Zheng_2012_Remote-Sensing-of-Environment

Remote sensing of crop residue cover using multi-temporal Landsat imagery

Baojuan Zheng a ,?,James B.Campbell a ,Kirsten M.de Beurs a ,b

a Department of Geography,Virginia Tech,115Major Williams Hall,Blacksburg,VA 24061,USA

b

Department of Geography and Environmental Sustainability,The University of Oklahoma,100E.Boyd Street,Norman,OK,USA

a b s t r a c t

a r t i c l e i n f o Article history:

Received 5August 2011

Received in revised form 26September 2011Accepted 29September 2011

Available online 4November 2011Keywords:Landsat

Remote sensing Crop residue Tillage

Normalized difference tillage index Multi-temporal

Tillage practices,which have direct impacts on soil and water quality,have changed dramatically during the past several decades.Tillage information is one of the important inputs for environmental modeling,but the availability of this information is still limited spatially and temporally.Previous studies have encountered dif-?culties in de ?ning reliable correlations between crop residue cover (CRC)and Landsat-based tillage indices because they neglected the signi ?cance of the timing of tillage implementation.This study explores relation-ships between temporal changes of agricultural surfaces and the normalized difference tillage index (NDTI)in Central Indiana.We found that minimum NDTI (minNDTI)values extracted from multi-temporal NDTI pro ?les reliably indicate the surface status when tillage or planting occurred.Simple linear regression reveals a coef ?cient of determination (R 2)of 0.89between CRC and minNDTI for calibration.In addition,a percent-age change (PC)method was tested for classifying CRC into three categories (CRC b 30%;30%b CRC b 70%;CRC >70%).Both the minNDTI and PC methods resulted in overall classi ?cation accuracies of >90%,produ-cer's accuracies of 83–100%,and user's accuracies of 75–100%.Our results indicated that both Landsat TM and ETM+imagery are capable of mapping CRC,however,multi-temporal Landsat imagery is required.To establish a capability for crop residue mapping,designers of future remote sensing platforms should consider increasing temporal resolution.

?2011Elsevier Inc.All rights reserved.

1.Introduction

Agricultural best management practices,such as conservation till-age and cover crops,have been adopted widely in recent years.The bene ?ts of conservation tillage are substantial,including improve-ment of soil and water quality,reduction of soil erosion,and maximi-zation of agricultural water use ef ?ciency (USDA-NRCS,2001).Reliable and systematic site-speci ?c conservation tillage data do not currently exist,but would form an important resource supporting the evaluation of the effectiveness of these practices.

Non-conservation tillage (intensive/conventional and reduced till-age)leaves less than 30%crop residue cover (CRC),while conservation tillage leaves more than 30%CRC (CTIC,2010).Current CRC data are not surveyed systematically and vary from one location to another.The USDA Natural Resources Conservation Service (NRCS)collects CRC data visually using a line-transect method (Morrison et al.,1993).The Conservation Technology Information Center (CTIC)provides assess-ments of conservation tillage practices,but collects data using annual roadside surveys of crop residue levels,which is subjective.Its tillage data are available at county,state,and regional levels.The county-level data were recently aggregated to 8-digit Hydrologic Unit (HU)wa-tersheds (Baker,2011).The National Agricultural Statistics Service (NASS)data relies on survey respondents and is only available at state

and county level.These inventory data are either too coarse (i.e.,they cannot provide ?eld level detail,nor report within-?eld spatial variabil-ity),or are inconsistent,adding more uncertainties in the environmen-tal modeling process.The spatial and temporal gaps in these inventory data restrict our ability to simulate the impact of crop management on water quality or carbon sequestration at broad scales (Jarecki et al.,2005;Saseendran et al.,2007).Thus,there is a strong need to develop methods to monitor agricultural practices over large areas,over time,using consistent methods.Multispectral remote sensing offers an op-portunity to systematically obtain information describing crop residues ef ?ciently and objectively over broad areas.

