J2.5 Assessment of Rice Hydroperiod and Irrigation Practices Using Multiscale Earth Observations to Support Food Security

Monday, 8 January 2018: 9:45 AM
Room 18B (ACC) (Austin, Texas)
Nathan Torbick, Applied Geosolutions, Durham, NH

Rice ecosystems reside within a water-energy-food (WEF) nexus given the interlinked tradeoffs among resources, production decisions, and outcomes. Rice is one of the most important crops globally for food security, with more than 1 billion people depending on rice to support diets and livelihoods. This past year approximately ~500 million tons of milled rice was generated from 161,527,000 hectares of harvested area. The total accumulated area of rice has largely tapered off as available arable land is becoming scarce and competition for land uses evolve. This has resulted in intensified practices for additional crop cycles and production amplifying water utilization managements to optimize production, resource levels, and ecosystem services. Paddy rice requires intense water resources and management for cultivation. Globally, agriculture accounts for approximately 70% of all water withdrawn from surface waters and aquifers. While less than 20% of cultivated area is irrigated, these areas produce more than half of the global food supply. FAO estimates that rice covers 30% of total irrigated cropland. Irrigation practices, access to water and dependence on seasonal rainfall, and infrastructure vary tremendously across geographies resulting in many unique hydroperiod (inundation timing, duration, and frequency) patterns. Generally, rice fields can undergo 5-20cm of inundated conditions during sowing / transplanting stages, intermittently among growth phases (e.g., stalk development, flowering, tillering), between rotations for land preparation, and potentially for other seasonal uses such as aquaculture or wildfowl habitat promotion. In this setting we will present a summary and examples of rice hydroperiod mapping in South / Southeast Asia and USA using multi-scale Earth Observations and how these drive food security. We rely on multi-scale mapping leveraging strengths of optical and radar sensors that complement each other. For example, in Sotuh Asia we have operationalized the use of time series Sentinel-1A and B Interferometric Wide imagery to map rice extent, crop calendar, inundation, and cropping intensity across production hot spots including Red River Delta, Mekong, Myanmar, Cambodia, and Thailand. We generate moderate resolution (<30m) crop extent maps by fusing Sentinel-1, Landsat-8 OLI, and PALSAR-2 using a randomforest based machine learning approach that is tuned with field data and national survey statistics. Time series phenological analyses of the dense Sentinel-1 and PALSAR-2 ScanSAR are used to track inundated conditions, duration, and frequency which drives food secyurity in S/SE Asia. In South Asia the crop extent maps have mean out-of-sample kappa of over 90% and strong agrement (R2=0.8)with census statistics. In the USA we have focused on using near daily moderate resolution SAR-optical imagery to map rice hydroperiod and irrigation practices. For example, irrigation in Arkansas occurs on 1.82 million ha, the majority of which is located in eastern Arkansas and derives largely from the shallow Mississippi River Valley Alluvial Aquifer underlying the alluvial plain. Understanding the intricate complexity between hydrology, nutrient inputs and aquatic drainage, and irrigation within rice production systems is expected to lead to adoption and innovation of drainage management systems to reduce nutrient loads in downstream aquatic ecosystems. This, in turn, is expected to lead to greater transfer of conservation technologies and management systems for better strategies towards preventing nutrient contamination and impairment from source production agriculture to receiving systems. We have been mapping irrigation practices in partnerships with producers to identify best practices and conservation managements. These data rely on drones, Sentinel-1, and Harmonized Landsat8-Sentinel2 (HLS) data. Results and current best practices will be shared and made available openly.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner