Developing Remote Sensing Protocols for Inventory of Nutrient Management Practices
An accurate inventory of in-field and off-site nutrient management practices is essential to establish baseline conditions and document implementation of the Iowa Nutrient Reduction Strategy. For some practices, such as crop rotation, available satellite sensor data already has been processed into crop rotation datasets. For most practices — cover crops, residue cover, riparian buffers, flow or erosion control structures — more research is needed to refine methodologies and calibrate results.
This project will develop standard remote sensing protocols to inventory cover crops and residue cover. These protocols will be validated in the Turkey River, Cedar River and Skunk River watersheds.
Remote detection of cover crops will use Landsat-7 and 8 satellite sensor data, plus other high resolution/high repeat datasets. Landsat imagery will be acquired in late fall and spring to detect greenness at each of these times, and used to determine the presence of a cover crop over the winter period.
Estimates of residue cover at the field scale will be evaluated using Landsat-7 and 8 satellite sensor data to produce a normalized difference tillage index (NDTI). The NDTI is a promising methodology to estimate residue cover prior to crop emergence.
Cover crops - As cover crops green up earlier in the spring, it is possible to detect this signature in a time series of satellite-derived vegetation robustness measures. However, the differences in imagery between cover crop and no cover crop are subtle and depend on planting date, weather, and other factors, this approach is feasible but not easy. Accurate assessments will require increases in observation frequency beyond what is currently feasible via freely available sources.
Residue Cover/Tillage - Residue cover levels are detectable via remote sensing due to differences in reflectance between soil and residue cover, primarily in the shortwave infrared. However, detection of these changes requires developing residue cover reference values across different soil color types and moisture contents, which can be a labor intensive process. Accurate assessments can be developed but will require increased satellite observation opportunities or models that account for how soil color and moisture vary over time and space to better leverage the few opportunities currently available. Measuring how tillage practices bury residue cover allows one to estimate tillage practices by measuring residue cover after planting, when tillage has ceased.
Other activities and accomplishments:
- Maps of residue cover for Iowa for 2015, 2016, and 2017 were developed using the residue cover surveys conducted as part of this project.
- 3 presentations
- Development of a remote sensing protocol for inventorying cover crop adoptions, M.S. Thesis, Carolina Bermudez
- This work, in small part, helped the state of Iowa obtain the 'Iowa Watershed Approach' grant from the Department of Housing and Urban Development. See info here: https://iowawatershedapproach.org/about/ The information acquired via this project helped support developing statewide residue cover maps using the survey data collected via this project.
Residue cover photograph classification and data assignment to field boundaries has been completed. Work to develop the ArcGIS and Python script that will enable evaluating measured residue cover versus the resulting satellite imagery/residue index correlations is still underway. Extra cover crop model validation work continues as well. The IDALS cover crop cost share data has been transferred to the PLSS (US Public Land Survey System) QQQ (Quarter Quarter Quarter; 10 acre squares) polygon boundaries. Work is underway to calculate the NDVI (Normalized Difference Vegetation Index) values for these polygons to further validate the cover crop detection techniques developed and supported by this project.
Residue cover photograph classification has been completed and the data now must be assigned to field boundaries. This will make it possible to evaluate measured residue cover versus the resulting satellite imagery/residue index correlations. This process will be done for multiple imagery dates across data from 2014, 2015, and 2016. Validation work on alternative cover crop detection techniques using remote sensing data will continue through the academic year.
Collection of spring 2016 residue cover photographs was completed in early June. Analysis of the resulting satellite imagery/residue index correlations and residue cover photos began when students returned to campus in late August. Validation work on alternative cover crop detection techniques using remote sensing data is underway. This will use additional cover crop location data obtained from IDALS cover crop cost share.