Early attempts to use remote sensing techniques for mapping CRC can be traced back to 1975(Gausman et al.,1975).Since then,the po-tential of remote sensing of crop residue has been investigated both in the laboratory and in the ?eld (Biard &Baret,1997;Daughtry,2001;Daughtry et al.,1995;Sullivan et al.,2007,2006).Remote sens-ing tillage indices,such as the crop residue index multiband (CRIM)(Biard &Baret,1997),the cellulose absorption index (CAI)(Daughtry,2001),and crop residue cover index (Sullivan et al.,2006)are designed in the laboratory to amplify the differences in the spectral signals between crop residues and soils (Table 1).Most tillage indices are based on the cellulose and lignin absorption features near 2100nm.Researchers have applied these tillage indices (Table 1)to airborne (Daughtry et al.,2005)and satellite remote sensing imagery (Daughtry et al.,2006;Gowda et al.,2003;Serbin et al.,2009a;Sullivan et al.,2008;Thoma et al.,2004;van Deventer et al.,1997).

Remote Sensing of Environment 117(2012)177–183

?Corresponding author.Tel.:+16467507087.E-mail address:baojuan5@58bd8d2b482fb4daa58d4bea (B.

Zheng).

0034-4257/$–see front matter ?2011Elsevier Inc.All rights reserved.doi:

10.1016/j.rse.2011.09.016

Contents lists available at SciVerse ScienceDirect

Remote Sensing of Environment

j o u r n a l h om e p a g e :ww w.e l s e v i e r.c o m /l o c a t e /r s e

These previous methods neglect an important factor—the timing of tillage or planting,which can vary greatly from?eld to?eld within even small regions.Three different surface conditions can coexist in a single image during the planting season(Fig.1):before tillage/planting(A), after tillage/planting with no or little vegetation(B&C),and crop emer-gence(D).Most?elds are under condition A at the early planting season and in condition D at the end of the planting season.If there are agricul-tural?elds tilled after an image was acquired,the previous methods would wrongfully designate these?elds as no-till.If crops have emerged,the green vegetation is likely to confound the residue cover estimation(Daughtry et al.,2005).Therefore,the methods previously outlined(i.e.,single image methods)could be problematic in predicting CRC and cannot be applied broadly.Watts et al.(2009)suggested that the use of higher temporal datasets might better capture surface distur-bances in minimum tillage?elds.Although Watts et al.(2011)pro-duced better classi?cation accuracy using a?ve-date Landsat model,a physical relationship between Landsat data and tillage categories was not de?ned.Instead they generated classi?cation models with the Ran-dom Forest classi?er.The objective of this study is to reveal the impor-tant role of temporal changes in CRC mapping,and to present a simple and objective method to map CRC using multi-temporal Landsat imagery.

2.Remote sensing imagery for crop residue detection

Accurate mapping of CRC not only requires remotely sensed data with spectral and spatial detail,but also with high temporal resolu-tion.Based on crop residue's unique absorption features near 2100nm(Daughtry,2001),past and current satellite remote sensing platforms capable of mapping CRC include Landsat5TM and7ETM+, EO-1Hyperion,the Advanced Spaceborne Thermal Emission and Re-?ection Radiometer(ASTER),and the Moderate Resolution Imaging Spectroradiometer(MODIS).Hyperion imagery,with a narrow swath width(7.5km),has low temporal coverage because its sensor

Table1

Satellite-based tillage indices.

Sensor Tillage index Formula Description Reference

Landsat CRIM SM/SR SM:distance from point M to the

soil line;SR:distance between

soil and residue lines at point M

Biard&Baret,1997

Simple tillage index(STI)B5/B7B2:Landsat TM/ETM+band2;B4:

TM/ETM+band4;B5:TM/ETM+

band5;B7:TM/ETM+band7;Van Deventer et al.,1997

NDTI(B5?B7)/(B5+B7)

Modi?ed CRC(B5?B2)/(B5+B2)Sullivan et al.(2006)

NDI5;NDI7(B4?B5)/(B4+B5);

(B4?B7)/(B4+B7)

McNairn and Protz(1993) Hyperion CAI0.5(R2.0+R2.2)?R2.1R2.0and R2.2:the re?ectance on the

shoulders at2021nm and2213nm

Daughtry et al.(2006)

ASTER LCA100(2×B6?B5?B8)B5,B6,B7,B8:ASTER shortwave

infrared bands5,6,7,and8Daughtry et al.(2005)

SINDRI(B6?B7)/(B6+B7)Serbin et al.

(2009a) Fig.1.Pictures of agricultural?elds:before tillage(A),after tillage/planting with no or little vegetation(B&C),and crop emergence(D).

