GsAL Returns to Bangkok

Written by Oz Brown

How do you sustainably manage urban growth in a way that allows you to make the most efficient use of your land and water resources? How do you account for rapid declines in forest cover and the resulting physical, ecological and human problems this can cause? Further – how do you account for declines in forest cover in a neighbouring country, for which you may not have access to accurate, up-to-date mapping data? And how do you reduce the risk of losing hard earned development gains in areas that are prone to natural disasters such as landslides or flooding? These are some of the problems that current land-use planners and decision makers face in a world of rapid change and shifting climates; problems in the Lower Mekong Region that are being tackled in a wide collaboration to produce a dynamic new mapping tool that provides greater detail and accuracy of data.

From the 1st to the 3rd of August 2017, SERVIR Mekong in collaboration the University of San Francisco’s Geospatial Analysis Lab (GsAL), the US Forest Service and SilvaCarbon held the fourth in a series of conferences to launch their Regional Land Cover Monitoring System (RLCMS). This is an innovative tool designed to produce high-quality land cover maps that will allow decision makers in the Lower Mekong region – from national governments to local community groups, non-profits and the private sector – to make informed policy and planning decisions about resource allocation, disaster preparedness, climate change resilience, ecological conservation, carbon accounting and a whole host of other critical issues affecting these societies. The conference was led by David Saah, the director of GsAL, who has been working closely with the GsAL lab manager, Megan Danielson, on developing the RLCMS since its conception.

Over the last two years the SERVIR-Mekong and GsAL teams have been working closely with stakeholders across the Lower Mekong region (Thailand, Vietnam, Myanmar, Laos and Cambodia) and Indonesia to identify local, regional and national needs, to develop a consistent regional classification scheme, to collect and employ on-the-ground reference data, develop algorithms for identification of land-cover classes and implement plans for how the RLCMS will be used.  One of the most encouraging and affirming aspects of the process is that representatives from groups in all of these countries (e.g. government reps, non-profits and local interests) have come together to develop and learn to use a tool that crosses political boundaries and provides better opportunities for the citizens, environments and industries of the Lower Mekong region to interact and prosper in more sustainable ways.

Since the late 90s there has been a drive in the environmental sciences community to produce global land cover mapping data and a number of detailed and invaluable maps have been created, such as the European Space Agency’s GlobCover Land Cover V2, the UN Food and Agriculture Organization’s  Global Land Cover-SHARE and the Natinoal Geomatics Center of China’s GlobeLand30. Traditionally, regional actors have used such efforts to extract what information they can for their own varied uses, but this method has its limitations. Existing maps are infrequently updated, the classification systems they use can be inconsistent, the actual classes they identify can be inadequate (i.e. they do not provide accurate representation of the range of biophysical layers present in a given area), they may not meet accuracy assessment requirements and some need significant data storage and processing power to run. The RLCMS, on the other hand, harnesses the processing power of Google Earth Engine (GEE) to enable the this kind of analysis to be performed on a standard computer and combines its staggering quantity of satellite imagery and data with local field research by stakeholders to provide functionality, speed and accuracy. Biophysical layers are mapped by training statistical models and machine learning algorithms on reference data obtained from said field research or existing legacy data-sets. Thanks to the input of regional stakeholders since the beginning of the process the tool uses a consistent classification scheme to describe a broad range of layers appropriate to the diverse ecology of the Lower Mekong region which can be customized and expanded as required by the user, creating high quality land cover maps that are regularly updated.

Many of the national priorities for the countries comprising the Lower Mekong region with regards to the RLCMS are similar – to simply improve the quality of their own nation’s land cover maps, to help meet targets for carbon capture and emissions and to build capacity to meet their development needs. However, many other uses for the RLCMS have been identified which reflect the different needs of these countries and organisations working within them. Laos requires improved forest mapping data and accuracy; the Mekong River Commission, an inter-governmental organisation representing the water needs of Laos, Thailand, Vietnam and Cambodia, will use it as an additional input into their flood monitoring systems; Indonesia has suffered from peat fires for decades and famously had an extreme crisis event in the 2015 fire season when 2.6 million hectares of land were burned, tripling their annual carbon emissions – so there is a significant need to be able to accurately define the areas affected, complete analysis of land cover changes over time and incorporate the new data they retrieve into existing methodology to prepare for and reduce the negative effects of those fires.

