Professor Danielle Wood leads the new Space Enabled Research Group at the MIT Media Lab. The Space Enabled Research Group advances justice in Earth’s complex systems using designs enabled by space. Space Enabled sees opportunities to advance justice by increasing access to space technology to development leaders around the world and by applying space technology in support of the Sustainable Development Goals, as curated by the United Nations. There are six space technologies that have been contributing to the Sustainable Development Goals for decades, but barriers remain that limit their impact. These technologies include satellite earth observation, satellite communication, satellite positioning, microgravity research, technology transfer and increasing research infrastructure. The Space Enabled Research Group implements projects with development leaders at the multi-lateral, national and local scale to apply space technology in support of their initiatives. During these projects, Space Enabled implements an integrated design process that includes techniques from engineering design, art, social science, complex systems modeling, satellite engineering and data science. During this talk, Prof Wood will discuss the role of space to spur innovation and development and give examples of projects pursued by the Space Enabled Research Group.

Informal settlements are the result of the continuing rapid growth of mega cities, especially in the global south where people migrate from rural areas in hope for a better future. Poverty is considered one of the major challenges to our society in the upcoming decades, making it one of the most important issues in the Sustainable Developments Goals as defined by the United Nations. These settlements, however, often lack basic sanitation and access to clean water. While many urban agglomerations of the global south are prone to large slum areas, still, the exact location and size of these settlements is often unknown. Remote sensing methods have improved tremendously in their capabilities of mapping informal settlements and its morphological features, which can be described by their high building density, small building sizes or its building orientation. But the challenge of large scale slum mapping still remains open, due to fuzzy feature spaces between formal and informal settlements, as well as a significant imbalance of slum occurrences where slums only account for 1% in the data. To tackle this issue we propose a fully convolutional xception network (XFCN). With its 34 convolutional and five dilated convolutional layers including four skip connections during the up-sampling phase the XFCN is capable of detecting poor urban areas in a coherent transfer learning approach using high resolution satellite images. This proves to be an ambitious task, differentiating between formal built-up structures and informal settlements at high resolutions. We train the network on a large sample of globally distributed slums (Cape Town, Caracas, Delhi, Dhaka, Lagos, Medellin, Mumbai, Nairobi, Rio de Janeiro, São Paulo and Shenzhen), greatly heterogeneous in its morphologic feature space and transfer the XFCN to map informal settlements. The XFCN is trained from scratch using 5 input channels and rigorous regularization. Using this approach we are able to reach an overall accuracy of up to 95%.

Global supply chains increasingly displace commodities production from the final consumption creating geographically distant drivers to local environmental and social impacts. International trade, for example, is associated with biodiversity threats and natural forest cover loss in tropical areas thousands of kilometres away from the consumers. Tracing product consumption back to the geographical location of production is crucial for informed decision making towards more sustainable supply chains.

Environmentally extended Input-Output models have already been applied to various environmental issues. These models are used to trace the environmental impacts from the producing country via multiple processing stages to the country of final consumption. However, the spatial resolution of these models is so far limited to the national level. This causes significant uncertainties in supply chains and impacts assessment, as the environmental and socioeconomic conditions often vary significantly within countries that harvest agricultural products and extract mineral resources. Therefore, it is urgent to develop methods for spatially explicit supply chains impact assessments, which can take the heterogeneity within producing countries into account. The ERC FINEPRINT project (www.fineprint.global) currently implemented at the Vienna University of Economics (WU) aims at developing a new method for performing supply chain assessments on a high spatial resolution.

This presentation shows our recent advances in using Earth Observation data to produce more accurate supply chain impact assessments. The FINEPRINT team is developing new methods to integrate Input-Output models and spatially explicit data sets on agricultural production and mineral extraction, and related environmental impacts. Our case studies illustrate how Earth Observation data plays a crucial role in improving the robustness of supply chains impact assessments.