178 B.Zheng et al./Remote Sensing of Environment117(2012)177–183

is only active when requested.ASTER's shortwave infrared(SWIR)de-tector failed in April2008(NASA,2011).Thus,subsequently,ASTER im-agery is not capable of CRC mapping.MODIS revisits the same area daily;however,MODIS data have coarse spatial resolution(500m in SWIR bands),so may experience mixed pixel problems for many agri-cultural landscapes.

Landsat5TM and Landsat7ETM+imagery currently provide the best available imagery for mapping CRC,not only because their short-wave infrared(SWIR)Band7(2080–2350nm)is sensitive to crop residue,but also because they provide moderate spatial resolution (30m)and an eight-day revisit rate using both Landsat5and7.

3.Study area and data

3.1.Study site

This study was conducted in Central Indiana(Fig.2),one of the most signi?cant agro-ecoregions within the Eastern Corn Belt Plains of the United States.Locations of?eld data for this study are shown in Fig.2.

Central Indiana is an extensive agricultural region with?at topog-raphy.This landscape is drained by long,shallow,streams occupying sinuous valleys.Agricultural lands often have drainage ditches and channelized streams to promote soil drainage in?at,poorly drained, areas.The principal crops are maize(Zea mays)and soybeans(Glycine max).Most of the soils of this region are Al?sols,Inceptisols,and Mol-lisols(Major Land Resource Area[MLRA]111A).Soil erosion rates in this region are from7.5to4.1tons per acre from1982to2007 (USDA-NRCS,2007).

3.2.Field measurements

CRC was measured using a line transect method(Morrison et al., 1993)from May13to May26in2010.We used a50-foot(15.24m) measuring tape which can be easily divided into100parts with0.5-foot intervals shown as red markings.At each sampling site,the tape was stretched diagonally across the rows(NRCS,1992)and the number of markings intersecting crop residue was counted.Then we measured off diagonal and counted the number of markings intersecting crop res-idue again.Percent cover was calculated by the average of the two counted numbers of the markings.In addition,we used a Garmin eTrex GPS unit(positional accuracy of b15m)to record the location of each measurement,acquired photographs,and made notes for each sampling site.We measured a total of72?elds using the line transect method,among which44?elds were planted with corn and28?elds with soybean in2009with the con?rmation of a cropland data layer (58bd8d2b482fb4daa58d4bea/research/Cropland/SARS1a.htm).We found that17of the28soybean?elds displayed a mixture of corn and soybean residue.

3.3.Remotely sensed data

Five Landsat images(Path21/Row32)acquired on March30 (ETM+7),April15(ETM+7),May9(TM5),May25(TM5),and June10(TM5)in2010were atmospherically corrected to surface re?ectance using the Landsat Ecosystem Disturbance Adaptive Proces-sing System(LEDAPS)(Masek et al.,2006).Images acquired on May9 and25are partially covered by clouds and cloud shadows.The Landsat 7ETM+images are scan line corrector(SLC)-off and have data gaps. Serbin et al.(2009b)compared several Landsat-based tillage indices and found that the Normalized Difference Tillage index(NDTI)was the best for separating crop residue and soil.Thus,we generated NDTI layers for each surface re?ectance image and stacked the images into

a time-series of NDTI image.

4.Methods

Multi-temporal Landsat imagery can capture agricultural changes during the spring planting season.Fig.3shows both the changes of Normalized Difference Vegetation Index(NDVI)and NDTI for our study region in Indiana between March30(day89)and June10 (day161)in2010.NDVI and NDTI are positively correlated with

the Fig.2.Locations of sampling sites in Central Indiana.

179 B.Zheng et al./Remote Sensing of Environment117(2012)177–183

green vegetation cover and CRC respectively.The decrease in NDTI values from day 89to day 129(Fig.3)corresponds to the decrease of CRC due to residue weathering and tillage application on the ?eld,while the rebound of NDTI values after day 129is caused by growing vegetation.Thus,NDTI values are affected by greening vege-tation.Fig.4shows how the NDTI values change through time from March 30(day 89)to June 10(day 161)in 2010for three pixels with different levels of CRC.The abrupt change in NDTI value (the diamond dotted line)from day 105(NDTI =0.10)to day 129(NDTI=0.01)(Fig.4)is due to signi ?cant decreases in the amount of CRC (b 30%)caused by non-conservation tillage,while the change in NDTI value was less abrupt (e.g.,changes from 0.14to 0.09)when con-servation tillage (>30%CRC)was applied to the ?elds.The increased NDTI value after day 129(May 9)is caused by growing vegetation.For this speci ?c example,the use of single images acquired on the days 105or 145would result in dif ?culties differentiating conservation from non-conservation tillage.A single image cannot provide reliable assessment of tillage practices because tillage or planting could happen anytime from April to early June in Central Indiana (Table 2),and in the absence of sequential imagery,analysts cannot determine the correct status of a ?eld.4.1.Minimum NDTI

