This short note is to assist in understanding how the use of land cover classes was used in the NTCSA 2020, and how these relate to the NLC products. Three national land cover products (NLC 1990. 2014 and 2018) were used to track changes over time in land use and the resultant impacts of land use on Soil Organic Carbon (SOC).

South Africa has a number of National Land Cover products, with three, 1990, 2014 and 2018 being designed to be reasonably compatible for use in time series analysis. All three products have about 72 land cover classes, but although these are identical in the 1990 and 2014 product, the 2018 NLC uses a different set of land covers. The classes as used in the NLC are hierarchical and by choosing lower level classes (or by combining classes) it is possible to find what can be considered as the “lowest common denominator” i.e. classes that are relatively consistent between all three land cover products. Further, the numerous sub-classes within the settlement land cover class, are combined the NTCSA 2020 as we currently have no available data on their individual SOC profiles.

The NLC products are extremely accurate in differentiating between some land covers, but very inaccurate when it comes to differentiating between others. Differentiating transformed land cover, such as an agricultural field, from natural vegetation, is something that the NLC products can do very well, both in space and between different time periods. Differentiating between different classes of natural vegetation, for instance between low shrubland, thicket and woodland has proved difficult and has low accuracy. This differentiation between natural vegetation classes has low certainty within a NLC classification (when compared to on the ground control points) but also is very unreliable when comparing between two time periods. Further, there is not consistency of how these classes are defined between the 1990/2014 versus the 2018 classification.

Using the above considerations, 17 land cover classes were decided on for use in the NTCSA based on the “lowest common denominator” approach to find common classes from the 1990, 2014 and 2018 NLC products. Of these only three represent natural vegetation, with the remainder representing transformed land use. The water class being an exception as it covers both natural and man-made water bodies.

The NTCSA uses a wall-to-wall approach to monitoring woodiness (i.e. standing tree biomass and tree canopy cover), and as such, there is no need to attempt to distinguish between different NLC types of natural vegetation, although wetland have been maintained as a separate category. The exception is that closed-canopy indigenous forest (as defined as forest in South African Biome classifications, Mucina and Rutherford 2006) has been maintained as a land cover class of its own since it represents a biome. All other land cover classes representing natural land are analysed as a single land cover class, but stratified based on biomes.

Bare ground is an exception as it is seen as a degradation of natural vegetation rather than a transformation. However, bare ground (as identified on satellite imagery) is natural in very arid biomes such as the desert and karoo. NTCSA2020 has regarded bare ground as a degradation in all biomes except in the Desert and Karoo where it is treated as natural vegetation.

Land use change is used as the key driver of organic carbon stock (SOC) change in the NTCSSA 2020. A secondary loss of organic carbon can occur through degradation of vegetation which is not visible on the national land cover products.

The reference land cover is a hypothetical land cover that we would expect if no human induced land use change has taken place. In other words, it is the natural vegetation that would be expected in the area. The reference land cover has no specific date but is assumed to be the land cover that existed at some point in the past before man made transformations took place. In the NTCSA the reference landcover is assumed to be the “other natural vegetation” land cover class unless the area is “indigenous forest” or “wetland”. Although both natural indigenous forest and wetlands may have changed over time, there is no historic data showing their spatial extent, and therefore current extent is assumed.

The soil organic carbon (SOC) reference value is the SOC that is anticipated within the natural vegetation. The SOC data used in NTCSA2020 is the global SoilGrids dataset, curated out of ISRIC/World Soil Resources in Wageningen (ISRIC). ISRIC data was derived used a modelling approach to extrapolate SOC based on soil pit samples and climatic and terrain features. It provides a 250m resolution estimate of the reference soil carbon stock for the entire country.

Changes in land cover (for instance from a natural grassland to an agricultural field) results in changes in SOC. In most cases a change from the natural vegetation to transformed land results in the loss of SOC. The amount of SOC loss is dependent on the vegetation type of the natural vegetation, the soil type, the nature of the land cover change and climatic variables.

The NTCSA2020 did an extensive review of available South African data on levels of change in SOC as a consequence of land cover change. Based on this, change factors were identified for different land cover changes. The country was divided into areas of similar rainfall and vegetation (i.e. by biome) with change factors identified for each rainfall class per vegetation type.

The NLC 1990, 2014 and 2018 land cover products were used to estimate SOC loss compared to the baseline. This was calculated for ever 1km2 land unit of South Africa. This resulted in four sets of national level SOC data, a reference value, a 1990 value, a 2014 value and a 2018 value.

The difference in SOC from the reference until 1990 represents the total loss in SOC that had happened before 1990. This loss will have happened over a few hundred years as agriculture and settlement expansion took place. Most of this loss is found in agricultural fields, some of which have been abandoned. Most agricultural expansion in South Africa took place prior to 1990.

Loss (or gain) between 1990 and 2014 can be calculated by subtracting the 2014 data from the 1990 data. Since this loss took place over 24 years it is also possible to find the mean yearly rate of loss. Similarly loss between 2014 and 2018 or between 1990 and 2018 can also be calculated.

The 2018 land cover data has a unique land cover class not found in earlier land cover datasets, this is the class of “fallow land”. Fallow land is land that was previously agricultural land and as such would have lost a large proportion of its SOC. Some of this carbon may have been regained since agriculture ceased. The dynamics of carbon in fallow land is poorly researched in South Africa, but for the NTCSA we have assumed that fallow land has not regained all its lost carbon. What this means is that the 2018 data will show a large loss of carbon from fallow land, but this loss is not reflected in the 1990 or 2014 data, despite the fact that the fallow land may well have already existed during these time periods.

From the NTCSA 2020 it is clear that the single biggest loss of terrestrial environmental organic carbon in the South African context relates to loss of SOC in dryland crop agriculture. As such, restoration of SOC in agricultural fields (including abandoned agricultural fields), represents a huge mitigation opportunity. The full extent to which carbon can be reclaimed through processes such as conservation agriculture is still unclear, so this district level assessment considers gains that could potentially be made using three scenarios, a 25, 50 and 75% recovery of the lost SOC as a consequence of the introduction of sustainable agriculture. It further makes the assumption that these gains would require about 20 years.