A tremendous amount of work has been poured into the collaboration to develop the RLCMS over the last three years and there are significant steps still to take – chiefly, using reference data to input land cover classes in the stakeholders’ respective areas of interest. However, many key obstacles have been already overcome and it is clear that this tool will afford users greatly improved prospects for informed decision making and positive change. In producing and developing the RLCMS, SERVIR-Mekong, GsAL, the US Forest Service, SilvaCarbon and the regional stakeholders involved are setting the groundwork for a comprehensive and reliable data mapping tool that has dizzying application potential.

 

 

 

 

Monitoring Land Cover for Resilient Development

This article was provided by SERVIR-Mekong/ADPC, and is also available on the ADPC website.

Take a look at this land cover photo. What do you see? Is it a forest, agriculture, water or something else?

Landcover_GoogleMapImage
Photo Credit: Google

It may seem like an irrelevant question, but it is not. Understanding land cover and land use change is important for land resource planning and for ecosystem services. This includes biodiversity conservation, water provision and purification, and resilience to climate change, among others. However, updates to land cover maps are infrequent, and classification systems can be inconsistent across years and countries.

 Workshop participants discuss path forward
Wildlife Conservation Society (WCS) and ADPC participants from the Regional Land
Cover Monitoring System workshop discuss a consistent path forward in classifying
different plots. Photo Credit: Watcharaporn Usomchat  

Returning to the question in the beginning, how do we classify different land types? Working with key partners in the region is necessary to promote a coordinated approach in mapping and land classification. In March 2016, the Asian Disaster Preparedness Center (ADPC), along with SilvaCarbon, the US Forest Service and the Spatial Informatics Groups (SIG), brought partners from the region to begin the process of agreeing to common land classifications that could address the needs of several groups. This was part of a larger process of building a Regional Land Cover Monitoring System in the Lower Mekong Region, one of the flagship systems being created under the SERVIR-Mekong initiative funded by the United States Agency for International Development (USAID) with NASA and implemented by ADPC.

This work continued from 7-14 July 2016 as SERVIR-Mekong, along with SilvaCarbon, the US Forest Service and SIG, hosted a Google Earth Engine Training and a second workshop for the Regional Land Cover Monitoring System. Stakeholders worked particularly on algorithm development and reference data collection approaches.  During the workshop, partners in the region convened to determine a consistent path forward in classifying different plots (or land types). This is a key step in the development of the Regional Land Cover Monitoring System, which will promote in-depth and reliable analysis on land cover issues in the Lower Mekong and provide reliable and consistent maps for the region.

 Facilitator David Saah works with participants using Google Earth Engine
Workshop facilitator, David Saah (Associate Professor & Director of Geospatial
Analysis Lab, Department of Environmental Science, University of San Francisco)
works with some ADPC, WCS and University Partnership Network participants in
using Google Earth Engine. Photo Credit: Watcharaporn Usomchat  

As development of the Regional Land Cover Monitoring System continues, SERVIR-Mekong will be launching MAPCHA, a custom built and innovative tool that will allow for crowd sourced visual interpretation of satellite images. Along with other reference data sets from ministries and departments, image interpretations from participants throughout the Lower Mekong region will constitute a large training and validation dataset.

With the Regional Land Cover Monitoring System, decision makers, partners, and the general public will have access to reliable information on land cover, providing evidence based data to inform land use planning for climate resilience.

SERVIR-Mekong is a unique partnership with USAID and NASA, to promote the use of publicly available space technology for resilient development in the Lower Mekong. SERVIR-Mekong is implemented by Asian Disaster Preparedness Center with Spatial Informatics Group, Stockholm Environment Institute and Deltares.

For more information, please visit servir.adpc.net