Due to partial cloud cover in some Landsat images and data gaps in Landsat 7ETM+images,samples affected by clouds,cloud shadows,

and missing data were removed from analysis,resulting in 63clean samples.The minimum NDTI (minNDTI)values,representing the closest status of the surface condition right after planting,were chosen from each spectral pro ?le.We applied simple linear regression (SLR)to deter-mine the relationship between minNDTI and ?eld observed CRC.We ?rst sorted our ?eld observation samples by minNDTI values and divid-ed them into calibration (n=31)and test (n=32)datasets by selecting every other sample to ensure representative subsamples.The regression equation from the calibration dataset was then applied to the test data-set.We divided CRC into three categories:CRC b 30%(non-conservation tillage),30%b CRC b 70%,and CRC>70%.Conservation tillage was split into two categories (30%–70%and >70%)to identify ?elds that were likely managed with no-till (CRC>70%).4.2.Percentage change method

In the next step,we applied a percentage change (PC)method to map CRC.We ?rst selected the NDTI values before planting for each sample pixel (NDTI B ).NDTI B was selected according to the following criteria:1)it must be acquired before the minNDTI occurred;and 2)its value should be larger than 0.08because some ?elds may have ex-perienced several tillage operations at different times before planting.Note that the selection criteria for NDTI B may be different for other regions.

The rationale for the PC method is to detect changes of the same pixel from time I (before tillage)to time П(after tillage/planting).The PC is calculated by

NDTI B ?minNDTI eT=NDTI B ?100%

e1T

The magnitude of change in NDTI is different for different tillage types (Fig.4).This method is unique in its ease of use,ability to min-imize effects of soil variation,and to map tillage practices over broad regions.It requires less ?eld validation effort,and can be applied ret-rospectively to archived imagery,as well as those acquired in the future.5.Results

5.1.Minimum NDTI

We found a linear relationship between CRC and minNDTI with a coef ?cient of determination (R 2)of 0.89and root mean square of error (RMSE)of 10.5%for the calibration data (Fig.5).The R 2between measured and predicted CRC is 0.85and RMSE is 12.6%for the test dataset (Fig.6).The slope is 1.05when the intercept was forced to zero.The SLR results in an R 2of 0.87between CRC and minNDTI and RMSE of 11.5%using all 63samples (Fig.7).

Tables 3and 4show error matrixes for the three residue cover categories using SLR for the test dataset and the complete

dataset

58bd8d2b482fb4daa58d4beaparison of time-series NDTI and NDVI values from the same pixel:non-conservation tillage (left);conservation tillage (right).

02

4681012

141618200

R a i n f a l l (c m )

N D T I

Day of Year

58bd8d2b482fb4daa58d4beaparison of time-series NDTI values with different levels of CRC.The magni-tude of change in NDTI differs depending on the type of tillage practice —i.e.,upon the amount of CRC left on the ground.In this instance,the tillage practice was applied around day 129(May 9),as is highlighted in the graph.Daily precipitation data (NOAA Cooperative Station 126340,Noblesville 3W)are also included.

Table 2

2010Indiana crop progress (Indiana Crop and Weather Report,2010).Week ending Corn (%)Soybean (%)Precip.ξ

(cm)

planted Emerged planted Emerged April 111NA ?NA NA 3.3April 1817NA NA NA 0April 2556512NA 1.3May 2712623NA 3.8May 98152359 1.9May 2388795034 3.4May 3094867052 1.4June 697928169 3.2June 20

100

100

91

85

8.6

*NA:not available.ξ

Precip.:weekly total precipitation for Central Indiana.

