Policy makers, government planners and agricultural market participants in Pakistan require accurate and timely information about wheat yield and production. Punjab Province is by far the most important wheat producing region in the country. The manual collection of field data and data processing for crop forecasting by the provincial government requires significant amounts of time before official reports can be released. Several studies have shown that wheat yield can be effectively forecast using satellite remote sensing data. In this study, we developed a methodology for estimating wheat yield and area for Punjab Province from freely available Landsat and MODIS satellite imagery approximately six weeks before harvest. Wheat yield was derived by regressing reported yield values against time series of four different peak-season MODIS-derived vegetation indices. We also tested deriving wheat area from the same MODIS time series using a regression-tree approach. Among the four evaluated indices, WDRVI provided more consistent and accurate yield forecasts compared to NDVI, EVI2 and saturation-adjusted normalized difference vegetation index (SANDVI). The lowest RMSE values at the district level for forecast versus reported yield were found when using six or more years of training data. Forecast yield for the 2007/2008 to 2012/2013 growing seasons were within 0.2% and 11.5% of final reported values. Absolute deviations of wheat area and production forecasts from reported values were slightly greater compared to using the previous year's or the three- or six-year moving average values, implying that 250-m MODIS data does not provide sufficient spatial resolution for providing improved wheat area and production forecasts.

Wheat is a staple food of Pakistan’s population of more than 180 million. The average annual harvest of 24 million tonnes places Pakistan eighth in the world in wheat production. Roughly, 76% of wheat is produced by Punjab province, the breadbasket of Pakistan. The current wheat area and yield reporting system operated by the Punjab Crop Reporting Service (CRS) delivers estimates several months after harvest. The delayed production data cannot contribute to in-season decision support system, such as timely production estimation. However, freely available satellite imagery at medium spatial resolution offers a data source for addressing this limitation. Using Landsat data from the 2013 to 2014 winter wheat growing season and a supervised bagged classification tree approach, we estimated the area of wheat cultivated in Punjab. Validation consisted of a two-stage probability-based cluster field sample. Field data from 110 sample points distributed among twelve 20 km × 20 km blocks were collected. Overall accuracy of the Landsat-based wheat map was 76% with wheat commission error of 38% and wheat omission error of 30%. The Landsat-based map totalled 6.13 Mha of wheat area, and the field-based sample estimate 6.96 ± 2.15 Mha. Both figures are comparable to the official Punjab CRS figure of 6.77 Mha. The study presents a robust approach for early wheat area estimation providing information relevant to decision support systems concerning food production and trade.

The Planet constellation of satellites represents a significant advance in the availability of high cadence, high spatial resolution imagery. When coupled with a targeted sampling strategy, these advances enhance land-cover and land-use monitoring capabilities. Here we present example regional and national-scale area-estimation methods as a demonstration of the integrated and efficient use of mapping and sampling using public medium-resolution (Landsat) and commercial high resolution (PlanetScope) imagery. Our proposed method is agnostic to the geographic region and type of land cover and change, which is demonstrated by applying the method across two very different geographies and thematic classes. Wheat extent is estimated in Punjab, Pakistan, for the 2018/2019 growing season, and tree-cover loss area is estimated over Peru for 2017 and 2018. We used a time series of PlanetScope imagery to classify a sample of 5 × 5 km blocks for each region and produce area estimates of 55,947 km² (±9.0%) of wheat in Punjab and 5398 km² (±9.1%) of tree-cover loss in Peru. We also demonstrate the use of regression estimation utilizing population information from Landsat-based maps to reduce standard errors of the sample-based estimates. Resulting regression estimates have SEs of 3.6% and 5.1% for Pakistan and Peru, respectively. The combination of daily global coverage and high spatial resolution of Planet imagery improves our ability to monitor crop phenology and capture ephemeral tree-cover loss and degradation dynamics, while Landsat-based maps provide wall-to-wall information to target the sample and increase precision of the estimates through the use of regression estimation.

Time series of multi-spectral satellite data are a valuable option for crop-area monitoring, in particular by enabling rapid, accurate, and consistent estimation of crop area. In this study, we demonstrate the use of a turn-key wheat area mapping model using growing season Landsat time-series data to improve the precision of area estimates produced from a sample of field locations at which the presence of percent wheat per sample pixel was determined. The strata used in the sampling design to obtain the current season’s in-situ reference data were created from a wheat map of the previous growing season. We then used a map for the current growing season to serve as an auxiliary variable in a regression estimator of the wheat area. This approach allows for rapid production of such maps and area estimates with less demand for technical expertise in an ongoing wheat monitoring system. Our automated, turn-key classification-tree algorithm produced wheat maps with comparable overall accuracy to those achieved by traditional supervised classification using manual training data collection. Pixel counts from turn-key maps for the years 2013 to 2017 over-estimated wheat area by 0.86% in 2014 to 7.26% in 2017 when compared to the official estimate from the Punjab Crop Reporting Service. Of the several approaches to estimating wheat area were compared, stratified random and simple random sampling augmented by a regression estimator were the two that yielded the smallest standard errors. Incorporating a current season, turn-key map in a regression estimator with data from a simple random sampling design represents a methodologically straightforward, operational implementation of timely, accurate, and precise monitoring of wheat area. Given the importance of wheat production in Punjab to Pakistan’s national economy, this method represents an operationally efficient and effective monitoring of inter-annual shortages or surpluses in support of grain management. An automated, turn-key winter wheat mapping algorithm enables timely area estimation in advance of harvest.