Automatically Determined Management Unit Evaluation

The other new factor in this experiment, automatically determined management unit boundaries, appears to suitably define the area over which residue cover estimates should be made. A comparison of r2 values and RMSEs calculated using both the automated

methodology and manual delineation of boundaries showed that manual delineation provided slightly less accurate estimates (0.02 lower r2 and 0.005 higher RMSE) than the automated methodology. Although the reason for better estimates is not known, the differences are small. Positive results were expected based on the research by Gelder et al. (2007), creating the possibility for great time savings and routine determination of mean field residue cover.

generally either time consuming and expensive or lacking in accuracy and consistency. Previous implementations of Landsat crop residue indices have shown mixed results in residue detection ability due to soil color and green vegetation influences on indices and required manual delineation of field boundaries to compute field average residue cover. These complications have limited previous implementations of residue indices to accurately and efficiently determine residue cover.

Additional research into the influence of green vegetation has shown that residue cover indices can be used to estimate residue cover on the glacial till derived soils of central Iowa both before and after the emergence of green vegetation. Green vegetation obscures the response of residue cover indices to residue cover, decreasing the accuracy as the image date increased from early April to early June. Classification of fields into pre- and post-emergent classes permits better analysis of response, with indices utilizing combinations of Band 7 and 3 or 5 performing better than other indices. The NDTI and CRIM5,7 index provided the best

estimates before the emergence of green vegetation with the indices explaining about 85% of the response before emergence and 51% after emergence. The NDRI and CRIM3,7 indices

also performed well before emergence with an r2 of 0.72 and 0.56, but coefficients of

determination decreased after emergence to 0.52 and 0.55, respectively. Indices using Band 4 and 7, the NDI7 and CRIM4,7 performed well before emergence, but performed worse than

the NDTI, NDRI, CRIM5,7, and CRIM3,7 after emergence. Indices utilizing combinations of

Bands 3, 4, and 5 were the worst at estimating residue cover, especially after plant emergence. These results follow that expected from spectral response of the scene

components, as Band 7 shows absorption due to cellulose and lignin not present in soils and Band 4 shows a significant increase in reflectance with green vegetation.

Band indices selected to be more resistant to the effects of green vegetation, such as the NDRI, showed the same impacts of green vegetation as the NDTI and CRIM5,7, with

increasing error in residue cover estimates with increasing NDVI. The error in both NDTI and CRIM5,7 increases consistently with NDVI, meaning that it is possible to remove some of

the effect of green vegetation on the index. Subtraction of the regression of index error vs NDVI from the index value increased the r2 of both the NDTI and CRIM5,7 by about 0.10 and

decreased the RMSE by 0.05, although r2 values were still 0.20 below and RMSE were still 0.03 above those on pre-emergent fields.

These results illustrate the need for timely remote sensing observations to reduce the impact of green vegetation on scene reflectance or the use of correction factors to reduce these effects. It may also be necessary to use multiple imagery dates to optimally estimate residue cover where plants with different emergence dates are routinely planted. When selecting imagery sources, it should be noted that the 2.08-2.35 um spectrum sampled by Landsat Band 7 is not measured by most other operational satellite remote sensing

instruments and is a component of the indices most resistant to the distorting effects of green vegetation.

The use of pre-determined management unit boundaries was also tested in this paper and found to be of suitable accuracy. The automated methodology of Gelder et al. (2006) returned estimates that were slightly more accurate than estimates obtained using manually delineated field boundaries. The use of this methodology for delineating field boundaries will significantly increase the rate at which remote estimates of residue cover can be made and enable routine measurements. Due to the inaccuracies involved in residue measurement and mistrust of “black box” remote sensing estimates, this methodology for producing residue cover estimates is presently best used as a screening tool to determine areas that either do or do not require further field investigation of residue cover. These limitations,

3.6 References

Arsenault, Eric, and Ferdinand Bonn. 2005. Evaluation of soil erosion protective cover by crop residues using vegetation indices and spectral mixture analysis of multispectral and hyperspectral data. Catena. 62:157-172.

Baret, F., S. Jacquemoud, and J.F. Honocq. 1993. The soil line concept in remote sensing. Remote Sensing Reviews. 7(1):65-82.

Biard, F., A. Bannari, and F. Bonn. 1995. SACRI (Soil Adjusted Crop Residue Index): an indice utilisant le proche et le moyen infrarouge pour la détection des residues de cultures de mais. In Proc. 17th

Canadian Symposium on Remote Sensing, 417-423. Ottawa,

Canada: Canadian Remote Sensing Society.

Biard, F. and Frederic Baret. 1997. Crop residue estimation using multiband reflectance. Remote Sensing of Environment. 59(3):530-536.