180 B.Zheng et al./Remote Sensing of Environment 117(2012)177–183

respectively.The overall accuracy is better than 90%and the Kappa coef ?cient (K )is 85%for both datasets.The K of 85%suggests that the classi ?cation accuracy is 85%better than chance alone.The user's accuracies are 72–100%,while the producer's accuracies are 83–100%for discriminating among three categories.5.2.Percentage change method

The high correlation between PC values and the CRC (R 2=0.80)(Fig.8)demonstrates the potential of this PC method for mapping CRC.Fig.8shows that the correct classi ?cation of classes with more than 70%CRC (green dots),more than 30%and less than 70%CRC (blue dots),and less than 30%CRC (yellow dots).Misclassi ?cation is shown in red dots.According to Fig.8,we determined that the classi-?cation rules for this study area are as follows:pixels that reveal a PC less than 40%are assigned to class CRC>70%;pixels with a PC larger than 40%but smaller than 70%are classi ?ed as 30%b CRC b 70%;and pixels with more than 70%change are assigned to non-conservation tillage (CRC b 30%).

The error matrix using the PC method is shown in Table 58bd8d2b482fb4daa58d4bea-pared to Table 4,the PC method resulted in the same overall accuracy and K .However,the user's accuracy of the class,30%b CRC b 70%,is slightly lower than that of minNDTI.We evaluated the difference be-tween the two classi ?cation accuracies using the minNDTI and PC methods using McNemar's test (Agresti,1996;Foody,2004)and

found no signi ?cant difference between two classi ?cation results (z=0.38b 1.96).6.Discussion

Both minNDTI and PC methods were able to classify CRC into three categories.minNDTI improves both continuous range mapping as well as categorical classi ?cation depending on user needs.Previous studies either classi ?ed CRC into two categories to achieve higher prediction accuracy (Gowda et al.,2001;Thoma et al.,2004),or found low correlations between tillage indices and CRC (Daughtry et al.,2006)using Landsat-based indices.The R 2of 0.11between CRC and NDTI reported by Daughtry et al.(2006)is probably because their Landsat image was acquired on June 12when most crops had emerged and confounded the NDTI signal.Other studies (Daughtry et al.,2005;Serbin et al.,2009b )suggested exclusion of pixels with green vegetation from the analysis,especially for Landsat-based till-age indices.Hyperspectral tillage indices are more effective for map-ping CRC than Landsat-based indices (Daughtry et al.,2005)because their narrow bands are more sensitive to crop residue and less sensitive to presence of green vegetation (Serbin et al.,2009b ).Nevertheless,pixels with green vegetation should be masked out using NDVI or other vegetation indices as suggested by Daughtry et al.(2005).Variation in soil moisture content may have negative effect on mapping CRC (Daughtry &Hunt,2008),but unfortunately,we don't have soil moisture data to examine the effect of soil moisture on our methods.No heavy rainfall happened immediately before our image dates (Fig.4).Therefore,there is no indication here that soil moisture in ?uences NDTI values.

Extracting minNDTI values from multi-temporal pro ?les can re-duce unmapped areas because this method can eliminate effects of green vegetation and avoid consideration of areas that farmers have not tilled yet.Watts et al.(2011)discovered that tillage classi ?cation accuracy was better using all ?ve available Landsat images instead

of

C R C %

minNDTI

Fig.5.Crop residue cover (CRC)as a function of minimum NDTI extracted from the time-series of Landsat images (calibration dataset:n =

31).

P r e d i c t e d C R C %

Measured CRC%

Fig.6.Measured vs.predicted crop residue cover (test dataset:n =

32).

C R C %

minNDTI

Fig.7.Crop residue cover (CRC)as a function of minimum NDTI extracted from the time-series of Landsat images (n =63).

Table 3

Error matrix for three residue cover classes using simple linear regression for test dataset.

Classi ?cation data

Reference data CRC b 30%

30%b CRC b 70%CRC >70%Total User accuracy CRC b 30%

100010100%30%b CRC b 70%260875%CRC >70%01131493%

Total

1271332

Producer's accuracy

83%

86%

100%

Overall accuracy:91%;Kappa coef ?cient:85%.Bold data are the number of pixels correctly assigned to each class.

181

B.Zheng et al./Remote Sensing of Environment 117(2012)177–183

using a single-date image,and demonstrated the importance of tem-poral frequency in tillage mapping.

Our methods are simple and can be easily adopted by others.The physical relationship is well explained by the SLR.Re ?ectance values at band 7(2080–2350nm)of TM/ETM+images decrease as CRC in-creases because crop residues have absorption features near 2100nm (Daughtry,2001).Thus,the NDTI has a positive linear relationship with CRC.Other non-linear methods,such as Arti ?cial Neural Networks (Sudheer et al.,2010),should also take into account temporal changes of agricultural surfaces when mapping tillage practices.Under similar soil moisture conditions,the PC method can mitigate soil color variation that could confound both single and multi-date approaches.The 40%break point of the PC method maximizes the classi ?cation accuracy for our study site.Logically,the break point should be 30%.The extra amount of change is probably due to residue weathering.Thus,one may adjust this value between 30and 40%regionally.However,further investigation is needed to con ?rm the causes.

