Heranga ā Nuku is responsible for identifying, collating and distributing Public Access Areas data to NZ. The main input to this data is the LINZ cadastre which holds a wealth of information within its data model.
However, the ‘devil is in the detail’ and extracting certainty from the often complex and legally nuanced wording within the text fields ultimately requires specialist manual interpretation, in our case coming from our excellent technical operations team.
The strategic direction at Herenga ā Nuku like many others is automation, constant data improvement and 'value add'. With a small team we need to use whatever tools are available to us to identify public access from the cadastre.
___________
Sharon Xia has been investigating the use of NLP (natural language processing technique) to identify different public access types from the LINZ cadastre dataset.
Results have been encouraging. Using the default language model on the test data we achieved an accuracy of nearly 90%.
Subsequent training the model could achieve an accuracy of nearly 98%.
The importance of decent sized training datasets cannot be understated in the future.

Thanks to this work we can see what role NLP can play in our data landscape. Language models can be used to refine uncertainty or to check the overall quality of our data.
____________
Herenga ā Nuku’s work relies on the professional interpretation of cadastral nuance. While these models will not be able to replace that value, they can help to prioritise workload for the operations team, to improve the efficiency and to reduce the uncertainty. Spending taxpayers’ money wisely is what we are ultimately about.

We addressed the issue of identifying obstacles that penetrate the flight path surface, primarily focusing on trees and vegetation. Additionally, we assessed potential crane locations to determine if they would intrude on the flight path surface.

GIS provided a comprehensive set of tools for spatial and terrain analysis, enabling precise identification and measurement of obstacles relative to the flight path surface. It facilitated efficient data processing, visualisation, and the creation of customised obstacle limitation surfaces (OLS).

We utilised several types of geospatial analysis:
• LiDAR Analysis: To accurately measure the height and elevation of obstacles.
• Terrain Analysis: To understand the topography and its impact on flight paths.
• Spatial Analysis: To evaluate and map the locations of potential obstacles and crane placements.

We employed Esri Aviation tools to create the OLS and perform runway obstacle analysis. These tools were essential for generating accurate representations of the flight path surface and identifying intrusions.

We discovered that the default configuration of the Aviation tools was not suitable for curved surfaces. After extensive customisation, we successfully configured the OLS tool to work with our specific requirements. This enabled us to tag all obstacles penetrating the OLS, primarily trees, for removal. We also identified and marked potential future risks.

This project is particularly valuable for the aviation sector, as it ensures safe flight paths by identifying and mitigating obstacles. The spatial analysis techniques we employed can also be beneficial to other users needing precise geospatial data analysis for various applications, such as urban planning, environmental management, and construction.

The details reported by commercial fishers are both complex and spatially coarse which creates limitations when performing spatial analysis. A common analysis task is to measure the quantity of fish caught inside a proposed fishing closure.
The Spatial Intelligence team at the Ministry for Primary Industries (MPI) looked at the available data and developed a process to enrich the reported point locations with vessel-based track lines, these track lines were intended to better understand where the fish was being caught.
Early proofs of concepts demonstrated the benefit of the enriched track-line data, and for some fisheries these track-lines became the only viable option for mapping. However, the processing needed to create the track-lines was incredibly slow and reliant on specialist knowledge which limited how the enriched data could be used.
We recently rebuilt our track-line processing scripts, mainly to remove most of the manual data handling factors, but also with a hope of improved performance. We decided to use pandas to streamline the large amount of data grooming and soon discovered this approach offered many more improvements to our old arcpy cursor approach. A bonus discovery was when we found the ArcGIS API provided smooth interoperability between pandas and arcpy.
With the new tool we can process millions of records in around 45 minutes, previously it took over 30 days to perform the same task. To put it lightly, the outputs from this new tool are changing how MPI analyses the spatial components of New Zealand’s commercial fishing data.
We have a large list of lessons learnt from this project and we’d also like to highlight what we think is a little-known aspect of the ArcGIS API (the smooth interoperability of arcpy and pandas).