Statement_of_Work_1.pdf

P. 1 of 4

STATEMENT OF WORK

Global Food Security-support Analysis Data (GFSAD) Project

I. Introduction

Climate variability and ballooning populations are putting unprecedented pressure on agricultural croplands and their water use, which are vital for ensuring global food and water security in the twenty-first century. In addition, the COVID-19 pandemic, military conflicts, and changing diets have added to looming global food insecurity. Therefore, there is a critical need to produce consistent and accurate global cropland products at fine spatial resolution (e.g., farm-scale, 30m or better), which are generated consistently, accurately, and routinely (e.g., every year). In this regard, we produced the world’s first Landsat-derived global cropland extent product @ 30m (GCEP30) (Thenkabail et al., 2021; download @ LP DAAC) funded by NASA MEaSUREs 2013 and later continued by USGS.

The high impact of our previously-funded GFSAD products such as Landsat 30m GCEP30, 1km cropland dominance, and 1km irrigated versus rainfed is demonstrated by the use of these data by 126 countries during 2018-2021 (97 countries in 2021 alone), continued average downloads every month by about 20 countries, publication of 11 key peer-reviewed articles which already have 1276 citations in a short time-period (2017-present), and use for a wide range of applications.

Therefore, the overarching goal of this GFSAD continuity project is to develop a comprehensive global food security-support analysis data (GFSAD) project that will produce multiple cropland models, maps, and monitoring tools leading to a wide array of products using machine learning algorithms (MLAs), and satellite sensor big-data analytics through cloud-computing. In this continuity GFSAD project, we focus on producing four distinct Landsat-derived global cropland products for the years 2020 and 2025:

1. Global Cropland Extent Product @ 30m (LGCEP30-2020, LGCEP30-2025).

2. Global Rainfed and Irrigated Product @ 30m (LGRIP30-2020, LGRIP30-2025).

3. Global Cropping Intensity Product @ 30m (LGCIP30-2020 & LGRIP30-2025) &

4. Global Crop Type Product @ 30m for USA, Canada, and India (LGCTY30-

2020USACAN, LGCTY30-2025USACAN; LGCTY30-2020India, LGCTY30- 2025India).

The study will make use of Landsat-8, 9, and Sentinel-2A&2B surface reflectance products already available in the Google Earth Engine (GEE) cloud, and NASA’s Harmonized Landsat Sentinel-2 (HLS) Landsat derived product (HLSL30) for 2013-present and Sentinel-2 derived product (HLSS30) for 2015-present, that together have sub-5-day global coverage at nominal 30m resolution. The four cropland products will be generated using 14 mature MLAs that the team has extensive experience with such as random forest, support vector machines, decision trees, and spectral matching techniques (e.g., Thenkabail et al., 2021, Oliphant et al., 2019, Teluguntla et al., 2018, 2015, Xiong et al., 2017a, Thenkabail et al., 2012, 2009, 2007, 2005) as outlined in the methods section, utilizing GEE and\or other cloud platforms.

All crop products will be evaluated for accuracies, errors, and uncertainties using various datasets such as ~ 112,000 points in the global cropland reference dataset; cropland data from USDA CDL, Agri-Food Canada (AAFC) Canada, ground data for various countries through the CGIAR global network such as the ICRISAT, ICRAF, and IITA, country-wise irrigated and rainfed statistics of UN FAO through University of Goettingen (Dr. Stefan Siebert), an additional ~250,000 samples to be taken during the project period using our ground data Mobile App, P. 2 of 4 country-wide reference maps, and sub-meter to 5m very-high-resolution-imagery (VHRI) from the National Geospatial-Intelligence Agency (NGA) available to USGS.

GFSAD will make significant contributions to the Earth System Data Records (ESDRs) including Climate Data Records (CDRs), the Group on Earth Observations Agriculture and Water Societal Beneficial Areas (GEO SBAs), and the GEO Global Agricultural Monitoring (GEOGLAM). All data will be released through NASA’s LP DAAC like our earlier GFSAD releases.

Thereby, this contract supports two major initiatives:

1. Advancement of existing cropland products, by producing the most latest updates, and,

2. Development of new and unique products to further comprehensive assessment of global food and water security in the twenty-first century.

Both will help advance our knowledge on global food and water security by providing objective, un-biased, and consistent data for any part of the world year after year. Further, the use of big-data, machine learning (ML), deep learning (DL), artificial intelligence (AI) by coding and computing on the cloud, allows the USGS to do 21st Century science using 21st Century data, methods, tools, and techniques. This contract supports the development of operational and scientific improvements to the GFSAD project through development of refined automated machine learning algorithms on cloud computing platform.

As specified in this contract, the Contractor will design and complete advancements, enhancements, as well as new innovative developments, which will be subject to review and approval by USGS Point-of-Contact (POC).

II. Deliverables

Overall, the contractor will produce a total of five global food security-support analysis data (GFSAD) products routinely and repeatedly, year-after-year, by processing multi-sensor remote sensing data on the Google Earth Engine (GEE) cloud using spectral matching technique (SMT) algorithms and other multiple machine learning algorithms (MLAs). All five cropland products

(1. Cropland extent\areas, 2. Irrigated versus rainfed, 3. Cropping intensities, 4. Crop types, and

5. Croplands versus cropland fallow) will be produced @ 30-m resolution for the entire world using Landsat and Sentinel time-series data for the 2000-2026 period. These CMAs and MLAs will have the ability to Hind-cast, Now-cast, and Future-cast. Finally, all products, algorithms, and manuscripts will be first released on USGS and LP DAAC on the following sites:

https://www.usgs.gov/apps/croplands/app/map https://lpdaac.usgs.gov/news/release-of-gfsad-30-meter-cropland-extent-products/

The USGS will also release all training and reference data produced for machine learning.

