1. General

Category

Category

Agriculture

2. Project Details

Company or Institution

NASA Harvest, University of Maryland

Project

NASA Harvest

General description of the AI solution

Harvest is developing solutions that provide information on agricultural production and land use that support the attainment of several SDGs as well as monitoring their achievement via the Global Indicator Framework. Harvest contributes to Goal 2: Zero Hunger, water (Goal 6), responsible consumption and production (Goal 12), climate action (Goal 13), life on land (Goal 15), and global partnerships for sustainable development (Goal 17). Towards SDG-2 targets 1 and 2, Harvest is using ML and EO data to accelerate the availability, timeliness, and quality of information on crops. This feeds into processes for early warning of crop failures and production shortfalls as well as empowering decisions related to food security, including global food aid, (re)insurance activation, and farmer response. Towards SDG-12 target 12.A, Harvest supports developing countries to integrate and leverage EO- and ML-based solutions with existing monitoring frameworks through extensive capacity-building programs. Since November 2017, NASA Harvest has initiated or been involved in ~30 projects globally to improve tools and grow regional and local capacity to address food insecurity. Harvest maintains a satellite-based Global Agriculture Monitoring system (GLAM) developed by the University of Maryland with NASA and USDA. GLAM was customized for East Africa, enabling the implementation of the World Bank’s Disaster Risk Financing and Insurance Program. In Uganda, this program has supported >300,000 individuals in Karamoja, providing alternative livelihoods to smallholder farmers affected by drought. This system also enables the delivery of newer maps and solutions using ML including crop maps and yield forecasts. Harvest’s 2019 crop map of Togo was used to implement the YOLIM program which has served more than 50,000 people. This work has garnered global recognition. Harvest Africa program director, Catherine Nakalembe, was selected as a 2020 Africa Food Prize laureate. Harvest team member Hannah Kerner was included on the Forbes 2021 “30 Under 30” list in science Christopher Justice, Harvest’s chief scientist, received a NASA Distinguished Public Service Medal.

Website

https://nasaharvest.org/

Organisation

University of Maryland

3. Aspects

Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.

This nomination is for the NASA Harvest team. Harvest is developing solutions that provide information on agricultural production and land use that support the attainment of several SDGs as well as monitoring their achievement via the Global Indicator Framework. Harvest contributes to Goal 2: Zero Hunger, water (Goal 6), responsible consumption and production (Goal 12), climate action (Goal 13), life on land (Goal 15), and global partnerships for sustainable development (Goal 17). Towards SDG-2 targets 1 and 2, Harvest is using ML and EO data to accelerate the availability, timeliness and quality of information on crops. This feeds into processes for early warning of crop failures and production shortfalls as well as empowering decisions related to food security, including global food aid, (re)insurance activation, and farmer response. Towards SDG-12 target 12.A, Harvest supports developing countries to integrate and leverage EO- and ML-based solutions with existing monitoring frameworks through extensive capacity-building programs. Since November 2017, NASA Harvest has initiated or been involved in ~30 projects globally to improve tools and grow regional and local capacity to address food insecurity. Harvest maintains a satellite-based Global Agriculture Monitoring system (GLAM) developed by the University of Maryland with NASA and USDA. GLAM was customized for East Africa, enabling the implementation of the World Bank’s Disaster Risk Financing and Insurance Program. In Uganda, this program has supported >300,000 individuals in Karamoja, providing alternative livelihoods to smallholder farmers affected by drought. This system also enables the delivery of newer maps and solutions using ML including crop maps and yield forecasts. Harvest’s 2019 crop map of Togo was used to implement the YOLIM program which has served more than 50,000 people.
This work has garnered global recognition. Harvest Africa program director, Catherine Nakalembe, was selected as a 2020 Africa Food Prize laureate. Harvest team member Hannah Kerner was included on the Forbes 2021 “30 Under 30” list in science Christopher Justice, Harvest’s chief scientist, received a NASA Distinguished Public Service Medal

Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.

