The normalized difference vegetation index (NDVI) Greenseeker(TM) handheld sensor: toward the integrated evaluation of crop management. Part A - Concepts and case studies

7/31/2019 The normalized difference vegetation index (NDVI) Greenseeker(TM) handheld sensor: toward the integrated eval

1/16

The normalized difference vegetation index

(NDVI) GreenSeekerTM handheld sensor:

Toward the integrated evaluation of crop managementPart A: Concepts and case studies

Bram GovaertsInternational Maize and Wheat Improvement Center (CIMMYT)

[email protected]

Nele VerhulstInternational Maize and Wheat Improvement Center (CIMMYT)

Katholieke Universiteit Leuven (K.U.Leuven)

[email protected]

Verhulst, N., Govaerts, B. 2010. The normalized difference vegetation index(NDVI) GreenSeekerTM handheld sensor: Toward the integrated evaluation ofcrop management. Part A: Concepts and case studies. Mexico, D.F.; CIMMYT.


2/16

Table of contents

Introduction ................................................................................................................................1

A. Plant reflectance and normalized difference vegetation index (NDVI) .................1

B. NDVI and remote sensing: A small review .................................................................2

C. A case study in the Mexican highlands .......................................................................2Using the NDVI handheld sensor to monitor crop growth and development .................5

Spatial variability in crop performance as an indicator of sustainability .........................7

Using spatial variability in crop performance (NDVI) to evaluate

soil processes determining the crop system sustainability ...........................................9

References .................................................................................................................................13

Acknowledgements

Verhulst received a scholarship from the Research Foundation Flanders.The presented research was funded by CIMMYT and its strategic partners.Introduction to the NDVI GreenseekerTM sensor and scientific backstoppingand exchange was provided by Bill Raun at Oklahoma State Universityand the finalization and reproduction of this training document was madepossible thanks to the 2009 USAID Linkage Funds.


3/16

1

Healthy plant

Stressed plant

50

40

30

20

10

0

400 700 1,000Visible Near Infrared

Wavelenght (nm)

Reec

tance(%)

Introduction

A. Plant reflectance and normalized difference

vegetation index (NDVI)

Reflectance is the ratio of energy that is reflectedfrom an object to the energy incident on the object.

Spectral reflectance of a crop differs considerablyin the near infrared region ( = 700-1300 nm) andin the visible red range ( = 550-700 nm) of theelectromagnetic spectrum (Kumar and Silva, 1973;Figure 1). Plants generally have low reflectance inthe blue and red portion of the spectrum becauseof chlorophyll absorption, with a slightly higherreflectance in the green, so plants appear green toour eyes. Near infrared radiant energy is stronglyreflected from the plant surface and the amount ofthis reflectance is determined by the properties ofthe leaf tissues: their cellular structure and the air-

cell wall-protoplasm-chloroplast interfaces (Kumarand Silva, 1973). These anatomical characteristicsare affected by environmental factors such assoil moisture, nutrient status, soil salinity, andleaf stage (Ma et al., 2001). The contrast betweenvegetation and soil is at a maximum in the redand near infrared region. Therefore, spectralreflectance data can be used to compute a varietyof vegetative indices that are well-correlated withagronomic and biophysical plant parameters relatedto photosynthetic activity and plant productivity(Ma et al., 2001; Adamsen et al., 1999). The NDVI

is successful in predicting photosynthetic activity,because this vegetation index includes both nearinfrared and red light. Plant photosyntheticactivity is determined by chlorophyll content and

activity. The relationship between leaf N and leafchlorophyll has been demonstrated for maize(Piekielek and Fox, 1992; Chapman and Barreto,1997) and wheat (Evans, 1989).

The NDVI is calculated from reflectancemeasurements in the red and near infrared (NIR)portion of the spectrum: R

NIR-R

RedNDVI=

RNIR

-RRed

where RNIR

is the reflectance of NIR radiation andR

Redis the reflectance of visible red radiation.

The NDVI has been correlated to many variablessuch as crop nutrient deficiency, final yield insmall grains, and long-term water stress. However,rather than exclusively reflecting the effect of

one parameter, NDVI has to be considered as ameasurement of amalgamated plant growth thatreflects various plant growth factors. The physicalcharacteristics detected by the index are likelyrelated to some measure of canopy density (i.e. leafarea or percent cover) or total biomass. Therefore,the underlying factor for variability in a typicalvegetation index cannot be blindly linked to amanagement input without some knowledge ofthe primary factor that limits growth. For example,in a field where N is the limiting factor to growth,the NDVI may show a strong correlation with the

N availability in the soil; however, in another field,where water is the limiting factor, the NDVI maybe just as strongly correlated with plant-availablesoil moisture.