Earth observation satellites are the only viable and cost-effective data source capable of providing regular and consistent information about mineral extraction and agricultural production. Satellite data products of high spatial, temporal, and spectral resolution combined with novel artificial intelligence algorithms, can, for the first time, provide the spatial data required to accurately assess the local and regional impacts of international commodity supply chains and trade.

Using Sentinel 2 and Landsat time series, we are assessing the spatial distribution of the production of agricultural commodities (e.g., soybeans, sugar cane, cotton, and maize). We applied our methodology in a pilot case study to trace the direct land cover changes driven by agricultural commodities produced in Brazil. We first built a crop reference database using a web application based on street view images. This data collection was used to train machine learning algorithms for crop mapping. Using time series of agricultural commodities maps, we estimate the direct impact on land cover changes and deforestation due to agricultural production, which is then traced to the final consumption using multi-regional Input-Output models.

To assess the spatial extent and the environmental impacts of the mining sector, we used openly available satellite images to quantify the land used for mining activities at a global scale, including mining lease and infrastructure. We developed a new web application for visual interpretation of mining activities. The app combines various sources of mining statistics with satellite images, mainly Sentinel 2 and very-high-resolution images available from Google Satellite. Our novel dataset allowed us to estimate the direct land cover change caused by mining and associated impacts, such as deforestation. Our new dataset also provides a source of accurate training data for artificial intelligence algorithms, which could be used to automatize the process of mapping mining activities.

We have developed new methods to improve impact assessment of agricultural production and mineral extraction using satellite images and machine learning. Our next step is to enhance the integration of the spatially explicit impacts with the multi-regional Input-Output models. This innovative coupling of natural science and socioeconomic data opens up a wide range of further research options. The new method allows performing assessments of the global impacts related to industry input requirements and final demand in industrialised countries. The method is beneficial for supply chain management, as it provides the empirical basis to identify options for greening global supply chains and thus support companies that aim to progress with their sustainability agenda. Economic geography is another field that can directly exploit the results for assessing how globalisation changes the spatial patterns of land use, trade and consumption.

Cloud and Earth Observation (EO) based services offer the European Research community a wealth of powerful tools. However, for many researchers these tools are currently out of reach. It is difficult to find and select suitable services. Establishing agreements with cloud and EO service providers and ensuring legal and technical
compliance requires specialist skills and takes an inordinate amount of time. Equally, service providers find it difficult to reach and meet the needs of the research community in technical, financial and legal areas.

The Open Clouds for Research Environments consortium (OCRE) will change this, by putting in place an easy adoption route. In the autumn of 2019, OCRE will run a pan-European tender and establish framework agreements with service providers who meet the requirements of the research community. 10.000 European research and education institutes will be able to directly consume these offerings via the European Open Science Cloud service catalogue, through ready-to-use agreements. They will not have to run a tender of their own. In addition, to stimulate usage, OCRE will make available 9.5 million euro in service credits (vouchers), through adoption funds from the European Commission.

The OCRE consortium consists of GÉANT (consortium coordinator), CERN, RHEA and Trust-IT with the contribution of SixSq, as technology provider, together with EARSC and Evenflow, as support to outreach activities, and receives funding from the European Commission, as part of the European Open Science Cloud.

This presentation will introduce the benefits of the Open Clouds for Research Environment for providers and for researchers. It will also address how OCRE aggregates community demand and requirements and applies these into a pan-European call-for-competition, for commercial service providers to respond to. This tender which will be launched in Q3 of 2019, will result in procurement-compliant framework agreements with suitable suppliers. In the presentation OCRE will talk about how providers can respond to this opportunity.

In addition the presentation will touch upon how researchers, affiliated/not affiliated to institutions will be able to consume the services through ready-to-use contracts, without the need to run their own tender and how OCRE will stimulate usage through an adoption fund.

