5.4 Guiding principles – Estimation and uncertainty
- Image classification can be by human interpretation, or it can be automated with human interpreters checking the results. The latter can be less resource intensive and may increase consistency.
- The results of using a classification algorithm can be improved by an iterative process involving human interpretation, choice of training data, and use of auxiliary information usually obtained via the NFMS on forest conditions on the ground.
- Reference data should be used with map data to correct for estimated bias and estimate confidence intervals.
- The stratification used for activity data and for estimating emissions and removals factors should be consistent.
- In addition to the general principles of consistent representation of land when using remote sensing for representing land or tracking units of land using a pixel approach, MGD advice is that:
- Once a pixel is included, then it should continue to be tracked for all time. This will prevent the double counting of activities in the inventory and will also make emissions estimates more accurate.
- Stocks may be attributed to pixels, but only change in stocks and consequent emissions and removals are reported with attention to continuity to prevent the risk of estimating large false emissions and removals as land moves between categories.
- Tracking needs to be able to distinguish both land cover changes that are land-use changes, and land cover changes that lead to emissions within a land-use category. This prevents incorrect allocation of lands and incorrect emissions or removals factors or models being applied that could bias results.
- Rules are needed to ensure consistent classification by eliminating oscillation of pixels between land uses when close to the definition limits.
- In addition to classification errors, uncertainties arise from biomass and other sampling used to establish emissions and removals factors and other parameters, and use of default data.
- Combining activity data and emission factor uncertainties to estimate overall uncertainty estimates is possible using repeated application of straightforward rules, or (in the case of more complex modelling approaches) by Monte-Carlo analysis.