MODIFICATION
R -- Planning & Coordination of Remote Sensing Projects Related to the Production of Crop Acreage Estimates
- Notice Date
- 8/29/2005
- Notice Type
- Modification
- NAICS
- 541370
— Surveying and Mapping (except Geophysical) Services
- Contracting Office
- Department of Agriculture, Agricultural Research Service, Acquisition and Property Division, Acqusition Branch (DC), 1280 Maryland Avenue, SW Suite 580C, Washington, DC, 20024-2142
- ZIP Code
- 20024-2142
- Solicitation Number
- 05-3K06-570
- Response Due
- 9/6/2005
- Archive Date
- 9/21/2005
- Point of Contact
- Stephen Schaefer, Contract Specialist, Phone 202-720-7365, Fax 202-720-3987, - Anthony Wimbush, Supervisory Contract Specialist, Phone (202) 720-3998, Fax (202) 720-3987,
- E-Mail Address
-
sschaefer@ars.usda.gov, twimbush@ars.usda.gov
- Small Business Set-Aside
- Total Small Business
- Description
- Amendment 01 of Solicitation 05-3K06-570 I. Questions & Responses Q1: What are the crop classes of interest? R1: The crop classes are determined by whatever the majority crop covers in the state are. For instance, the majority covers in IL/IN/IA are corn, soybeans and winter wheat. In North Dakota the majority covers include allwheat, spring, durum and winter wheats, barley, beets, canola, sunflower, corn, drybeans, oats, potatoes, rye, soybean and flaxseed. Q2: What level of land cover identification is desired? R2: The NASS June Agricultural Survey defines the categories that NASS uses for training. Additional categories can be created using partner?s datasets of extra training sets, or the analyst can derive additional signatures using manual crop identification techniques. Q3: Is clubbing together of similar crop areas allowed? R3: NASS is not sure of the meaning of clubbing, but its analysis method takes all of the usable training fields located within an analysis district, or multiple satellite scene footprint area with the same observation date. These training fields are then separated out by cover type and clustered individually. This provides unique training signatures for each cover type. The signatures are then appended together for classification and accuracy assessment. Q4: Which satellite images are used? R3: Landsat 5/7, ResourceSat1 AWiFS/LISS Q5: Can we have images from 2 seasons: one with standing crop, and another without crop? R5: NASS prefers to have two dates of imagery to separate the non ag cover types with the planted ag. Spring followed by mid-summer is preferred for our processes. In ag areas, the spring scenes of bare fields help separate out ag vs. non ag, and the crop?s spectral reflectance allows for differentiation of the mid-summer scenes. Q6: What techniques are used to identify the crops from the satellite images? R6: NASS uses the June Agricultural Survey (JAS) which is an enumerative survey that runs the first two weeks of every June. During the survey, selected sample sites are visited and the growers are queried on crop type and acreage, providing the necessary training data for the year. Follow-up surveys are a possibility in the case of a refusal or inaccessible location at the time on the JAS. These surveys provide the training data for our remote sensing program. Q7: Only spectral properties or other information are used? R7: The spectral reflectance properties are what drive our clustering/classification algorithms. A supervised modified ISODATA clustering algorithm is used for clustering, and a maximum likelihood classifier is used for classification. Q8: Will NASS be interested in testing the textural properties of the image for crop identification? R8: NASS is interested in researching the concept of textual properties as well as object-based classifiers, but operational demands of the program have kept the research of these new techniques in low priority research mode currently. Q9: How does the 'regression estimator' measure the acreage? R9: A regression estimation approach is applied using classified pixels as the independent variable and farmer