The three global products, associated algorithms, and manuscripts will be one-of-a-kind knowledge storehouse on global food and water security-support analysis data (GFSAD).

P. 3 of 4

III. Impact and Outcomes

(A) Support of Landsat data continuity mission (LDCM): The global food security-support analysis data (GFSAD) products/algorithms are a very high value, one of a kind, application of Landsat-Sentinel-MODIS sensors @ nominal 30-m. These products help promote and justify the demand for the LDCM and NASA, USGS, and other US Government run satellite remote sensing programs.

(B) Global food-security support analysis data (GFSAD): GFSAD will significantly contribute in planning and ensuring global food security scenarios. For example, the current GFSAD project is in process of releasing the world’s first Landsat derived global rainfed and irrigated area product @ 30m (LGRIP30) (Teluguntla et al., 2023). Also, in another study, the current GFSAD project showed that cropland areas were reduced by as much as 12 million hectares (22% of total cropland areas) in Australia during 2002, one of the worst drought years in 100 years (Teluguntla et al., 2017). The same study showed that the vigor of existing croplands during that year also fell. This combination (reduction in cropland areas and reduction in cropland vigor or yield) has a devastating effect on not just Australia’s food security, but that of the entire world given that Australia is a major exporter of food.

(C) Global water-security support analysis data (GWSAD): Precise estimates of global cropland areas as well as cropland water use of major world crops will be of immense value for planning-sustaining-ensuring global food as well as global water security (given ~80% of all human water use goes towards producing food) and to meet the world’s sustainable development goals of the twenty-first century.

(D) Support global crop water models: Accurate and detailed mapping of the world’s croplands will help improve global crop water models (GCWM) to help accurately determine:

(1) green water use (by rainfed croplands), (2) blue water use (by irrigated croplands), (3) food production, (4) dynamics of virtual water trade, and (5) scenario modeling.

(E) Cropland Mapping Algorithms (CMAs) with multiple machine learning algorithms (MLAs) for global food and water security with the ability to Hind-cast, Now-cast, and Future-cast: The cropland and crop water productivity products produced in this study @ nominal 30 m are produced on a cloud computing environment for which machine learning algorithms will be provided in open access through LP DAAC and will have the ability to Hind-cast, Now-cast, and Future-cast. The GFSAD data will be available for use by Governments and Institutes, which is intended to benefit multiple institutes and agencies (e.g., GEOGLAM, GEO- Agri. and water societal beneficial areas, USAD, USGS, NASA, International organizations such as the CGIAR-system).

IV. Supporting Information

A. Place of Performance:

Vendor Facility

B. Period of Performance:

Estimated September 1, 2023 to August 31, 2026

C. Government Furnished Property:

None

D. Special Considerations

P. 4 of 4

1. Key Personnel Qualifications Ability to conduct independent research, published papers in peer-reviewed journals on the relevant subject matter, understanding of global agriculture and water issues, deep knowledge of remote sensing science, ability to code and compute on the cloud. A key skill is the ability to code and computed on the cloud using ML\DL\AI utilizing peta-byte scale computing. The incumbent must be familiar with LP DAAC platform for data dissemination.

2. Data Rights All the codes and products must be made available through public domain web access.

V. Task Order 1

The first task order of the global food security support analysis data (GFSAD) project is to provide Landsat-derived global rainfed and irrigated area product @ 30m (LGRIP30) for 2020 and 2025. This is followed by mapping major crop types of the world using ML\DL\AI on the cloud.

There is roughly 1.35 billion hectares of rainfed croplands and 450 million hectares of irrigated croplands in the world. As per the United Nations (UN) and the UN Food and Agriculture Organization (FAO) estimates of the global population expected to reach 11-12 billion by the year 2100, the demand for agricultural commodities such as food, fodder, and biofuels is expected to increase 70-100% by 2050 and possibly double that by 2100. With increase in population combined with calorie intake increases, the world will need a whopping 80% more calories by the year 2100 to feed the global population. Irrigated areas produce 40% of the world’s food supply from 25% of the total cropland areas whereas the rest of the 60% food comes from the 75% of the total cropland areas. Thereby, knowing where they are located precisely in the world is of great importance for understanding, modeling, and mapping crop productivities of these croplands and their variabilities over time and space. Further, globally, nearly 90% of all human water use goes towards producing food from agriculture. Of this 90%, nearly 70% of the water is used by 1.2 billion hectares of rainfed croplands (green water use, GWU) and the remaining 30% by 400 million hectares of irrigated croplands (blue water use, BWU). For these reasons, understanding and mapping irrigated and rainfed croplands are crucial for understanding, assessing, and acting on global food and water security.

In this context, USGS is looking for Landsat-derived global rainfed and irrigated area product @ 30m (LGRIP30) with least uncertainties and errors for nominal years 2020 and 2025. The product is to be developed using Landsat and Sentinel data; big data analytics by developing reference (training and validation) data hubs; machine learning; artificial intelligence; and cloud computing. The LGRIP30 product for the year 2020 and 2025 must also include irrigated and rainfed areas of every country in the world for the nominal year 2020 and 2025. All the codes and products must be made available through public domain web access.