This study investigates the effect of changes in map scale on the error in the development of areal map units and their associated erosion severity measurements of New Zealand’s (NZ) Land Use Capability (LUC) surveying system. A map scale of 1:50,000 was used in the underlying data (i.e., a LUC survey) of an Erosion Susceptibility Classification (ESC) system, which was developed by Bloomberg and others (2011) of the University of Canterbury for the Ministry for the Environment’s (MFE) 2010 proposed National Environmental Standard for Plantation Forestry. The ESC was intended for local erosion management decisions, yet most literature would classify the map scale of 1:50,000 as more appropriate for regional management issues. Thus, this study will test two finer 1:10,000 scale datasets against the current 1:50,000 national LUC areal map units and their erosion severity measurements of the underlying data for the ESC system, to quantify the level of agreement.

This study first attempted to identify a unique discriminating parameter of high erosion severity. A case study was conducted in the Sherry River catchment, located in the Tasman District of the South Island, NZ. The Sherry River Case Study had two aims; the first was to investigate the correlation between the Melton ratio and LUC erosion severity. This was accomplished by calculating the Melton ratio, a tested morphometric factor that describes basin (watershed) ruggedness, using Irvine’s (2011) Geographic Information Systems (GIS) debris-flow model. The product of this GIS debris-flow model, a calculated Melton ratio ≥ 0.50 with the areal extent outlined by a River Environment Classification (REC) order one polygon, were designated the areas of interest (AOIs). The Melton ratio was then tested against LUC erosion severity using the Spearman’s Ranked Correlation Coefficient, within the designated AOIs. A field investigation was conducted to verify debris-flow in GIS identified AOIs. Only five of the thirteen AOIs identified showed evidence of debris-flow. Two were un-checked due to accessibility and the others had a high degree of fluvial activity, which indicates a high probability that surface evidence of alluvial erosion deposition was erased. Nominal association between the two measurements of erosion (Melton ratio and LUC erosion severity) was found at the map scales of 1:50,000 or 1:10,000. Therefore the Melton ratio was not recommended as an independent parameter of erosion severity.

The second aim of the Sherry River Catchment study was to assess the sensitivity of empirically generalised LUC areal map units and their erosion severity measurements to spatial resolution, that is, what is the effect of agreement between the smallest measurable value when looking at LUC map units and their erosion severity measurements recorded at two different map scales. A hard classification accuracy assessment was chosen to accomplish this objective. An accuracy assessment is a statistical model, which provides a probability of error (uncertainty), in essence a goodness-of-fit measurement, and quantified the agreement between a sample and reference dataset. This was accomplished by the calculation of an Overall accuracy (i.e., overall thematic agreement), Producer’s accuracy, and a User’s accuracy analytical statistics. The Producer’s accuracy refers to the probability that an area of sampled erosion severity category in the sample map is classified as such according to the reference map, while the user’s accuracy refers to the probability that a point labelled as a certain erosion severity in the sample map has that severity rating in reality (i.e., according to the reference map). An accuracy assessment also includes a second goodness-of-fit test, the Kappa statistic (K ̂), which measures the agreement between the sample and references map as well as chance agreement. An accuracy assessment of the AOIs within the Sherry Catchment Study area using an 85% significance criterion was conducted. This accuracy assessment investigated a sample LUC survey measured at the map scale of 1:10,000, as compared to the referenced underlying data of the ESC (1:50,000 map scale). Overall accuracy was marginal (69%) with equally marginal levels of Producer’s and User’s accuracy. The Kappa statistic showed a marginal level of significance according to Landis and Koch (1977) (K ̂ = 44%). The disagreement seen between the two LUC surveys, which were empirically developed using different map scales, provides evidence of high spatial resolution sensitivity, when comparing areal map units and erosion severity measurements.

To further investigate evidence of spatial resolution sensitivity in LUC surveying, a second case study was conducted using a LUC survey across a broad geographical area of the Manawatu-Wanganui Region of the North Island, NZ. A sample dataset from the LUC survey, empirically generalised at 1:10,000 map scale by the Horizons Regional Council, was compared to the referenced underlying data of the ESC. There was a moderately-strong consistency found between the assessors of each LUC survey using Spearman’s Ranked Correlation Coefficient. This provides evidence of limited surveyor bias, as each map was made using empirical judgment. The accuracy assessment’s overall agreement was 63% and as for the previous case study, had equally low Producer’s and User’s accuracy levels. The Kappa statistic for this case study was K ̂= 46%, a moderate chance agreement. This evidence, along with the evidence provided by the Sherry River Catchment Case study, suggested that the MFE’s ESC system is sensitive to changes in map scale and that any decision based on it will have different results when its underlying data is produced at different spatial resolutions. It is therefore recommended that MFE reassess the map scales and resolutions of its underlying data, given that the ESC’s purpose is for local level environmental management, before imposing the system as a regulatory requirement in the National Environmental Standards for Plantation Forestry.