There are different vegetation indices; however,those that rely on NIR and red reflectance as theirprincipal inputs will typically yield the sameinformation as the NDVI. One of the reasons forthe popularity of the NDVI is that many sensors(from handheld to satellite) provide measurementsin the NIR and red portion of the spectrum. NIRis also used in color infrared photographs. Most, ifnot all, of the new commercial satellites will havered and NIR bands, so the availability of thesedata will increase.

Further readingAraus, J.L., J. Casadesus, J. Bort. 2001. Recent tools for

the screening of physiological traits determiningyield. In: M.P. Reynolds, J.I. Ortiz-Monasterio, A.McNab (eds.), Application of physiology in wheatbreeding. Mexico, D.F.: CIMMYT. Pp. 59-77.Figure 1. Typical reflectance spectrum of a healthy

and a stressed plant.


4/16

2

B. NDVI and remote sensing: A small review

Satellite-based NDVI are influenced by a numberof non-vegetation factors: atmospheric conditions(e.g. clouds and atmospheric path-specific variables,aerosols, water vapor), satellite geometry andcalibration (view and solar angles), as well as

soil backgrounds and crop canopy (Holben 1986;Soufflet et al. 1991; Justice et al. 1991). The angleof incidence of solar radiation also has a strongeffect on vegetation indexes (Pinter 1993). However,these complications can be avoided by using theGreenSeeker handheld optical sensor unit tomeasure NDVI. Designed at Oklahoma StateUniversity, and commercialized by Ntech Industries,the GreenSeeker cancels out the disturbing effectsof atmospheric interference and satellite geometrysince it is held closely above the crops. Moreover,the handheld sensor contains its own light source,allowing measurements to be taken day or nightwithout interference of sunlight and sun stand.Lack of effect of climate as well as sun angle wasconfirmed by an independent study as reported onthe Oklahoma State University website (http://nue.okstate.edu). This is a great advantage comparedto the satellite-based measurements. The highresolution obtained with this handheld sensormakes proper measurement possible at the plot levelin contrast with the low resolution typical for air orspace remote sensing material. The handheld sensoris non-destructive and the sensor samples at a veryhigh rate (approximately 1,000 measurements per

second) and can easily and time-efficiently measurea whole plot representative area. There is, however,still important scope for research on the comparisonof the NDVI handheld sensor with satellite imagery,especially when scaling out of results and modelsbecomes important.

C. A case study in the Mexican highlands

Description of the experimental area and designThe case studies that are used as examples herewere all done in a long-term tillage trial located atEl Batn in the semiarid, subtropical highlands ofcentral Mexico (2,240 m a.s.l.; 19.318N, 98.508W). Themean annual temperature was 14C (1990-2001) andthe average annual rainfall was 600 mm per year,with approximately 520 mm falling between Mayand October. Short, intense rain showers followedby dry spells typify the summer rainy season andthe total yearly potential evapotranspiration of1,900 mm exceeds rainfall throughout the year.The El Batn experiment station has an average

growing period of 152 days. The soil is classifiedas a fine, mixed, thermic Cumulic Haplustoll (SoilSurvey Staff, 2003) or as a Cumulic Phaeozem(International Union of Soil Sciences (IUSS) workinggroup World Reference Base (WRB), 2006). El Batnsclimate is representative of many highland areas inthe regions of West Asia and North Africa, as wellas the Southern Cone and Andean highlands ofSouth America, the central highlands of Ethiopia,the Mediterranean coastal plains of Turkey, and thehighlands of central Mexico (van Ginkel et al., 2002).

The experiment was started in 1991 as describedin Fischer et al. (2002). Individual plots are 7.5 mby 22 m. Standard practices include the use ofrecommended crop cultivars, with maize planted at60,000 plants per hectare in 75 cm rows and wheatplanted in 20 cm rows at 100 kg of seed per hectare.Both crops are fertilized using urea at 120 kg N per

hectare. Weed control is done using appropriate,available herbicides as needed and no disease orinsect pest controls are utilized, except for seedtreatments applied by commercial seed sources.Planting of both maize and wheat depends on theonset of summer rains but is usually done betweenJune 1 and 15.

