S-map confidence measures

Inherent in the design of S-map is provision of estimates of thickness of soil layers and of their physical characteristics, together with estimates of the proportion of siblings (soil types) that occupy map units. Provision of this quantitative data is challenging because soils are highly variable in both the horizontal and vertical dimensions. Accuracy of information within a map unit is therefore dependent on the amount of site data available for any map unit.  While all reasonable skill and care have been exercised in the collection and preparation of this data, the Data Providers can give no warranty that the data supplied are free from errors, omissions or other inaccuracies. 

A general indication of accuracy may be found in the confidence rating assigned to map unit composition.  Each sibling with a map unit is assigned a confidence rating of High (H), Medium (M) or Low (L).

  • The H and M ratings relate to confidence in the proportional value of the soil sibling (or functionally similar siblings) occurring within the map unit. 
  • For a High confidence map unit with a single sibling, the sibling should be found in at least 80% of the map unit (i.e. 100±20), as opposed to at least 60% (i.e. 100±40) in a Medium confidence map unit.
  • For a double symbol map unit it is likely that areas for both siblings will be within ±10 in a High confidence map unit, and ±20 in a Medium confidence map unit.
  • A rating of L (Low) is used where the pedologist has less confidence in the identified sibling or its proportion due to lack of data or insufficient resources to verify the data.  This will usually occur where the sibling(s) have been assigned based on a soil-landscape relationship and aerial photo interpretation, rather than a site visit with direct observation (auger holes or soil profile descriptions).  

Map units have been mapped on a topographical base (Topo50 series, NZ 260 series or NZMS 1) and are designed to be used at scales between 1:50000 and 1:20000.

Users should consider both the scale at which they are applying the data and the map unit confidence code, and take responsibility for the impact these may have on any decisions arising from their data use. 

Manaaki Whenua shall not be liable on any legal basis (including without limitation negligence) and expressly excludes all liability for loss or damage howsoever and whenever caused to a user of the maps, fact sheets or other information provided by this web site.


Hydrological soil properties model

The new S-map hydrological model estimates the soil water content (in units of water volume per unit volume of soil, or cm3/cm3) for pre-defined tensions (0, 5, 10, 20, 40, 100, and 1500 kPa). The model uses a range of available information from S-map, including soil order, rock class, functional horizon characteristics and depth as well as the estimated texture (percentage of sand, silt, and clay). It was developed using a beta regression approach and relies on 4641 laboratory measurements (on 684 sites) from the National Soils Database Repository. More detailed information about the new soil hydrological  model is provided here.


pH prediction model

The values of this layer represent estimates for the centre of a 100 m x 100 m (1 ha) grid covering all of New Zealand, with the exception of  areas that do not have soils. The layer was generated using a 3 dimensional quantile regression forest model, that leverages a range of spatial predictors, including land use, topography and climate layers.

More detailed information about the model used is provided in the publication: National Scale 3D Mapping of Soil pH Using a Data Augmentation Approach, Remote Sens. 2020, 12(18), 2872; https://doi.org/10.3390/rs12182872


P-retention prediction model

P-retention is an indication of the ability of soil material to hold on to phosphate ions. Phosphate ions that are retained by soil material are called "immobilised" and their availability for plant uptake may be limited. Conversely, phosphate ions that are not retained by the soil material are "mobile"—they can more easily enter the soil water solution and thus be available for plant uptake. A high value of P-retention indicates that most of the phosphorus in a soil is immobile, and vice versa.

Saunders (1965) indicates that in New Zealand soils, the degree of P-retention is related to the nature of the soil parent material and to the degree of weathering of the soil material. P-retention is generally lowest in weakly weathered soils, and generally increases with the degree of weathering—though the most highly weathered soils may not have the highest P-retention. Variation of P-retention within different types of soil is generally greater in soils formed under high rainfall, and that have good drainage.

Estimates in S-map are only valid for the 0 to 7.5 cm depth interval (i.e. the near-surface soil material) and should not be used to infer the P-retention at depth. Being model predictions, the estimates are subject to uncertainty, and the degree of uncertainty varies across the map. The degree of uncertainty is due among other factors to the intrinsic variability of soils and landscapes in New Zealand, the spatial distribution of available observations of P-retention in New Zealand, and the ability of the model to capture relationships between P-retention and environmental variables. Areas with lower model uncertainty include the east coast south of the Hawkes Bay on182872 the North Island, and the Canterbury plains and central Otago basins on the South Island. Areas with highest uncertainty include parts of the central plateau on the North Island, and at high elevations along the Southern Alps on the South Island.

Saunders WMH 1965. Phosphate retention by New Zealand soils and its relationship to free sesquioxides, organic matter, and other soil properties. New Zealand Journal of Agricultural Research 8(1):30–57. doi:10.1080/00288233.1965.10420021


 Spatial soil carbon prediction model

A number of data layers were acquired to act as explanatory layers for the prediction of soil carbon. Soil data includes the MfE Historic Soils Database (under an agreement with MfE), the NSD (LandcareResearch New Zealand 2012), the SINDI soil quality dataset (Landcare Research New Zealand 2012), and the LMI soils dataset (under an agreement with Plant and Food Research 2011).Additional explanatory data  layers  were obtained from the LENZ data  set (Landcare  Research New Zealand), including the LENZ climate layers  (e.g. mean  annual rainfall) as  well as  environmental classification layers (LENZ level 1, 2, 3, and 4), from the Koordinates data portal (Koordinates 2012). In addition, the natural potential vegetation layer (Leathwick 2001) is  used as an indicator of vegetation prior to agricultural development.  These layers were augmented  by explanatory layers acquired from the LRIS portal (Landcare Research 2012), consisting of national maps of basic soil properties (soil order,  exchangeable  calcium,  acid soluble phosphorus, rock  class, mid-estimate of surface outcrops, annual water deficit). Additional 0–30cm carbon stock data was obtained by using SINDI data (derived from 0–10cm depth samples) along with NSD data and a regression model to infer 0–30cm soil carbon stocks (tonnes/ha) at the centre of each pixel.

 

Last updated: 24 August 2020