Punjab in Pakistan is the major food producing province and titled “bread basket” for the country’s more than 220 million people. Wheat is the major cereal crop produced in Punjab, putting Pakistan 8th among the wheat producing countries. Estimation of wheat area and production before harvest is important input for key operational decision making on post-harvest crop marketing and ensuring food security of dependent populations. As part of the Harvest program, UMD’s Global Land Assessment and Discovery (GLAD) team has been working on the characterization of wheat crop in Punjab and has developed a turn-key operational model capable of producing pre-harvest wheat area.
In collaboration with the Crop Reporting Service of Punjab, Ahmad Khan led a field campaign this year, with research extending over two months. This was a three-stage simple random sampling of the Punjab province with 25 sample blocks of 5 km x 5 km randomly selected in the first stage. From each of these sample blocks, 10 pixels were then randomly selected and from each of the sample pixels, a variable number of 2 m x 2 m crop cuts were selected. The number of crop cuts was based on proportion of wheat cover in a pixel assessed in the field, ranging from 0 for no wheat cover to 5 for a sample pixel fully covered with wheat.
The field sampling was composed of two campaigns, the first being identification of the sample pixels and assessment of the crop cuts with delineation of each crop cut. The second campaign was collection of wheat from each of the delineated crop cuts, separating grain from the wheat kernel collected from each of the sample pixels, and weighing the grain to obtain average wheat produced per sample pixel. Field data reveal a yield estimate of 20.04 million tonnes for Punjab, about 3% higher than Punjab Crop Reporting Service’s estimate of 19.37 million tonnes. Using multi-year field data as a calibration input, the GLAD team will next develop a Landsat-based model to estimate wheat yield and production for Punjab. The field samples will be used as a reference for measuring the accuracy of the Landsat-based model.
Preliminary results of the field data show that this sampling - which is simpler in application, easier logistically, and quicker in implementation - can provide wheat yield estimates that are comparable to the official statistics. The Crop Reporting Service of Punjab is interested in evaluating these methods and integrating a remote sensing based approach to their crop statistics. As part of the Harvest program, the GLAD team is continuing to work with stakeholders around the world to increase the use of Earth observations in agricultural monitoring.