Climate-resilient rural livelihoods in Mongolia
For the last decades, grasslands in Mongolia are being degraded, with studies pointing to around 78% of national territory experiencing desertification and other forms of land degradation, resulting in decreased livestock productivity and lower incomes.
Evidence seems to show that most of the degradation in Mongolian grassland steppes between 1988 and 2008 was due to climate change trends. In fact, during the last 20 years, 60% of the decrease in the vegetation indicators average values is correlated to precipitation and temperature trends, with the remaining part correlated to overgrazing and biomass burning.
Mitigation or adaptation to this reality requires optimising pasture management through e.g. a) rotational use of pastures; b) improving water supply; c) increasing hay harvest with irrigation and fertiliser; d) planting forage species. The ADB Establishment of Climate-Resilient Rural Livelihoods project carried out design and implementation of these type of actions within a collaborative pasture management framework in the territory of Buutsagaan, Khüreemaral and Zag districts (soums) of Bayankhongor province (aimag). To monitor pasture changes within these regions, studies were performed in various trial plots in 2013 and 2015. Positive changes were reported in vegetation structure, species compositions and yield as result of resting and rotational use of summer pasture. Results of the social surveys also highlighted the positive outputs of the use of collaborative management.
The ESA support project aimed to provide EO-based information to complement these results and demonstrate the usefulness of EO-based products to support establishing pasture management plans. The project also aimed at enriching the existing environmental database maintained by the National Remote Sensing Center (NRSC) and providing relevant EO traning and capacity building. In particular, EO-based products were planned to be feeding into the Mongolian Environmental Information Center (EIC), integrating several low- and medium-resolution EO products and geographical datasets on atmosphere, land, disaster risk, etc.
Land cover / land use classification and its changes over time
This service provided baseline land use and land cover (LULC) maps for September 2013 and September 2014 using Landsat-7 and Landsat-8 data as input (30 m resolution) for the soums of Buutsagaan, Zag and Khüreemaral. The classification focussed on pasture classes for both the summer and the winter seasons to be used as reference to measure the impact of the aforementioned ADB project. The month of September was chosen to minimise radiometric and seasonal variability and thus improve land classification accuracy and to avoid higher cloudiness during other months. Relevant land use changes between the two seasons were also mapped.
The classification followed as much as possible the classes used in the ADB project, even though the degree of pasture class discrimination proved too detailed to implement using only satellite imagery and few appropriate training datasets. Nevertheless, the produced mapping products clearly complemented Mongolia’s map portfolio for the respective geographical areas, which already included the national LULC maps hosted by the EIC and the local/regional pasture maps provided by ADB-funded project. The delivered maps established a higher-resolution baseline, able to produce, for several spatial scales (national to local), accurate, consistent and continuous LULC information, with very little in-situ and ancillary data. The maps also provide valuable information regarding soil degradation trends, and livestock forage availability and trends (e.g. possible grasslands for grazing reserves) and can be used to assess the effectiveness of pasture management practices. For instance soil degradation can be indirectly monitored by identifying bare soils or loose of vegetation. Livestock forage can be retrieved by monitoring and mapping shrub and grassland areas.
At the level of the three districts, LULC changes for 2013–2014 revealed pasture gains of 4% (493 km2) and pasture losses of 5% (610 km2). Unfortunately high cloud cover hampered the derivation of a LULC map for 2012, the year most pasture management actions started as a result of the ADB project. Users were satisfied with the delivered products and could clearly see the benefit of EO-based LULC products when compared to field surveys. Nevertheless they emphasized the importance of increasing pasture classification accuracy and discrimination.
The drought monitoring service aimed at enhancing the current NRSC monitoring scheme based on MODIS data, by introducing higher-resolution data to monitor vegetation productivity and condition in the Bayankhongor province. The mapping products contain relative information on vegetation status and trends via NDVI and not absolute values of biomass or productivity variables. Nonetheless, these values are a good proxy for many of the biophysical features of grasslands and are not subject to the same error budget as maps providing absolute information.
The introduction of a 22 m resolution DEIMOS-1/DMCii NDVI time series (one image approximately every 15 days) and respective statistics for the year 2014 was an effective improvement of the currently-available information on pasture vegetation status and condition for a full growth cycle (April-October 2014). The dataset provided local users with a higher-resolution (22 m) characterisation of the pasture growth cycle, and consequently a more detailed differentiation of pasture growth regions and more detailed information to monitor the effect of pasture management actions. For instance, grazing reserves for the different stages of pasture growth could be defined more precisely.
Drought monitoring for 2014 was performed by calculating NDVI anomalies using monthly MODIS NDVI products from that year and their long-term averages (2001–2013). A sustained operational delivery of the higherresolution NDVI maps could result in image time series that retrieve the normal (average) growing conditions, thereby identifying even more detailed growth anomalies related to droughts, diseases and human actions. By using the MODIS-based NDVI maps, although no effective improvement on drought monitoring resolution was made, an alternative methodology was introduced to monitor drought in the Mongolian territory, offering a less data intensive and more accurate and easily comparable option to the currentlyused MODIS drought index. Results for 2014 showed that, although some regions within the Bayankhongor province presented both negative and positive anomalies, no significant and consistent drought signal was found. This was also confirmed by a correlation study done with satellite-based precipitation and land surface temperature estimates for the same period.
In the future, for both presented services, higher-resolution products derived from ESA’s recently launched Sentinel-1 and Sentinel-2 data could be mainstreamed gradually into the current NRSC national information database that is now based on either low to medium resolution data from sensors like MODIS or AVHRR or extensive field surveys. The use of the Sentinel satellite series would allow the services to benefit from significant improvements in spatial and temporal resolution, as well as pasture class discrimination.
As an additional service, a comparative study of the trade-offs between image costs and information gain for similar products to the ones delivered for each of the previous services for three different resolutions (MODIS, Landsat/ DEIMOS/DMCii and SPOT 5). The main impact of this analysis was visible during the Final User Workshop held in the Mongolian University of Life Sciences (MULS).
Project results are integrated in the web GIS portal available with user login at http://eotap.DEIMOS.com.pt.