Managing Global Resources for a Secure Future

2017 Annual Meeting | Oct. 22-25 | Tampa, FL

49-9 Accuracy of in-Season Maize Yield Estimation with Unmanned Air Vehicle (UAV) in Smallholder Maize-Based Farms.

See more from this Division: ASA Section: Climatology and Modeling
See more from this Session: Agricultural Remote Sensing General Oral (includes student competition)

Monday, October 23, 2017: 11:15 AM
Tampa Convention Center, Room 5

Julius B. Adewopo, GIS/Natural Resource Management, International Institute of Tropical Agriculture, Kano, Nigeria, Alpha Y Kamara, R4D, International Institute of Tropical Agriculture, Ibadan, Nigeria, Bernard Vanlauwe, Natural Resource Management, IITA, Nairobi, Kenya and Peter Craufurd, Sustainable Intensification Program, International Maize and Wheat Improvement Center (CIMMYT), Nairobi, Kenya
Abstract:
Rapid, timely, and cost effective yield monitoring in cropping systems remains topical, especially in sub-Saharan Africa (SSA) where there is limited capability for quick and near real-time assessment of yield gap and yield constraints. Unmanned air vehicles (UAVs) offer promising leverage for acquisition of low-altitude remotely-sensed data on terrain and vegetation properties, including Normalized Difference Vegetation Index (NDVI), which can provide information on plant health and productivity. We conducted flight missions with UAV, mounted with multi-spec4C sensor, to estimate in-season maize yield (before harvest) within smallholder farmers’ fields in Nigeria. Within a flight area (100ha) that covers farmer-managed fields (FMF) and nutrient omission trial (NOT) plots, fine-resolution imageries were acquired, post-processed, and analyzed. Relationships between ground-measured and UAV-derived NDVI at 4 and 8 weeks after planting (WAP) were evaluated. Actual maize yield data from the NOT plots was used to calibrate UAV-derived NDVI (as a potential proxy for yield estimation), and a spatially-explicit pixel-based yield prediction was conducted for the flight area covered. Results show good relationship between UAV-derived and measured NDVI in NOT plots at 8WAP (r2=0.61; p=0.002) and in FMF (r2=0.40, p<0.001). The strong relationship between measured yield and UAV-derived NDVI within NOT (r2=0.62, p=0.002) suggests that UAV-derived NDVI can be used as proxy for yield estimation. Therefore, further spatial analyses of yield in NOT plots showed that UAV-based average yield estimate (4.51t/ha) was close to actual yield (4.44t/ha). Despite systemic over-estimation of UAV-derived yield compared to actual average yield in FMF (3.8t/ha vs. 2.94 t/ha), the strong relationship (R = 0.93 and r2= 0.82; p=0.02) indicates the potential to filter out artefacts, such as weeds prevalence, which can inflate remotely-sensed NDVI values within FMF. Overall, this evidence present a promising opportunity for in-season estimation of maize yield within heterogeneously-characterized smallholder farms.

See more from this Division: ASA Section: Climatology and Modeling
See more from this Session: Agricultural Remote Sensing General Oral (includes student competition)