Currently,image availability is one of the most important factors that constrains our ability to map CRC 58bd8d2b482fb4daa58d4beandsat provides global coverage at 30meter spatial resolution.Multi-temporal Landsat imag-ery is required to map CRC because its tillage indices can be biased by any green vegetation (Serbin et al.,2009b ).Multi-temporal methods for mapping CRC are subject to failure with insuf ?cient temporal cover-age of remotely sensed data.Watts et al.(2011)demonstrated the potential of STARFM-based synthetic dataset for mapping tillage practices,which is a potential solution for the lack of availability of cloud-free Landsat imagery.The planned version 2.0of Web-enabled Landsat Data (WELD)(Roy et al.,2010)could be another source of gap-free Landsat 7ETM+data.Our study area includes

ample Landsat 5TM and 7ETM+scenes to support a temporal anal-ysis for most of the years from 1999to 2010(Table 6).Table 6illus-trates the availability of Landsat data.Years with more than four Landsat scenes have a higher chance of success for mapping CRC ac-curately.Further studies are required to test the transferability of our method to new scenes.

7.Conclusions

Availability of practical and reliable methods for monitoring prac-tice of tillage will reduce uncertainty in ecosystem models and permit identi ?cation of areas at risk for soil erosion and nutrient losses.Re-mote sensing is an ef ?cient and cost-effective way to obtain informa-tion concerning CRC/tillage practices.Mapping tillage practices using single-date images could be problematic unless the area has very nar-row window of planting dates.Watts et al.(2011)showed that incor-porating high temporal datasets can improve mapping accuracy of conservation tillage.Our study supports the ?ndings by Watts et al.(2011)that temporal resolution plays a signi ?cant role in mapping CRC/tillage practices accurately.Multi-temporal analyses (minNDTI and PC methods)are able to classify tillage categories and predict CRC along a continuum more accurately.Time-series of Landsat imag-ery have the potential to map CRC at broad scales and ?ll the tempo-ral data gaps in the observation of tillage practices.The Landsat Data Continuity Mission (LDCM)will provide the opportunity for continu-ously mapping the Earth's continental surface.The Hyperspectral In-frared Imager (HyspIRI)mission will provide another opportunity for mapping crop residue with the global coverage and 60meter spa-tial resolution.However,a 19-day revisit time of HyspIRI may not be short enough to provide two to three cloud-free images during plant-ing season.Future remote sensing platforms should consider im-provement of temporal resolution for crop residue detection.

Table 4

Error matrix for three residue cover classes using simple linear regression for all dataset.

Classi ?cation data

Reference data CRC b 30%

30%b CRC b 70%CRC >70%Total User accuracy CRC b 30%

190019100%30%b CRC b 70%31321872%CRC >70%01252696%

Total

22142763

Producer's accuracy

86%

93%

93%

Overall accuracy:90%;Kappa coef ?cient:85%.Bold data are the number of pixels correctly assigned to each class.

C R C %

Percentage Change %

Fig.8.The correlation between crop residue cover (CRC)and percentage change (PC)of NDTI (n =63).Green dots:correctly classi ?ed as CRC >70%(PC b 40%);blue dots:correctly classi ?ed as 30%b CRC b 70%(40%b PC b 70%);orange dots:correctly classi ?ed as CRC b 30%(PC >70%);red dots:misclassi ?cation.(For interpretation of the refer-ences to color in this ?gure legend,the reader is referred to the web version of this article.)

Table 5

Error matrix for three residue cover classes using the percentage change method.Classi ?cation data

Reference data CRC b 30%

30%b CRC b 70%CRC >70%Total User accuracy CRC b 30%

210021100%30%b CRC b 70%11231675%CRC >70%02242692%

Total

22142763

Producer's accuracy

95%

86%

89%

Overall accuracy:90%;Kappa coef ?cient:85%.Bold data are the number of pixels correctly assigned to each class.

