%0 PDF %T Satellite-based monitoring of crop yields and yield gaps: methods and applications across spatial scales %A Jeffries, Graham Robert. %D 2018-03-16T09:33:52.457-04:00 %8 2018-03-16 %R http://localhost/files/bz60d773x %X Abstract: Background Meeting the food needs of growing populations consuming more resource-intensive diets will require increasing the magnitude and stability of crop yields. Better data is needed in order to understand the determinants and indicators of crop yields, and to target efforts for yield improvement. Satellite remote sensing imagery has substantial potential for creating low-cost geospatial crop monitoring datasets for research and decision making. This dissertation asks if innovative yield mapping methods with satellite imagery are suitable for three crop monitoring applications: crop yield gap (potential minus actual yield) analysis across farm size groups, yield mapping to inform precision agriculture practices, and in-field characterization of plant traits in advanced crop breeding experiments. Methods This project used remote sensing imagery, socio-environmental datasets, and a mixture of biophysical and statistical modeling tools to predict and then analyze crop yields. Objective 1 was to assess variation in Brazilian soybean yields and yield gaps with respect to farm size, using remote sensing imagery to estimate yields and characterize parcel-level features. Objective 2 employed a crop yield prediction algorithm combining remote sensing imagery with crop simulation models to predict mean maize yields at 10 m resolution in Nebraska, USA, and then validated the predictions with harvester yield monitor records. Objective 3 built on the methods and site data in Objective 3 to develop and test a yield prediction algorithm for mapping sub-field maize yields across rainfed and irrigated fields. Results Yield gaps in Brazilian soybean production systems were mapped with satellite imagery, revealing significant variations across farms of different size (ranging 20-50,000 ha). Farm size was positively related with soy yield, and inversely related to yield gap size. Objective 2 found that a crop yield prediction model which required no in situ data collection was successful in mapping mean maize yields at 10 m resolution (R2 0.68, RMSE = 0.99 mt ha-1). Objective 3 showed that single-season maize yield prediction accuracy varied with irrigation status, remote sensing imagery source, and algorithm parameters, explaining up to 22.2% of the variation in sub-field yields. Implications Crop yield monitoring with satellites and modeling tools has significant potential for applications across scales. Mapping crop yield gaps across Brazil highlights regions and farm types with the highest potential for yield improvement. The methods developed in Objectives 2 and 3 can be applied to lower the cost and increase the availability of sub-field yield maps to support site-specific field management, which can increase farm input use efficiency. The results here suggest that remote sensing will play an important role in identifying pathways leading to higher yielding and more resilient cropping systems.; Thesis (Ph.D.)--Tufts University, 2018.; Submitted to the Dept. of Agriculture, Food and Environment.; Advisor: Timothy Griffin.; Committee: David Fleisher, Magaly Koch, and Elena Naumova.; Keywords: Agriculture, Remote sensing, and Sustainability. %[ 2022-10-11 %~ Tufts Digital Library %W Institution