictive errors between WGS84 and the map projections were compared using paired Mann-Whitney test for both IDW and OK. The data manipulation and modelling work were implemented in ArcGIS and R. The result from this study confirms that the little shift caused by the tectonic movement between WGS84 and GDA94 does not affect the accuracy of the spatial interpolation methods examined (IDS and OK). With respect to whether the unit difference in geographical coordinates or distortions introduced by map projections has more effect on the performance of the spatial interpolation methods, the result shows that the accuracies of the spatial interpolation methods in predicting seabed sediment data in the SW region of AEEZ are similar and the differences are considered negligible, both in terms of predictive errors and prediction map visualisations. Among the six map projections, the slightly better prediction performance from Lambert Equal-Area Azimuthal and Equidistant Azimuthal projections for both IDS and OK indicates that Equal-Area and Equidistant projections with Azimuthal surfaces are more suitable than other projections for spatial predictions of seabed sediment data in the SW region of AEEZ. The outcomes of this study have significant implications for spatial predictions in environmental science. Future spatial prediction work using a data density greater than that in this study may use data based on WGS84 directly and may not have to project the data using certain spatial reference systems. The findings are applicable to spatial predictions of both marine and terrestrial environmental variables.
Google map showing geographic bounding box with values North bound -10.0 East bound 165.0 West bound 108.0 South bound -50.0
Downloads
Related Links
For information on acquiring this product,
please contact Geoscience Australia Client Services via:

fax:
+61 2 6249 9960; or
phone:
1800 800 173 (within Australia);
 
+61 2 6249 9966 (outside Australia).

Product Type/Sub Type

dataset - GA Publication - Record

Constraints

license
Creative Commons Attribution 3.0 Australia Licence

IP Owner

Commonwealth of Australia (Geoscience Australia)

Author(s)

Date (publication)

2014

Product Type

dataset

Topic Category

geoscientificInformation

Keywords

GA Publication
Record
Earth Sciences

Resource Language

English

Resource Character Set

utf8

Resource Security Classification

unclassified

Geographic Extent

North bound
-10.0
East bound
165.0
West bound
108.0
South bound
-50.0

Lineage

Unknown

Digital Transfer Options

onLine

DISTRIBUTION Format

docx
pdf

Distributor

Role
distributor
Organisation Name
Geoscience Australia
City
Canberra
Administrative Area
ACT
Postal Code
2601
Country
Australia
Email Address

Metadata File Identifier

ddc00860-622a-5a26-e044-00144fdd4fa6

Metadata Standard Name

ANZLIC Metadata Profile: An Australian/New Zealand Profile of AS/NZS ISO 19115:2005, Geographic information - Metadata

Metadata Standard Version

1.1

Metadata Date Stamp

2014-02-14

METADATA SECURITY CLASSIFICATION

unclassified

Metadata Contact

Role
pointOfContact
Organisation Name
Geoscience Australia
City
Canberra
Administrative Area
ACT
Postal Code
2601
Country
Australia
Email Address
Downloads
Related Links
For information on acquiring this product,
please contact Geoscience Australia Client Services via:

fax:
+61 2 6249 9960; or
phone:
1800 800 173 (within Australia);
 
+61 2 6249 9966 (outside Australia).