Land Degradation Monitoring with Global and National Datasets - Ukrainian Use-case

Land Degradation Neutrality is one of the goals selected by UN Convention to Combat Desertification. Within the Sendai Framework corresponding SDG is defined - 15.3.1 "Proportion of land that is degraded over total land area". The indicator is based on "one out - all out rule" and relay on 3 sub-indicators (land cover (LC), land productivity dynamics (LPD ) and soil organic carbon stock (SOC)).

For the territory of Ukraine several datasets are considered: (i) annual global LC data derived from the ESA CCI-LC (300 m spatial resolution), National LC data are provided by Space Research Institute of Ukraine (SRI) [1]. LPD data are derived from JRC productivity map; SOC data are extracted from SoilGrids system.

Performed analysis showed that global datasets do not correspond to national statistics and surveys. Global LC highly overestimates cropland areas - actually it seems to be a mixture between grassland and cropland areas. Due to low spatial resolution it's not possible to capture transitions of artificial objects and bare land class to any other in reliable way.

At the moment SRI has 10 m resolution maps for Ukraine (2016, 2017) and 30 m historical maps (1990, 2000, 2010). SRI plan to continue this activity in coming years.

According to global LDP dataset main agricultural regions have range of negative productivity trends. SRI plans to perform activities on estimation productivity over cropland areas within ERA-PLANET Geo-Essential project.

Global SOC stock product highly overestimates SOC over the territory of Ukraine. Comparison with results derived from National Scientific Center «Institute of Soil Science and Agrochemistry Research named after O.N.Sokolovsky» for 2015 shows that for some LC classes difference could reach 2 times.

Total area of degraded land according to the methodology of 15.3.1 indicator calculation (based on global data) is 150520 km² that results to 0.25 of the territory of Ukraine that could be considered as degraded in 2000-2015 interval.

However, for next reporting period on LDN it'll be reasonable to use national datasets instead of global.

References

[1] N. Kussul, M. Lavreniuk, S. Skakun, and A. Shelestov. 2017b "Deep Learning Classification of Land Cover and Crop Types Using Remote Sensing Data," IEEE Geoscience and Remote Sensing Letters, vol. 14, no. 5, pp. 778-782.