The experimental design consists of a randomizedcomplete block with two replications. There are32 treatments. The core set of 16 managementpractices was based on variation of (1) crop rotation

(monocropping vs. a maize/wheat rotation); (2)tillage (conventional vs. zero tillage); (3) residuemanagement (retention vs. removal). A second setof treatments was established in 1996 and includestreatments with partial residue retention andplanting on permanent raised beds. In the casestudies used, only the core set of 16 practices will beconsidered that were all installed in 1990 and keptsince. Table 1 summarizes the considered practices.

Table 1. Treatments at the CIMMYT long-term tillage

trial at El Batn, Mexico.

Tillage Zero tillage Conventional tillage

Residuemanagement Kept Removed Kept Removed

Rotations M M M M M M M M

M W M W M W M W

W M W M W M W M

W W W W W W W W

W=wheat, M=maize.Rotation: MM=continuous maize, WW=continuous wheat,

WM or MW=yearly rotation of maize and wheat.


5/16

3

assessment and proceed from a theoretical definitionto a measurement of soil quality (Larson and Pierce,1991; Gregorich et al., 1994). Different authors haveproposed several minimum data sets. We proposethat instead of working with predefined lists ofindicators, that indicators be selected based on thesite-specific agro-ecological conditions by comparingoptimal conditions for the specific land use with thereal conditions. This comparison would reveal thelimiting factors of the system. All parameters relatedto these limiting factors would be measured in a firstoverall evaluation. Based on the obtained results, wecould thereafter refine the list of relevant parametersand come up with a minimum data set for futureassessments (Govaerts et al., 2006a).

Soil quality in Mexican highlands case studyIn order to apply the concept of soil quality, one musthave a defined set of indicator parameters for what

to measure. As discussed, several minimum lists ofindicators can be found in literature. The approachused for this case study, however, was different. Wedid not utilize the so-called generally predefinedminimum data sets (Larson and Pierce, 1994), butinstead utilized an agro-ecological site-specificselection of indicators as described above, whichled to a two-step-approach. The first step was theset up of a limiting factor parameter list, based on acomparison of the optimal conditions for the land-use and the actual agro-ecological characteristics.Indicators related to the limiting factors will possibly

be relevant for the evaluation of the system and formthe limiting factor parameters. The second step wasthe selection of the most explicative indicators fromthe set of measured indicators to form the eventualstrict minimum data set.

Table 2 shows the results of the comparison ofthe optimal conditions with the actual situationat the CIMMYT station at El Batn in the centralMexican highlands. Table 3 lists the limiting factorparameters, which were evaluated by measuring thefollowing indicators:

- Physical: small ring infiltration, direct surfaceinfiltration, aggregation by wet and drysieving, penetration resistance, conepenetration, probe depth, bulk density,permanent wilting point, field capacity

- Chemical: CEC, total N, NO3

-, and NH4

+, totalorganic C, P, macronutrients (Ca, Mg,K, Na), micronutrients (Fe, Mn, Zn, Cu),pH, EC

- Biological: Microbial biomass C and N

The soil quality conceptAs a way of further introducing the case study,we will present some soil quality results fromthe long-term experiment. When evaluating anagricultural management system for sustainability,the central question is: which production systemwill not exhaust the resource base, will optimizesoil conditions and will reduce food productionvulnerability, while at the same time maintainingor enhancing productivity? Soil quality can be seenas a conceptual translation of the sustainabilityconcept towards soil. Karlen et al. (1997) definedsoil quality for the Soil Science Society of Americaas the capacity of a specific kind of soil to function,within naturally managed ecosystem boundaries, tosustain plant and animal productivity, maintain orenhance water and air quality, and support humanhealth and habitation. A simpler operationaldefinition is given by Gregorich et al. (1994) as the

degree of fitness of a soil for a specific use. Thisimplies that soil quality depends on the role forwhich the soil is destined (Singer and Ewing, 2000).Within the framework of agricultural production,high soil quality equates to the soils ability tomaintain a high productivity without significantsoil or environmental degradation. Evaluation ofsoil quality is based on physical, chemical, andbiological characteristics of the soil.