reported acres as the dependent variable in a set of area frame stratum based simple linear regressions. The county and state crop acreage estimates produced by this system greatly reduce the sampling variation ('error') found in the estimates produced by the ground data alone; and have the added benefits of producing county estimates with measurable precision. Q10: Please direct us to the literature on the 'regression estimator.' R10: Day, C.D. (2002) "A Compilation of PEDITOR Estimation Formulas". RDD Research Paper RDD-02-03, USDA, NASS, Washington, D.C. January, 2002. Q11: What accuracy of classification is desired? R11: It is desirable to have accuracies greater than 85 percent overall, and in the mid-nineties if the imagery was collected during optimum times during the growing season. It is critical to have accuracies in the low to mid-nineties for the crops that we are estimating. It is also necessary, to have r-squares that are greater than .8, again it is image quality driven. Q12: Please direct us to the report on the accuracy values of the recent classification. R12: There are no accuracy values online, but there exists some image metadata online @ http://www.nass.usda.gov/research/Cropland/metadata/meta.htm Additionally, here is an example an accuracy assessment in Arkansas for 2004 over the Delta: SIGNATURES, PERCENT CORRECT AND KAPPA BY ANALYSIS DISTRICT - ARKANSAS 2004 (Percent Correct on Known Good Fields Only) ANALYSIS DISTRICT AD01 LANDSAT TM PATH: 23, ROW(S): 35, 36 & 37 - (05/16 + 08/10/2004) 253 CROP / COVER TYPE SIGNATURES, 14 CHANNELS MOSAIC CATEG# CROP / COVER * ORIG. # CATEG. ORIG. # PIXELS PERCENT CORRECT * COMMISSION ERROR KAPPA COEFFICIENT 92 AQUACULT 3 3536 90.05 12.98 89.87 81 CLOUDS 17 0 0.00 100.00 0.00 1 CORN 2 6902 75.76 5.73 75.10 2 COTTON 48 35651 94.50 6.98 93.34 62 FARM 0 7 0.00 0.00 0.00 81 FILLER 1 0 0.00 0.00 0.00 61 IDLE CROP 15 2713 84.70 20.95 84.49 62 NON AGG 6 15878 35.34 13.91 33.26 25 OTHER HAY 3 320 88.44 21.82 88.42 62 PERM PAST 2 468 71.15 32.04 71.09 3 RICE 19 49363 96.77 0.94 95.80 4 SORGHUM 10 1057 89.50 8.60 89.45 5 SOYBEANS 93 88231 96.12 6.33 93.15 53 STATE563 0 20 0.00 0.00 0.00 57 STATE567 2 221 99.55 0.45 99.55 59 STATE569 0 39 0.00 0.00 0.00 6 SUNFLRS 0 1 0.00 0.00 0.00 255 UNKNOWN 0 74 0.00 0.00 0.00 82 URBAN 1 225 94.67 39.66 94.66 83 WATER 6 561 100.00 70.19 100.00 24 WIN WHEAT 9 954 91.51 16.94 91.47 63 WOODPAST 0 59 0.00 0.00 0.00 63 WOODS 16 2717 93.67 74.09 93.36 OVERALL 253 208997 90.22 86.58 * NOTE: If signatures for covers such as CLOUDS or WATER were determined from pixels outside of the original ground truth sample, those cover types will have '0.00%' PERCENT CORRECT in this table. If a cover type named 'OTHER' exists, PERCENT CORRECT will also show as '0.00%' for the small area covers or crops that were combined into cover type 'OTHER'. Q13: Who will do the 'ground truthing' or 'verify the crop on the ground' for the spectral signatures? Contractor or the NASS staff? R13: Refer to R6. Q14: Please direct us to the recent NASS report on the crop classification using the satellite images and the methodology used. R14: Refer to the following websites: http://www.usda.gov/nass/nassinfo/remotehistory.htm http://www.nass.usda.gov/research/reports/02_09_CostBeneftCDL_Pecora15_GAH.pdf http://www.nass.usda.gov/research/reports/02_02_Compare_L5_L7_MEC.pdf http://www.nass.usda.gov/research/reports/02_01_CreateCDL_State_RWM.pdf http://www.nass.usda.gov/research/reports/99_04_AppropriateRoleRS_RCH.pdf Q15: Is this a new requirement or a recompetition of a current contract? If this is a recompetition, who is the incumbent contractor? R15: This is a recompetition. The incumbent contractor is Mr. Patrick Willis.
- Place of Performance
- Address: Various NASS Field Offices
- Country: USA
- Country: USA
- Record
- SN00881517-W 20050831/050829211624 (fbodaily.com)
- Source
-
FedBizOpps.gov Link to This Notice
(may not be valid after Archive Date)
| FSG Index | This Issue's Index | Today's FBO Daily Index Page |