Table 6

Summary of Landsat 5TM and 7ETM+scenes available for Central Indiana.Year Image acquisition date

Total images ξ201030-Mar ?,15-Apr ?,9-May,25-May,10-Jun 520094-Apr,12-Apr ?,22-May,23-Jun 4200730-Mar,15-Apr,1-May,2-Jun

420064-Apr ?,28-Apr,6-May ?,22-May ?,30-May 5200524-Mar,9-Apr,25-Apr,11-May,27-May 5200414-Apr ?,8-May,1-Jun ?,

3200312-Apr ?,28-Apr ?,6-May,22-May,23-Jun 5200225-Apr ?,3-May,20-Jun

3200121-Mar ?,14-Apr,30-Apr,8-May ?,9-Jun ?,17-Jun 6200026-Mar,27-Apr,13-May,29-May,6-Jun ?51999

24-Mar,25-Apr,11-May,27-May

4

*Images from Landsat 7ETM+archive.ξ

Total:total number of images;the average cloud cover of all the images is 7.43%.

182 B.Zheng et al./Remote Sensing of Environment 117(2012)177–183

Acknowledgements

This project was funded by the Graduate Research Development Program(GRDP)at Virginia Tech and Virginia Tech Geography Department's Sidman P.Poole Scholarship.The authors would like to thank the remote sensing team from IUPUI and Kai Wang for assist-ing data collection.

References

Agresti,A.(1996).An introduction to categorical data analysis.New York,N.Y.:Wiley. Baker,N.T.(2011).Tillage practices in the Conterminous United States,1989–2004-datasets aggregated by watershed.United State Geological Survey Data Series,573.(pp.13). Biard,F.,&Baret,F.(1997).Crop residue estimation using multiband re?ectance.Remote Sensing of Environment,59,530–536.

CTIC(2010).National survey of conservation tillage practices.West Lafayette,IN: Conservation Technology Information Center Retrieved February,2010,from.

58bd8d2b482fb4daa58d4bea/pdf/TillageDe?nitions.pdf. Daughtry,C.S.T.(2001).Discriminating crop residues from soil by shortwave infrared re?ectance.Agronomy Journal,93,125–131.

Daughtry,C.S.T.,Doraiswamy,P.C.,Hunt,E.R.,Jr.,Stern,A.J.,McMurtrey,J.E.,III,& Prueger,J.H.(2006).Remote sensing of crop residue cover and soil tillage intensity.

Soil and Tillage Research,91,101–108.

Daughtry,C.S.T.,&Hunt,E.R.,Jr.(2008).Mitigating the effects of soil and residue water contents on remotely sensed estimates of crop residue cover.Remote Sensing of Environment,112(4),1647–1657.

Daughtry,C.S.T.,Hunt,E.R.,Jr.,Doraiswamy,P.C.,&McMurtrey,J.E.,III(2005).Remote sensing the spatial distribution of crop residues.Agronomy Journal,97,864–871. Daughtry,C.S.T.,McMurtrey,J.E.,III,Chappelle,E.W.,Dulaney,W.P.,Irons,J.R.,& Satterwhite,M.B.(1995).Potential for discriminating crop residues from soil by re?ectance and?uorescence.Agronomy Journal,87,165–171.

Foody,G.M.(2004).Thematic map comparison:Evaluating the statistical signi?cance of differences in classi?cation accuracy.Photogrammetric Engineering and Remote Sensing,70,627–633.

Gausman,H.W.,Gerbermann,A.H.,Wiegand,C.L.,Leamer,R.W.,Rodriguez,R.R.,& Noriega,J.R.(1975).Re?ectance differences between crop residues and bare soils.Soil Science Society of America Journal,39,752–755.

Gowda,P.H.,Dalzell,B.J.,Mulla,D.J.,&Kollman,F.(2001).Mapping tillage practices with landstat thematic mapper based logistic regression models.Journal of Soil and Water Conservation,56,91–96.

Gowda,P.H.,Mulla,D.J.,&Dalzell,B.J.(2003).Examining the targeting of conservation tillage practices to steep vs.?at landscapes in the Minnesota River Basin.Journal of Soil and Water Conservation,58,53–57.

Indiana Crop and Weather Report(2010).USDA National Agricultural Statistics Service.

West Lafayette,60.Retrieved September,2011,from.58bd8d2b482fb4daa58d4bea/ Statistics_by_State/Indiana/Publications/Crop_Progress_&_Condition/2010/index10.asp. Jarecki,M.K.,Lal,R.,&James,R.(2005).Crop management effects on soil carbon sequestration on selected farmers'?elds in northeastern Ohio.Soil and Tillage Research,81,265–276.Masek,J.G.,Vermote,E.F.,Saleous,N.E.,Wolfe,R.,Hall,F.G.,Huemmrich,K.F.,et al.