A comparative soil quality evaluation is one inwhich the systems performance is determined

in relation to alternatives. When the biotic andabiotic soil system attributes of alternative systemsare compared, a decision about the relativesustainability of each system is made based onthe difference in magnitude of the measuredparameters (Larson and Pierce, 1994). A comparativeassessment is useful for determining differencesin soil attributes among management practicesthat have been in place for some period of time(Wienhold et al., 2004). A dynamic assessmentapproach differs in that the dynamics of the systemform a meter for its sustainability (Larson and

Pierce, 1994). A dynamic assessment is necessary fordetermining the direction and magnitude of changea management practice is having (Wienhold et al.,2004), especially when compared to the common,existing farmer practices. This assessment normallyinvolves an adequate time frame.

A minimum set of soil characteristics thatrepresents soil quality must be selected andquantified, to be able to apply any evaluation


6/16

4

After statistical analysis, the refined soil qualityminimum data set included the following physicalindicators: time-to-pond, aggregate stability,permanent wilting point, and topsoil resistance.Important chemical indicators were concentrationsof C, N, K, and Zn in the 05 cm topsoil and C, Nconcentrations in 520 cm.

Multivariate analysis grouped the treatments intoclusters: (1) zero tillage with retention of residue;(2) zero tillage with residue removal; and (3)conventional tillage. Zero tillage combined with cropresidue retention improved chemical and physicalconditions of the soil. In contrast, zero tillage withremoval of residues led to high accumulation ofMn in the topsoil, low aggregate stability, highpenetration resistance and surface slaking resultingin low time-to-pond values and high runoff.Finally, soil quality under conventional tillage was

intermediate (irrespective of residue mangement),especially reflected in the physical status of the soil.

Table 2. Comparison of optimal conditions for crop growth versus El Batn agro-ecological conditions

(adapted from Govaerts et al., 2006a).

Optimal wheat Conditions at Optimal maize Conditions at Possible

Parameter conditions El Batn conditions El Batn limitation

Climate

Soil temp. 15-22 C 16-18 C* 15-22 C No

Optimum day temp. 2025 Cs 20-25 C 25-30 C* 20-25 C No

Mean night temp. >13 C* No

Mean day temp. 5.0s 5.9 5.07.0* 5.9 No

Al-content Lows 0 Low* 0 No

Nutrient condition Highs Tend to decline High* Tend to decline Yes

Micronutrients Commonly deficient Tend to decline Commonly deficient Tend to decline Yes

in Cu, B, Mn, Zn** in Fe, Zn**

Pathogen Pathogenfree Yellow rust, leaf rust Pathogen-free Nematode sensible Yes

and Septoria tritici

s Tanner and Raemaeker, 2002; * Ristanovic, 2002; ** Sayre, K. D. personal communication

Table 3. Limiting factor parameterset for the El Batn area (Adapted

from Govaerts et al., 2006a).

Parameters

Compaction

Infiltration

Moisture content

Aggregate stability

Bulk density

Organic carbon

Nutrient status

Biological activity

Soil borne diseases

More details on the case study can be found inGovaerts, B., K.D. Sayre, J. Deckers. 2006a. A minimum

data set for soil quality assessment of wheat andmaize cropping in the highlands of Mexico. Soil&TillageResearch 87: 163-174.


7/16

5

Further readingBarrios, E., R.J Delve, M. Bekunda, J. Mowo, J. Agunda,

J. Ramisch, M.T. Trejo, R.J. Thomas. 2006. Indicatorsof soil quality: A south-south development of amethodological guide for linking local and technicalknowledge. Geoderma 135: 248-259

Doran, J.W., T.B. Parkin. 1994. Defining and assessingsoil quality. In: J.W. Doran, D.C. Coleman, D.F.Bezdicek, B.A. Stewart (eds.), Defining soil qualityfor a sustainable environment. Madison: AmericanSociety of Agronomy (ASA) and Soil Science Societyof America (SSSA). Pp. 3-21.

Karlen, D.L., D.E. Stott. 1994. A framework for evaluatingphysical and chemical indicators of soil quality.In: J.W. Doran, D.C. Coleman, D.F. Bezdicek, B.A.Stewart, (Eds.), Defining soil quality for a sustainableenvironment. Madison: ASA and SSSA. Pp. 53-72.

