|Page tools: Print Page Print All RSS Search this Product|
Land Management Practices Survey Design
This year the Rural, Environment and Agricultural Statistics Branch (REASB) is conducting the Land Management Practices Survey (LAMPS). This is a user-funded survey coming under the umbrella of the Federal Government's Carbon Farming Initiative. It will collect information on agricultural practices that relate to CO2 abatement. The survey is scheduled to be repeated in 2014 and 2016. Outputs from the present survey will assist in defining what is current "common practice" as well as providing benchmarks for measuring changes in farming practices in response to government climate change initiatives.
The requirements of this survey differ from those of the annual Agricultural Commodity / Agricultural Resource Management survey in significant ways and REASB considered that a separate survey was needed in order to satisfy these requirements. These differences also made the design a challenge. One of the differences is that many of the data items collected are new and so there is little historical data available for key items of interest. The design was therefore done using nine items from among the data regularly collected in agricultural surveys which were considered close "proxies" for the items of interest.
A second challenging feature was the output requirements. Agricultural survey outputs are generally produced for nation, state and sub-state national resource management regions. For LAMPS, outputs are required at national and state levels and for agro-climate regions, with estimates for the latter being a priority. These regions pose a difficulty for stratification in that they may cross state boundaries and in some cases consist of non-contiguous sub-regions. Consequently, the geography component of the stratification was chosen to be state x agro-climate region so that estimates of acceptable quality could be produced for both of these geographies.
At each geographic level outputs were required for a mixture of ANZSIC classes and groups of classes, 26 in all. Combined with the number of design variables, this led to a number of constraints so large (6,318) that it was impossible to run the allocation code for the entire set. The first task was therefore to try to reduce this constraint set. This was done by eliminating constraints for which there was insufficient historical data. In addition constraints for ANZSIC class x agro-climate region cells were eliminated if the expected cell estimate was only a small proportion of the estimate at higher output levels, e.g. all of the Agriculture division for the particular region, or nationally for the particular ANZSIC class. The final constraint set had 1,380 constraints which is comparable with the number in the annual Agricultural Commodity / Agricultural Resource Management survey.
The survey is currently in the field and methodology work is now focussed on supporting estimation for this survey and this year's Agricultural Resource Management survey (ARMS). These two surveys cover the same reference period and also have in common a number of data items which are conceptually the same. It was considered necessary to have these items in the LAMPS survey because they were intrinsic to the objectives of this survey and also because estimates of these items needed to be consistent with those of other items collected. However, this presents REASB with the challenge of meeting user expectations that two estimates of the same quantity for the same reference period should be the same even if produced from different surveys. One way of bringing estimates from the two surveys into closer alignment is to calibrate estimates from one against those produced by the other. Since LAMPS has a significantly larger sample size (50,000) than ARMS (37,000) it is planned to calibrate ARMS estimates of business counts and total area of holding with those from LAMPS at the state level. This will be supplemented by other strategies still under discussion.
These documents will be presented in a new window.