1 / 26

Global GEOGLAM

Global GEOGLAM. Primary objective is to leverage existing systems Challenges to intercomparison How do we get an apples-to-apples comparison What about a generic approach applied globally? Example of area by crop type, using free and available data (Landsat, MODIS)

kimberley-curry
Download Presentation

Global GEOGLAM

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Global GEOGLAM • Primary objective is to leverage existing systems • Challenges to intercomparison • How do we get an apples-to-apples comparison • What about a generic approach applied globally? • Example of area by crop type, using freeand available data (Landsat, MODIS) • Sampling, not mapping, to overcome limitations of Landsat and to minimize workload • Area estimation by key crop type, including soybean, maize, wheat and rice • No waiting for next capability or interoperability

  2. Sampling Strategy Preference is for wall to wall coverage BUT this is not feasible for all GEOGLAM regions, with high acquisition frequency, especially given cloud cover during growing season. A nested stratified, multi-resolution sampling approach is an alternative which allows for more frequent acquisitions over selected sites that are statistically representative of entire area • Coarse resolution MODIS indicator maps for stratification, if needed • Moderate spatial resolution sample blocks • Number of sample blocks depends primarily on variability of crop types, of crop rotations, size of region, desired standard error • Blocks should be quite small (5-20km) • Frequency of acquisition, will depend on the complexity of the cropping system- approximately 5 scenes per growing season; 1-2 out of season • Very high spatial resolution data for cal/val - Smaller Subset of sample blocks requested for fine res with same acquisition frequency as moderate resolution blocks

  3. Sampling: Example of proposed sampling approach to global soy area estimation Study area, consisting of the top four soybean production countries, nearly 90% of global soybean production. Top left, United States; top right, Brazil; lower left, Argentina; lower right, China. For each country, the administrative subset shown accounts for over 95% of national soybean acreage, except for China, where the subset shown represents 88% of national soybean acreage. Hansen et al

  4. Objectives • Employ multiple resolution data: MODIS, Landsat, RapidEye and field data to estimate national-scale crop area by type • Test a generic approach to estimate cultivated soybean area in the USA, Brazil, Argentina and China, which account for almost 90% of global soybean production • Illustrate the viability of remote sensing-based global crop type area estimation using a sampling approach

  5. Method • MODIS used for turn-key, generalized models to generate per nation/sub-region to indicate within growing-season soybean cultivation based on sub-pixel percent cover training data • The models estimate percent soy-cover and enable the stratification of national-scale cropland growing regions for sampling purposes • S1- Landsat samples used to map per sample block soybean cultivated area • S2- RapidEye allows for per country/region calibration of Landsat area estimates • The Landsat sample blocks are then analyzed to quantify national-scale crop type area

  6. High, medium and low soybean stratausing MODIS Red=high (>19.8%), orange=medium (7.2-19.8%), yellow=low (0.5-7.2%)

  7. Landsat sample blocks (S1)3-4 acquisitions during growing season Red=high (>19.8%), orange=medium (7.2-19.8%), yellow=low (0.5-7.2%)

  8. RapidEye sample blocks- S2 Red=high (>19.8%), orange=medium (7.2-19.8%), yellow=low (0.5-7.2%)

  9. Iowa – Corn Belt July 2, 2011 - LT50260312011183PAC01

  10. Iowa – Corn Belt August 3, 2011 - LT50260312011215PAC01

  11. Iowa – Corn Belt Sept 4, 2011 - LT50260312011247PAC01

  12. Iowa – Corn Belt Soybean agreement CDL soybean / UMd no soybean CDL no soybean / UMd soybean Soybean agreement – CDL winter wheat double crop CDL soybean / UMd no soybean CDL winter wheat double crop

  13. SW Minnesota Landsat 5-4-3 24km x 20km, centered on 95 35 24W, 44 19 38N

  14. SW Minnesota RapidEye 4-5-3

  15. RapidEye – 140 acres, Landsat 126 acres

  16. 17 soybean states USDA 2011 total = 285k sq. kilometers planted 281k sq. kilometers harvested CDL as sample = 271k sq. kilometers +/- 9k sq. kilometers Landsat percent soybean 2011 MODIS percent soybean 2011

  17. Calibration of Landsat with Rapid EyeRapid Eye vs. Landsat Area estimates per Block Preliminary Results Comparison of Area Estimates using Landsat sample blocks vs. wall to wall Landsat based estimate Soybean samples blocks – CDL vs. RE Soybean samples blocks – UMd vs. CDL CDL Hansen et al

  18. Argentina strata • Preliminary result: 160,030 km2 with a standard error of 10,600 km2 Red=high (>19.8%), orange=medium (7.2-19.8%), yellow=low (0.5-7.2%) Hansen et al

  19. Argentina field data collection over selected sample blocks (Feb 2012)

  20. Summary • Freely available data can be used for prototyping global monitoring tasks of GEOGLAM • Viable estimates of crop condition, area and yield can be made • These will not relate 1:1 with existing estimates, but will be internally consistent • As newer systems come online, products can be improved • Sampling is a low-cost entry point for global monitoring • Especially given that EO data are not regularly used for national crop monitoring • Does not require wall to wall coverage and is tolerant of missing data (clouds, scan-line gaps) • A couple of graduate students or interns can be trained to implement methods per country – minimize work augmentation within agencies

More Related