Unsupervised LSTM-AE and Harmonic Models for Improving Timeliness of Forest Logging Notifications With Dense Satellite Image Time Series

The availability of free and open high resolution optical satellite imagery has greatly improved over the past few years, with the opening of the Landsat archives in 2008 and the launch of Sentinel-2 satellite in 2015. High resolution satellite imagery is widely used to monitor deforestation and forest degradation, evaluate carbon stocks worldwide, or map countries or continents. Satellite Image Time Series (SITS) analysis methods using all available data (Wulder et al., 2012) have defined the state of the art for continuous change monitoring and map updating (Zhu et al., 2014). However, reference data for accuracy assessment in time series analysis is often sparse, especially in the temporal domain or in case of rare phenomena such as storm damages, which motivates the development of unsupervised SITS analysis methods.

In 2010, Asta storm caused severe damages over 8.1 million m3 of forest in Finland, as well as to the country infrastructure (e.g. buildings, power lines and railroads), totaling financial losses over 20 million euros. The estimation of forest damages was carried out mainly by drive-by field visits or even involving flyovers by the Finnish air force, and was slow, expensive and inaccurate. In Finland, a significant part of the forest is owned by small private owners, who are required to report to the Finnish Forest Centre before any logging operation. However there is often a long delay (months or over a year) between a forest logging notification and the corresponding logging operation.

In this work, we propose an unsupervised SITS analysis model based on Long Short-Term Memory-Autoencoder (LSTM-AE) (Srivastava et al., 2015) network, and compare it with a well-known harmonic model trained in a quasi unsupervised way. Both models are used to demonstrate the benefits of dense SITS analysis for improving temporal and spatial accuracies of both forest logging notifications and storm damage mapping, compared to current operational practices in Finland.

All available 361 Landsat images acquired between 1997 and 2015 over adjacent footprints (paths 189 and 190, row 17) were downloaded and atmospherically corrected on USGS ESPA on-demand pre-processing platform. Scenes in path 190 were registered to a reference scene on path 189. All images were then stacked, keeping only pixels on the overlapping areas. Through this operation the effective temporal acquisition frequency was doubled, compensating for a low number of observations at high latitudes due to frequent cloud cover in summer and low sun angle in winter, that prevented radiometric correction.

The study area was located at Hyytiälä Forest Research Station in Finland, and entirely comprised in the overlap area between the two adjacent Landsat footprints. Reference data for accuracy assessment was collected from forest logging notifications of the Finnish Forest Centre, as reported by forest owners before normal logging operations (clearcuts, thinning cuttings) or consecutive to three major storm events between 2010 and 2015. In addition, clearcuts between 2010 and 2015 were visually delineated between two very high resolution images, with an improved spatial accuracy of the reference data.

We applied a LSTM-AE model, an autoencoder (AE) creating concise representations of time series data while preserving temporal information by using specific long short-term memory (LSTM) cells. Our version of the model was originally created for electronic health records - which are often sparse and contain both random and systematic biases - then adapted to SITS analysis which present similarities in terms of irregular temporal sampling and sparsity of break events.

As a baseline, the Continuous Change Detection and Classification (CCDC) algorithm (Zhu et al., 2014) was applied to the Landsat image stack (on the Short-Wave Infrared Band - SWIR 1 at 1.55-1.75 mm) in a quasi-unsupervised way, skipping the final supervised classification steps. The CCDC parameters were tuned interactively on pixels of known stable forest plots and clearcuts within the studay area. Once the parameters were set, change magnitude was computed automatically between the relevant dates corresponding to each event (storm or clearcuts). Finally, change magnitude maps were thresholded to filter out undesired changes and produce change maps.

The spatial accuracy of unsupervised forest change maps from LSTM-AE and CCDC were comparable, while the LSTM-AE approach was fully automatic and unsupervised. In addition, unsupervised change maps obtained from time series were more timely and temporally accurate than reference semi-supervised maps obtained from a bi-temporal change detection method used operatively. Results confirm the benefits of dense time series analysis using all available data instead of annual time series or bi-temporal change detection, for improved tolerance to missing observations and increased timeliness in the detection of forest changes.

Automatic analysis of dense image time series enables the production of accurate forest change maps, more frequently and cost-effectively than e.g. Lidar mapping (5-6 € /ha, currently used every 5 years or more to map forests over the whole Finland). The proposed methods are almost entirely automatic and thus suited to operational conditions, accurately updating logging notifications and storm damage maps.