Larson, W.E., F.J. Pierce. 1994. The dynamics of soilquality as a measurement of sustainable management.In: J.W. Doran, D.C. Coleman, D.F. Bezdicek, B.A.Stewart (Eds.), Defining soil quality for a sustainableenvironment. Madison: ASA and SSSA. Pp. 37-51.

Using the NDVI handheld sensor

to monitor crop growth and

development

Crop performance, growth, and development arethe integrated evaluators that show the efficiency ofthe chosen agricultural management system withinthe boundaries of the agro-ecological environment.Any crop cultivar (that has been selected for thegiven agro-ecological area), will act as an integrated

evaluator of all environmental factors thus showinghow management influences and determinesresource-use efficiency. Yields can be measuredas an end-of-season static result of seasonal cropperformance, but these results do not reflect thefluctuations of the crops performance throughoutthe season. End-of-season yield results do notpermit the evaluation of within-season managementinteractions with the production environment anddo not allow for full understanding of the appliedmanagement practice. In order to understandand evaluate cropping systems, and to fine-tune

resource management, crop performance over timeis a crucial factor. The effect of management factors,such as tillage systems, crop residue management,and crop rotation on crop growth and developmentduring the crop cycle has not been studiedintensively. Until now, most of the knowledge onplant growth and development has been developedfor conventional management practices, includingheavy tillage and common crop residue removal.

The NDVI handheld sensor can be used to followcrop growth and development throughout theseason, and thus increases our understanding ofthe different management practices.

Case study from the Mexican highlands

Materials and methodsNDVI measurements were taken with theGreenSeeker Handheld Optical Sensor Unit(NTech Industries, Inc., USA) in the central rowsof all plots of the 16 core practices that werestudied 3 times a week throughout the 2004 and2006 growing seasons. The average NDVI valueswere plotted against time for all treatments. As anexample, the NDVI-based growth and developmentcurves for maize in the 2006 growing season areshown (Figure 2). The NDVI curves were analyzedwith PROC MIXED (SAS institute, 1994) using theREPEATED statement for the analysis of repeatedmeasurements. The NDVI curves were dividedin three periods for the PROC MIXED analysisand the analysis was done separately for eachperiod. The 3 considered periods were: Period Iwith increasing NDVI values (days 16-66), Period IIwith relatively stable NDVI values (days 69-94) andPeriod III with decreasing NDVI values (days 100-136) (Verhulst et al., 2010).

ResultsThe zero tillage practices with residue retentionhad lower NDVI values in Period I compared to

the conventional tillage treatments with the samerotation (P


8/16

6

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

16 23 30 37 44 51 58 65 72 79 86 93 100 107 114 121 128 135Days after planting

NDVI

MM, ZT, K MM, ZT, R

MM, CT, K MM, CT, RWM, ZT, K WM, ZT, R

WM, CT, K WM, CT, R

in crop development under different tillage practicesare scarce, but some reports were found thatcoincide with our findings. Riley (1998) reportedthat development of spring cereals was delayedwith reduced tillage, but this was compensatedfor later in the season. Raimbault and Vyn (1991)and Vyn and Raimbault (1993) reported that zerotillage resulted in slower plant growth compared toconventional tillage systems. However, McMasteret al. (2002) reported faster, more uniform andgreater seedling emergence under zero tillage thanin conventional tillage in four out of six years inthe Central Great Plains, due to more favorable soilwater levels in the seeding zone under zero tillage.

It is important to note that the slower take off ingrowth with zero tillage compared to conventionaltillage is compensated for later in the season.Moreover, when looking at final yield (reported

in Govaerts et al., 2005), treatments with higheryields generally achieved their maximum NDVIlater in the growing period. This indicates thattreatments with an initial slower growth mayhave an advantage. It seems that zero tillagewith residue retention induces a more timely andefficient use of available crop growth resources. Itcould be hypothesized that the changes in C andN cycling between zero tillage and conventional

tillage both with residue retention result in abetter synchronization between demand and Nmineralization in zero tillage with residue retention,where N is released more slowly as compared tothe flush of N released in conventional tillage atthe beginning of the season when tillage is applied.However, more research is needed to confirm thishypothesis.

Rotation seemed to have an influence on early cropdevelopment with lower NDVI values for crops sownafter wheat than for crops sown after maize. Noreports were found in literature on wheat slowingdown the early crop development of the followingcrop. Differences between rotations disappearedlater in the growing season and growing wheat asthe previous crop had no adverse effect on final yieldcompared to growing maize as the previous crop.