This nomination is for the NASA Harvest team. Harvest is developing solutions that provide information on agricultural production and land use that support the attainment of several SDGs as well as monitoring their achievement via the Global Indicator Framework. Harvest contributes to Goal 2: Zero Hunger, water (Goal 6), responsible consumption and production (Goal 12), climate action (Goal 13), life on land (Goal 15), and global partnerships for sustainable development (Goal 17). Towards SDG-2 targets 1 and 2, Harvest is using ML and EO data to accelerate the availability, timeliness and quality of information on crops. This feeds into processes for early warning of crop failures and production shortfalls as well as empowering decisions related to food security, including global food aid, (re)insurance activation, and farmer response. Towards SDG-12 target 12.A, Harvest supports developing countries to integrate and leverage EO- and ML-based solutions with existing monitoring frameworks through extensive capacity-building programs. Since November 2017, NASA Harvest has initiated or been involved in ~30 projects globally to improve tools and grow regional and local capacity to address food insecurity. Harvest maintains a satellite-based Global Agriculture Monitoring system (GLAM) developed by the University of Maryland with NASA and USDA. GLAM was customized for East Africa, enabling the implementation of the World Bank’s Disaster Risk Financing and Insurance Program. In Uganda, this program has supported >300,000 individuals in Karamoja, providing alternative livelihoods to smallholder farmers affected by drought. This system also enables the delivery of newer maps and solutions using ML including crop maps and yield forecasts. Harvest’s 2019 crop map of Togo was used to implement the YOLIM program which has served more than 50,000 people. This work has garnered global recognition. Harvest Africa program director, Catherine Nakalembe, was selected as a 2020 Africa Food Prize laureate. Harvest team member Hannah Kerner was included on the Forbes 2021 “30 Under 30” list in science Christopher Justice, Harvest’s chief scientist, received a NASA Distinguished Public Service Medal.

Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.

Harvest is rapidly scaling up machine learning- and EO-enabled approaches for mapping cropland and crop-type as well as modeling and forecasting crop conditions and yield. Harvest’s scalable crop mapping approach (see Kerner et al., 2020, KDD; Tseng et al., 2020, NeurIPS) has been used to generate multi-year, 10 m/pixel crop maps in multiple countries including Togo, Kenya, Uganda, Rwanda, Ethiopia, Sudan, and China. Harvest has also developed the Global Earth Observations for Crop Inventory Forecasting (GEOCIF) system: an automated system that produces alerts to assess crop conditions globally by applying machine-learning algorithms on EO data. This model has been successfully implemented in multiple countries including the US, Argentina, Ukraine, and Kenya, and is now rapidly being scaled to support a regional food balance sheet in 11 African countries to address critical data gaps. Recognizing that computing infrastructure can limit the adoption of ML/EO solutions, Harvest developed and is now testing a framework for cost-efficient cloud deployment of the cropland and crop type models to ease accessibility. This is critical for training as well as ease of transferability of approaches and for smooth and cost-effective deployment (e.g., Indicator 4.4.1, Targets 14. a and 17.8). A key emphasis of Harvest’s work is reproducibility, scalability, and openness. Whenever possible, all data and code are made publicly available through Harvest’s Github page (https://github.com/nasaharvest) and data-sharing platforms like Zenodo.

Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.

A key aspect of our approaches includes co-development and continued development to ensure our solutions keep up with technological advancements as well as data availability. An example is our continued improvement of frameworks for cropland, crop type, and yield mapping. We also have streamlined accessibility to solutions not only through open access publication of papers, data, and models but also through direct integration of the advanced solutions we develop. For example, as we develop newer and higher-resolution maps, these are integrated into the operational crop monitor system as well as the Global Agricultural Monitoring System that is also open access. Moreover, we rely heavily on our partners during model development and their participation in data collection during model preparation. We hold in-country training events to facilitate the use of co-developed solutions, ensure the solutions meet the resource requirements of the regions, and solicit feedback for improving our systems based on local expertise. We also participate in training programs and initiatives hosted by other organizations to share with and engage a broader audience for our work. Another key aspect of our work is the geographic fairness of AI systems. We focus our efforts on developing new methods that improve the performance of AI systems in regions that are data-scarce and typically underrepresented in ML research and solutions. We also work to improve the representation of data-scarce regions such as Sub-Saharan Africa in publicly available datasets for EO and ML research.