Nitish Srivastava, Elman Mansimov, Ruslan Salakhutdinov, 2015. Unsupervised Learning of Video Representations using LSTMs, ICML 2015.
arXiv:1502.04681

Michael A. Wulder, Jeffrey G. Masek, Warren B. Cohen, Thomas R. Loveland, Curtis E. Woodcock, 2012. Opening the archive: How free data has enabled the science and monitoring promise of Landsat. Remote Sensing of Environment, Volume 122,Pages 2-10, https://doi.org/10.1016/j.rse.2012.01.010

Zhe Zhu and Curtis E. Woodcock, 2014. Continuous change detection and classification of land cover using all available Landsat data,
Remote Sensing of Environment. Volume 144, Pages 152-171, https://doi.org/10.1016/j.rse.2014.01.011

According to The World Bank, food production must double by 2050 to feed the world's growing population. However, agricultural production systems in many developing countries are neither resource-efficient nor producing to their full potential in order to meet these global food demands in a sustainable way (Chaherli & Nash, 2013). There is a number of factors responsible for this lack of efficiency, such as climatic, environmental, technological, economic, and policy-related. Moreover, a key problem, common to many of growers, is that they struggle to access the right information to manage their croplands in the best possible way. The combination of Remote Sensing (including the sheer amount of open Earth Observation data), Information Technologies and Artificial Intelligence, offers opportunities for increasing productivity while preserving our planet’s productive capacity.
We present AMIGROW, a service that aims to integrate field information, about crop condition and management practices, with satellite imagery in order to provide agricultural managers valuable insights to manage crop production and maximise crop yield. Our startup was born at the first ESA phi-week bootcamp in 2018. During three days, a multidisciplinary team, that included experts in data analytics, project management, user experience design, business, agriculture and remote sensing, worked to develop the idea. As a result of several iterations and lengthy discussions, our idea won the first prize at the event Phi-week 2018 event (http://phiweekbootcamp.space/prize-winners/).

Since the ESA phi-week bootcamp, AMIGROW has been in a constant evolution. Initially, we are focusing on medium/big size agricultural production systems in Colombia, as well as associations of growers. Colombia has been chosen as the first location to develop and validate AMIGROW services, due to the experience and knowledge that AMIGROW members have in this specific agricultural sector. An first market validation was performed in the region of Tolima, which is known for producing high-quality rice in the country. This partnership will help AMIGROW to find new customers willing to participate in making a first prototype.

AMIGROW’s MVP will be composed of three complementary modules:
A crop planning module, providing basic information about the conditions of the area where the crop is located to help managers plan their agricultural activity.
A farming data management module, which serves as an infrastructure to store, edit and manage the ground measurements about the crop condition and the activities performed on the fields. This module is a key piece of AMIGROW, because the crowdsourced data recorded by different actors will feed our data analytics module.
A data analytics module, integrating satellite imagery and data recorded in the farming data management and the crop planning modules to generate insights directly applicable to the crop management practices. AMIGROW data analytics solution will leverage Machine Learning and Artificial Intelligence techniques, that applied to the data recorded on the field and the satellite images will generate insights tailored to the needs of each customer and the inner dynamics of each crop.

After winning the Phi Week Bootcamp, Amigrow’s team continued working on developing the idea and seeking external validation for the implementation. For this reason, we applied to the InvEnterPrize Student Business Ideas Competition, organized by Aberystwyth University. The project won the first prize and an additional prize as the strongest application supporting the bio-sciences, life sciences and agriculture sectors, receiving a key cash injection for the development of the first Minimum Viable Product (https://www.aber.ac.uk/en/news/archive/2019/04/title-222255-en.html).

We have a long way ahead developing AMIGROW, a solution that will boost food production in a sustainable way. Winning the ESA phi-week Bootcamp first prize, but specially the support we got from experts in space technologies and business creation, provided the impulse that our idea needed in order to grow.