The two management practices that combined zerotillage with residue removal (continuous maize andwheat-maize rotation) were found to be overall lowperforming when analyzing the NDVI-based cropgrowth curves. This corresponds with the overallnegative impact on soil quality and soil health ofthese practices (Govaerts et al., 2006a,b; 2007a, b; 2008;2009) and their low yields (Govaerts et al., 2005).

Crop sequence: MM monoculture of maize, WM yearly rotation of maize and wheat. Tillage system: CT conventional tillage, ZTzero tillage. Residue management: R all residues removed from the field, K all residues kept on the field.

Figure 2. NDVI-based crop growth and development curves (NDVI vs. days after planting) for maize in the

2006 crop cycle in the long-term sustainability trial at El Batn, Mexico (adapted from Verhulst et al., 2010).


9/16

7

It can be concluded that tillage, rotation, andresidue management practices influence cropdevelopment. However, more research is needed tofully understand the underlying mechanisms. Asmost practices and knowledge are currently basedon conventionally tilled cropping, it is importantto monitor and understand crop growth underdifferent management systems to select the rightvarieties and adjust timing and practice of inputsupply (fertilizer, irrigation, etc.) in a holistic way foreach system.

More details on the case study can be found inVerhulst, N., B. Govaerts, K.D. Sayre, P. De Corte, J.

Crossa, J. Deckers, L. Dendooven. 2010. The effectsof tillage, crop rotation, and residue managementon maize and wheat growth and developmentevaluated with an optical sensor. FieldCropsResearch(Submitted).

Further readingRaimbault, B.A., T.J. Vyn. 1991. Crop-Rotation and Tillage

Effects on Corn Growth and Soil Structural Stability.AgronomyJournal 83: 979-985.

Riley, H.C.F. 1998. Soil mineral-N and N-fertilizerrequirements of spring cereals in two long-term tillagetrials on loam soil in southeast Norway. Soil&TillageResearch 48: 265-274.

Vyn, T.J., B.A. Raimbault. 1993. Long-Term Effect of 5Tillage Systems on Corn Response and Soil-Structure.AgronomyJournal 85: 1074-1079.

Spatial variability in crop performance

as an indicator of sustainability

The spatial structure of ecosystems often reflectshow these systems function (Herrick et al., 2002).The spatial ecosystem structure reflects the spatialdistribution of the key production-related processes.A change in spatial variability in plant performanceon any scale indicates that the distribution oflimiting resources has changed or that anotherresource has become limiting. This may reflect achange in the processes that both control and areaffected by the availability of resources on that scale.When all plant-growth elements are abundantlyavailable, a uniform pattern of plant growth willbe seen. However, when one or more critical plantelements are limiting, plant-to-plant competitioneffects will increase plant-to-plant performancevariability, increasing the coefficient of variation(CV) compared with a system where no elements arelimiting.

As a general principle, we propose that competitionfor resources results in greater within-plot plant-to-plant variability. Although scarce, there are somereports that support this principle. In a study ofmaize growth evolution, the onset of intra-specificcompetition was inferred from an increase in theCV of plant biomass. Edmeades and Daynard(1979) reported that at a low density (5 plants permeter), the CV of plant biomass had a low andalmost constant value (ca. 10%) during the wholegrowing season, indicative of a similar growth ofeach individual plant within the stand. Contrarily,at a high plant population of 20 plants per meter,the CV increased to 40% during the same period.This statistical parameter can therefore reveal theexistence of plants with different competitive abilitieswithin the same stand density (Edmeades andDaynard, 1979). The onset of this hierarchical growthpattern among plants within a stand would be

related to the intensity of intra-specific competition,i.e. plant population density (Maddonni and Otegui,2004). Comparing different improved maize hybrids,Tollenaar and Wu (1999) concluded that cropresource-use efficiency is inversely related to plant-to-plant variability. Martin et al. (2005) suggestedcompetition for soil moisture as a source of increasedplant-to-plant variability, especially in dryland fields.