During this quarter, work continued to investigate and account for sources of error in residue cover and cover crop estimation procedures. Additional residue cover photographs were collected for 2016 in more than 10 Iowa counties, and photographs are being manually analyzed for residue cover. This analysis should be complete by the end of 2016 and final accuracy analysis of residue cover classification will begin in early 2017. For cover crops, the project analysis is complete. The major finding was that cover crops can be reliably detected better than 80 percent of the time using a classification tree approach with monthly median field NDVI values. The NDVI (normalized differenced vegetation index) is an indicator that describes the greenness of vegetative covers, and it is sensitive to percentage of biomass, leaf size and healthiness of vegetation.
Metadata creation continued for an additional eight baseline conservation practices and riparian vegetation geodatabases. A private company was contracted to fly five watersheds in early April collecting four-band data. Comparing the baseline data to this current imagery and updating the database with practice changes will be the next step. Project staff members are investigating a semi-automated method to extract vegetated areas from color infrared imagery and then use ancillary data to remove areas not relevant to this project, such as road ditches, fence lines, and farm grounds. Ancillary datasets also are being identified to help locate grassed waterways from vegetated polygons by the use of different soil attributes.
Additional work was undertaken this quarter to further investigate sources of error in residue cover and cover crop estimation procedures. The automated script developed last quarter enabled further investigations into both residue cover and cover crops. For residue cover, investigations determined that defining crop residue types is not straightforward from the majority land cover in the Ag Conservation Planning Framework (ACPF) field boundary database. Work is underway to refine the ACPF crop cover types to split fields with multiple residue types. For cover crops, the year-to-year variability complicates setting a consistent vegetation index threshold across dates, as does the difference in spectral response from fall- and spring-planted cover crops. Late fall detection is particularly challenging, as variable seeding dates and weather result in highly variable cover crop establishment. This makes it difficult to discriminate fields with cover crops from those with bare soil or some residue cover. Spring detection still shows promise, but will be complicated by variability of cover crop planting dates and/or how they over-winter.
During this quarter a source for ground truth data on cover crops in the Middle Cedar River basin was found and is proving useful in developing cover crop identification thresholds. Residue cover data from 1,000 fields in the South Fork of the Iowa River watershed collected by Agren Inc. in 2013 was converted into a usable form. Also residue cover data on approximately 150 fields across central, east central and southeast Iowa was collected and is being analyzed. The team has begun developing a program to automatically calculate field average vegetation and tillage indexes for classifying residue cover and cover crop status. This will work on data previously added to a GIS facility file server.
Iowa LandSAT imagery from 2000-2014 that are 50% cloud cover or less has been downloaded and processed. These data are available on a GIS facility file server. Several viable sources for ground truth data on cover crops have been identified within the Middle Cedar basin, and are expected to be available this spring. The team is beginning to analyze 2013 ground truth data on residue cover in the South Fork watershed. Collection of additional ground truth residue cover measurements in the Water Quality Initiative watersheds will be done this summer.
This project is developing remote sensing protocols to inventory cover crops and residue cover. Two students are gathering all Iowa LandSAT imagery from 2000-2014 that are 50% cloud cover or less. A data management system to archive and organize this data on local servers is being built. This will allow the data to be accessible and sorted by team members. Several viable sources for ground truth data on cover crops have been identified within the Middle Cedar basin, which will be the first study area for mapping adoption of cover crops. The team also is working with partners to get ground truth data on residue cover in the South Fork watershed, which will be the first study area for the residue/tillage mapping.
An accurate inventory of in-field and off-site nutrient management practices is essential to establish baseline conditions and document implementation of the Iowa Nutrient Reduction Strategy. For some practices, such as crop rotation, available satellite sensor data already has been processed into crop rotation data sets. For most practices, more research is needed to refine methodologies and calibrate results. This project will develop standard remote sensing protocols to inventory cover crops and residue cover. A data structure to store the thousands of satellite images required to map cover crops, residue cover and vegetative buffers has been developed. Work has begun to download required imagery from the U.S. Geological Survey.