• About this site|
• Site index|
• Contact us|
• Jobs

Framework Report on the Development of the National Carbon Monitoring System

Contents

1. Executive Summary

2. Climate Change and Greenhouse Gas Mitigation

3. Developing the 1990 Baseline for Soils

4. A National Monitoring System For Soil Carbon

5 Development of Baseline 1990 Estimates for Forest and Scrub

7 An Information Management System for the CMS

8 International Review of the CMS Development

9 Response to the Review

10 Application of the Generalised Linear Model to Soils Data (Giltrap et al. 2001)

12 Potential Other End-User Requirements

13 Integration of the LCDB with the CMS (Stephens et al. 2001)

14 Improved Allometric Functions for Forest and Scrubland (Beets et al. 2001)

15 Forest Sampling Options for the CMS

16 Costs of a Proposed Structure

17 A Proposed Structure for the CMS

18 Concluding Remarks

20 APPENDICES

10 Application of the Generalised Linear Model to Soils Data (Giltrap et al. 2001)

10.1 Introduction

The Generalised Linear Model (GLM) analysis was first applied to the soils data in mid-2000 and was re-applied in 2001 after

• revising the database to exclude sites that had not been in their current land use for long enough to approximate equilibrium and to replace close geographic clusters of sites with single composite points where appropriate

• using a digital elevation model (DEM) interpolated from 20 m contours to measure the nationwide distribution of slope, aspect and other topoclimatic variables and to compare this with their distribution in the database sites.

The analyses were extended by

• including the five depth ranges 0–10 cm, 10–30 cm, 0–30 cm, 30–100 cm and 0–100 cm rather than just 0–10 cm and 0–30cm;

• using categories derived from the LCDB as an alternative to the land-use classes derived from the Vegetation Cover Map (VCM);

• deriving estimates of the total national soil C inventory (at each of the five depth ranges) from the GLM analyses

• inspecting the individual model predictions for Soil-Climate/Land-Use cells and comparing them with the actual cell averages for the 30–100 cm depths (the Soil-Climate by land-use interaction was not significant for soil C at shallower depths than 30 cm).

Results of the analysis

• The database revision led to a substantial reduction in the number of exotic forest sites.  This was partly due to the exclusion of recently planted forests but mainly to the amalgamation of clustered sites into composites.  There was a much smaller reduction in the number of sites in other land-use categories and a substantial increase in the number of arable sites as a result of reclassification of the land use on sites in pasture/arable rotations.

• A simple additive model using soil-climate category, land-use (or LCDB) category and the product of annual rainfall and slope gave the best prediction of 0–10 cm, 10–30 cm and 0–30 cm soil C.  This model means that a single "land-use effect" can be applied across all soil-climate categories.

• the model for these depths worked equally well for LCDB or land-use categories and was not particularly sensitive to the inclusion or exclusion of predictor variables other than soil-climate and land-use (or LCDB).

• The 0–30 cm land-use (or LCDB) effects show only negligible differences between pasture (grassland) categories, substantial reductions (-30 to -50 t/ha) for the forest categories and intermediate reductions (-14–20 t/ha) for arable, horticulture and scrub.  These values are not materially different to those given in Report XII.

• The corresponding land-use (or LCDB) effects for 0–10 cm soil C show a material reduction in the effect for exotic forestry  (-5 t/ha) compared with that in Report XII (-21 t/ha).  This is probably because of the exclusion of "young" or otherwise unsuitable sites.  The results for other land uses are comparable to those in report XII.

• The analyses for 30–100 cm (0–100 cm) soil C was unstable and dominated by the influence of very high "inert" organic matter at depth in podzols.  It may be possible to produce a stable model if podzols are excluded, but this will also exclude most existing indigenous forest sites.  The soil-climate by land-use interaction term was significant for the 30–100 cm and 0–100 cm depths but not for 0–10 cm, 10–30 cm or 0–30 cm.

• The estimates of the national soil inventory from the GLM are 1131 Mt (0–10 cm), 1296 Mt (10–30 cm) and 1304 Mt (30–100 cm).  Compared with earlier estimates, this shows close agreement for the 0–10 cm, a modest (8%) reduction for 10–30 cm and a substantial (20%) reduction for 0–100 cm.  The reduction for the 0–30 cm soil C was due partly to the database revisions and partly to the way soil C was estimated for unsampled cells.  Using the additive model instead of individual cell averages for sampled cells did not contribute to the observed reduction.

Influence of Land Use History on Coefficients of Change

Of the 455 sites in the database, 37% recorded no history of land-use change, 9% recorded a change, and 54% recorded no data about land-use change.  Because the Landcare Research National Soils Database sites were normally placed on undisturbed and stable sites, it is assumed that for the 54% of sites with no land-use history data, no change in land use in the 10 years before sampling has occurred.  Of the 9% of sites that recorded a change in land use within 10 years of sampling, 19 sites were identified to be removed from the database and 20sites were recommended for reclassification of their land use.  Soil C estimates made before and after deletion or reclassification of these sites showed no difference.

[ Previous| Next