Within-plot spatial variability can be the result ofinherent variation in plot conditions. However,agronomical practices also influence spatial within-

plot plant variability. Increased within-plot plantspatial variability throughout the season cantherefore be considered a reaction to inefficientuse of critical plant growth resources provoked byan unsustainable management of these resources.Ginting et al. (2003) reported that differences insoybean yield between high and low elevationwere larger for a conventional tillage systemcompared to a reduced tillage system. Kravchenkoet al. (2005) compared a zero-input with a low andconventional input treatment, and found that theoverall variability (expressed by CV) was the highest

in the zero-input treatment and that crop yields ofthe same treatment were more sensitive to small-scale variations in nutrient and water availabilityconditions of the field.

Spatial variability reduces resource-use efficiency:the potentials in climate conditions and germplasmare expressed only in certain parts of the plot,while in others, crop yields lag behind. Increasedplant spatial variability throughout the season can


10/16

8

0

5

10

15

20

25

30

35

40

16 23 30 37 44 51 58 65 72 79 86 93 100 107 114 121 128 135

Days after planting

CV(%)

MM, ZT, K MM, ZT, R

MM, CT, K MM, CT, R

WM, ZT, K WM, ZT, R

WM, CT, K WM, CT, R

therefore be considered a reaction to inefficientuse of critical plant growth resources induced byan unsustainable management of these resources.As such, it can serve as a sound indicator of cropmismanagement and can help to correct this. TheGreenSeekerTM NDVI sensor can be used as a toolto follow the spatial variability in crop performancethroughout the season.


Materials and methodsNDVI measurements were taken in the same wayas the measurements of the crop development studyin the 2004, 2005, and 2006 growing seasons. TheCV is defined as the standard deviation expressedas a percentage of the mean result (Steel et al. 1997).The CV was calculated for each NDVI measuringsequence per plot that consisted of approximately

200 individual measuring points throughout the plot.As an example, the CVs corresponding to the maizegrowth curves (CV vs. days after sowing) from the2006 growing season are shown (Figure 3). The CVsof the NDVI calculated from the measurement tracksmeasured several times during the growing seasonwere analyzed as dependent variables with PROCMIXED using the REPEATED statement for theanalysis of repeated measurements in time.

ResultsThe CV curves showed a general trend oppositeto the one observed in the NDVI curves (Figure2). There was high spatial variability at thebeginning of the season for all treatments. Afterthis initial stage, the canopy began to close andthe CV declined (until approximately 65 daysafter planting). Soil coverage by the canopy wasthen near maximum, with a uniform leaf color,and a CV of


11/16

9

DiscussionMeasurements of CV throughout the crop seasonreflected the growth and senescence curve ofmaize (Figure 3). Once the canopy began to close,leaves from larger plants covered the leaves andwhorl of smaller plants, extending further into thelinear row. As these leaves began to fill the row,intersecting with, and in some cases covering upleaves from smaller plants, soil coverage increasedthe amount of green vegetation. Comparableresults were obtained by Raun et al. (2005).

There were significantly higher CVs throughoutthe season in the sustainability trial for bothmaize grown in monoculture as well as in rotationwith wheat when planted with zero tillage andresidue removal (P


12/16

10

the field held more water so that yield productionreached a level at which the potassium content wasinadequate and limited production. Machado etal. (2002) reported a positive effect of soil NO

3-N

on sorghum grain yield in a year when water wasabundant, but a negative effect in a year whenwater was limited.

Linking spatial variability in crop performanceto differences in soil attributes could identify thelimiting factors driving the system. Patterns of cropperformance will follow the spatial variability ofthe underlying limiting soil attributes. The sensordetects cold and hot zones of plant performance,which can be correlated to field spots of differingsoil quality. This allows a detailed investigation ofunderlying soil processes and how they might beaffected by different management practices.


Materials and methodsNDVI measurements were taken as a measureof plant performance with the GreenSeekerHandheld Optical Sensor Unit (NTech Industries,Inc., USA) 84 days after planting in the 2006 cropcycle (the beginning of tasseling for maize and thebeginning of grain filling for wheat). For maize,NDVI was measured in all rows, except borderrows, giving a total of eight rows. For wheat, wemeasured 6 strips that were 0.60 m wide, with thefirst and last one at 1.0 m from the border and the

remaining ones at equidistance. Soil attributeswere only measured spatially in plots with croprotations of maize and wheat (8 treatments). Soilattributes were determined in 8 points withineach plot, lying on a grid of 5.5 by 2.5 m, leaving aborder of 3 m at the south-east side of the plot and aborder of 2.5 m at all other sides. The following soilattributes were measured: volumetric soil moisturecontent; direct surface infiltration (time-to-pond);aggregate distribution and stability by dry and wetsieving; and total N, organic matter, pH, electrolyticconductivity, content of Ca, Mg, Na, K, and

