University of Twente Student Theses

Login

Integrating earth observation data into area frame sampling approach to improve crop production estimates

Dogar, S.S.S. (2022) Integrating earth observation data into area frame sampling approach to improve crop production estimates.

[img] PDF
5MB
Abstract:Accurate and reliable agricultural statistics are crucial for understanding the current crop dynamics and improving food security, especially in developing countries. Yield gap analysis provides insights into crop dynamics across the agricultural landscape. It lays the foundation to identify yield constraint factors within fields and improve practices to close the yield gap. However, in many developing countries, current agricultural surveys are established using administrative boundaries and do not reflect the country's agricultural landscape. The most common survey approach is Area Frame Sampling which typically uses Admin areas as primary stratification and does not incorporate AEZ. This research adopts a hybrid approach to identify site-specific crop yield variability and extrapolate it to area-specific crop production estimates by combining statistical and open-source earth observation data. To identify yield constraint factors and quantify crop production function, 503 site-specific wheat yield samples from Punjab, Pakistan, were analysed using Comparative performance analysis (CPA). Long-term NDVI climatology of 20 years is used to capture the agro-climatic conditions over a complex and fragmented agricultural landscape. ISODATA unsupervised classification is used to identify crop phenological cycles and produce Crop Production System Zones (CPS zones). Regression analysis is used to assess the relationship of site-specific measured yield with CPS zones and admin areas. Results revealed that site-specific field parameters explained 41.2 percent of the yield variability. CPS zones based on Earth observation approach explained 23.3 percent of yield variability. A combination model was developed and evaluated to determine the combined impact of site-specific factors and CPS zones. The final model derived through stepwise multiple linear regression included two CPS zones (one from irrigated zone and second from a rainfed zone). This final model explained 43.2 percent of the yield variability. The main findings of this research were as follows: i) UREA fertilizer, broadcast sowing pattern and seed treatment were discovered to be an important field parameters (i.e., explained 28, 19.5 and 14.4 percent of the deviance), ii) Long-term NDVI identified clear crop phenological cycle exists in the study area, iii) CPS zones could differentiate between different rainfed and irrigated croplands. Overall, the study's findings and comparisons support the premise that hyper-temporal earth observation can effectively capture climatological changes in a fragmented agricultural landscape to identify crop yield variability and improve crop production estimates. The method can be applied by government departments and researchers for further studies and aid in decision-making related to closing yield-gap, cropping and food security goals.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Subject:48 agricultural science
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:https://purl.utwente.nl/essays/90845
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page