(2006).A Landsat surface re?ectance dataset for North America,1990–2000.IEEE Geoscience and Remote Sensing Letters,3,68–72.

McNairn,H.,&Protz,R.(1993).Mapping corn residue cover on agricultural?elds in Oxford County,Ontario,using Thematic Mapper.Canadian Journal of Remote Sensing, 19,152–159.

Morrison,J.E.,Huang,C.-H.,Lightle,D.T.,&Daughtry,C.S.T.(1993).Residue measure-ment techniques.Journal of Soil and Water Conservation,48,478–483.

NASA(2011).ASTER user advisory.Retrieved January,2011,from.asterweb.jpl.

58bd8d2b482fb4daa58d4bea/latest.asp.

NRCS(1992).Farming with crop residue brochure.Retrieved March15,2010,from.

58bd8d2b482fb4daa58d4bea/news/publications/brochures/farmcropres/

FarmCropRes-broc.

Roy,D.P.,Ju,J.,Kline,K.,Scaramuzza,P.L.,Kovalskyy,V.,Hansen,M.,et al.(2010).

Web-enabled Landsat Data(WELD):Landsat ETM+composited mosaics of the conterminous United States.Remote Sensing of Environment,114,35–49. Saseendran,S.A.,Ma,L.,Malone,R.,Heilman,P.,Ahuja,L.R.,Kanwar,R.S.,et al.(2007).

Simulating management effects on crop production,tile drainage,and water qual-ity using RZWQM-DSSAT.Geoderma,140,297–309.

Serbin,G.,Daughtry,C.S.T.,Hunt,E.R.,Jr.,Brown,D.J.,&McCarty,G.W.(2009).Effect of soil spectral properties on remote sensing of crop residue cover.Soil Science Society of America Journal,73,1545–1558.

Serbin,G.,Hunt,E.R.,Daughtry,C.S.T.,McCarty,G.,&Doraiswamy,P.(2009).An improved ASTER index for remote sensing of crop residue.Remote Sensing,1, 971–991.

Sudheer,K.P.,Gowda,P.,Chaubey,I.,&Howell,T.(2010).Arti?cial neural network approach for mapping contrasting tillage practices.Remote Sensing,2,579–590.

Sullivan,D.G.,Shaw,J.N.,Price,A.,&van Santen,E.(2007).Spectral re?ectance properties of winter cover crops in the Southeastern Coastal Plain.Agronomy Journal,99, 1587–1596.

Sullivan,D.G.,Strickland,T.C.,&Masters,M.H.(2008).Satellite mapping of conserva-tion tillage adoption in the Little River experimental watershed,Georgia.Journal of Soil and Water Conservation,63,112–119.

Sullivan,D.G.,Truman,C.C.,Schomberg,H.H.,Endale,D.M.,&Strickland,T.C.(2006).

Evaluating techniques for determining tillage regime in the Southeastern Coastal Plain and Piedmont.Agronomy Journal,98,1236–1246.

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

USDA-NRCS(2001).A resources conservation act report.Interim appraisal and analysis of conservation alternatives.

USDA-NRCS(2007).2007National resources inventory.Soil erosion on cropland Retrieved April10,2011,from.58bd8d2b482fb4daa58d4bea/technical/NRI/2007/nri07erosion.

html.

van Deventer,A.P.,Ward,A.D.,Gowda,P.H.,&Lyon,J.G.(1997).Using Thematic Mapper data to identify contrasting soil plains and tillage practices.Photogrammetric Engineering and Remote Sensing,63,87–93.

Watts,J.D.,Lawrence,R.L.,Miller,P.R.,&Montagne,C.(2009).Monitoring of cropland practices for carbon sequestration purposes in north central Montana by Landsat remote sensing.Remote Sensing of Environment,113,1843–1852.

Watts,J.D.,Powell,S.L.,Lawrence,R.L.,&Hilker,T.(2011).Improved classi?cation of conservation tillage adoption using high temporal and synthetic satellite imagery.

Remote Sensing of Environment,115,66–75.

183

B.Zheng et al./Remote Sensing of Environment117(2012)177–183

本文来源：https://www.bwwdw.com/article/0qme.html