Blough, R.F., A.R. Jarrett, J.M. Hamlett, and M.D. Shaw. 1990. Runoff and erosion rates from slit, conventional, and chisel tillage under simulated rainfall. Transactions of the

American Society of Agricultural Engineers. 33:1557-1561.

Conservation Technology Information Center. 2004. National Crop Residue Management Survey. West Lafayette, IN: Conservation Technology Information Center.

Curran, P. J. 1989. Remote sensing of foliar chemistry. Remote Sensing of

Environment. 30:271-278.

Daughtry, C. S. T. 2001. Discriminating crop residue from soil by shortwave infrared reflectance. Agronomy Journal. 93(1):125-131.

Daughtry, C.S.T., E.R. Hunt, Jr., P.C. Doraiswamy, and J.E. McMurtrey, III. 2005. Remote sensing the spatial distribution of crop residues. Agronomy Journal. 97(3):864-871.

Elvidge, C. D. 1990. Visible and near infrared reflectance characteristics of dry plant materials. International Journal Remote Sensing. 10:1775-1795.

ERDAS. 2005. IMAGINE 9.0. St. Gallen, Switzerland: Leica Geosystems Geospatial Imaging.

Gausman, H. W., A. H. Gerbermann, C.L. Wiegand, R.W. Leamer, R.R. Rodriguez, and J.R. Norieaga. 1975. Reflectance differences between crop residues and bare soils. Soil

Science Society of America Proceedings. 39:752-755.

Gelder, B. K., R.M. Cruse, and A.L. Kaleita. 2007. Chapter 2: Automated Determination of Management Units for Precision Conservation. In Land management

database development: Methods for delineating management units and estimating crop and residue cover. PhD Dissertation. Ames, IA: Iowa State University, Department of

Agronomy.

Gowda, P.H., B.J. Dalzell, D.J. Mulla, and F. Kollman. 2001. Mapping tillage practices with Landsat Thematic Mapper based logistic regression models. Journal of Soil

and Water Conservation. 56(2):91–96.

Halvorson, G.D., G.A. Peterson, and C.A. Reule. 2002. Tillage system and crop rotation effects on dryland crop yields and soil carbon in the Central Great Plains. Agronomy

Journal. 94:1429-1436.

Ketcheson, J.W., Stonehouse, D.P., 1983. Conservation tillage in Ontario. Journal of

Soil and Water Conservation. 38:253–254.

Laflen, J.M., M. Amemiya, and E.A. Hintz. 1981. Measuring crop residue cover.

Journal of Soil and Water Conservation. Vol. 36(6):341-343.

Moorman, T. B., C.A. Cambardella, D.E. James, D.L. Karlen, and L.A. Kramer. 2004. Quantifcation of tillage and landscape effects on soil carbon in small Iowa watersheds.

Natural Resource Conservation Service. 2006. Natural Resources Inventory 2003 Annual NRI: Soil Erosion. Washington, DC: U.S. Department of Agriculture.

Pimentel D., C. Harvey, P. Resosudarmo, K. Sinclair, D. Kurz, M. McNair, S. Crist, L. Shpritz, L. Fitton, R. Saffouri, and R. Blair. 1995. Environmental and economic costs of soil erosion and conservation benefits. Science. 267(5201):1117-1123.

Richardson, A.J., and C.L. Wiegand. 1977. Distinguishing vegetation from soil background information. Photogrammetric Engineering and Remote Sensing. 43(12):1541- 1552.

Rouse, J.W., R.H. Haas, Jr., J.A. Schell, and D.W. Deering. 1974. Monitoring vegetation systems in the Great Plains with ERTS. In Proc. Third ERTS-1 Symp., 309-317. Greenbelt, MD: NASA SP-351.

Sullivan, D.G., C. C. Truman, H. H. Schomberg, D. M. Endale, and T. C. Strickland. 2006. Evaluating techniques for determining tillage regime in the Southeastern Coastal Plain and Piedmont. Agronomy Journal. 98:1236-1246.

Thoma, D.P., S.C. Gupta, and M.E. Bauer. 2004. Evaluation of optical remote sensing models for crop residue cover assessment. Journal of Soil and Water Conservation. 59:(5):224-233.

van Deventer, A.P., A.D. Ward, P.H. Gowda, and J.G. Lyon. 1997. Using Thematic Mapper data to identify contrasting soil plains and tillage practices. Photogrammetric

Engineering and Remote Sensing. 63(1):87-93.

Viña, Andres, Albert J. Peters, and Lei Ji. 2003. Use of multispectral Ikonos imagery for discriminating between conventional and conservation agricultural tillage practices. Photogrammetric Engineering and Remote Sensing. 69(5):537-544.

CHAPTER 4. RADAR DETERMINATION OF SURFACE