inorganic N. To examine the within-plot patterns ofcrop performance under the different treatments,maps of NDVI were produced using ArcMapsoftware, version 9.2 (Environmental SystemsResearch Institute 2006). To link the patterns inplant performance visually to the variability insoil attributes, overlays of NDVI and soil attributeswere made with ArcMap 9.2. As an example, theoverlay figures are shown for maize for some key

soil attributes (Figure 4). In the overlay figure onlyone plot was chosen to represent both conventionaltillage with residue removal and incorporation,as the effect of residue was minimal underconventional tillage.

ResultsThere was a clear pattern in crop performancefor maize with zero tillage and residue removal(Figure 4), which was not observed in thesurrounding plots. Low values were found atthe south-east side of the field (light colors) andhigh values at the north-west side (darker colors).In plots under zero tillage with residue removal,soil moisture content, time-to-pond, and soilaggregate distribution, expressed as mean weightdiameter (MWD) obtained through dry sievingand organic matter content at 0-5 cm, varied fromlow values at the south-east side of the field to

high values at the north-west side, reflecting thepattern in crop performance (Figure 4). Highervalues were observed for the same soil attributesunder zero tillage with residue retention thanunder conventional tillage (Figure 4). Plots underzero tillage with residue retention or conventionaltillage did not have a pattern of soil attributes norin crop performance. Maps of inorganic N contentshowed a clear pattern under zero tillage withresidue removal in the 0-5 cm layer (Figure 4).Inorganic N ranged from high values at the south-east side of the field and toward low values at the

north-west side, the opposite of what was observedfor crop performance.

DiscussionSoils under zero tillage with residue removal didnot have a mulch layer that slowed down run-offand absorbed water. The poor structure resultedin fast surface sealing, low infiltration rates, highrun-off, and soil erosion (Govaerts et al., 2006a).Small variations in topography (field slope


13/16

11

Figure 4. Overlays of NDVI and soilattributes (volumetric soil moisture

content (%), time-to-pond, mean

weight diameter obtained through

dry sieving (mm), soil organic

matter (%), inorganic nitrogen

content (mg per kg) in the 0-5 cm

and the 5-20 cm layer) for maize

plots in 2006 at CIMMYTs long-term

sustainability trial, El Batn, Mexico

(adapted from Verhulst et al., 2009).

Zero tillage with residue removal

Zero tillage with residue retention

Conventional tillage

Time -to-pond (s) MWD (mm)Soil moisturecontent (%)

Organic matter0-5 cm (%)

Inorganic nitrogen (mg/kg)0-5 cm 5-20 cm

1.001-1.2001.201-1.4001.401-1.6001.601-1.8001.801-2.0002.001-2.200

2.201-2.400

2.401-2.600

2.601-2.800

2.801-3.000

3.001-3.200

3.201-3.400

15.01-16.0016.01-18.0018.01-20.0020.01-22.0022.01-24.0024.01-26.00

26.01-28.00

28.01-30.00

30.01-32.00

32.01-34.00

34.01-36.00

36.01-38.00

NDVI0.30-0.40 0.40-0.50 0.50-0.60 0.60-0.70 0.70-0.80 0.80-0.90

2.01-2.202.21-2.402.41-2.602.61-2.802.81-3.003.01-3.20

3.21-3.40

3.41-3.60

3.61-3.80

3.81-4.00

4.01-4.20

4.21-4.40

4.01- 8.008.01-12.00

12.01-16.0016.01-20.0020.01-24.0024.01-28.0028.01-32.0032.01-36.00

36.01-40.00

40.01-44.00

44.01-48.00

48.01-52.00

52.01-56.00

2.7-3.03.1-3.53.6-4.04.1-4.54.6-5.05.1-5.55.6-6.06.1-6.56.6-7.0

7.1-7.5

7.6-8.0

8.1-8.5

8.6-9.0

9.1-9.5

9.6-10.0

N


14/16

12

When water is the limiting factor, more availablewater makes the plants grow better, resulting inhigher within-season NDVI levels and higher end-of-season crop and root biomass at the north-westside compared to the south-east side of the field.After harvest, more remnant stubble and root wasleft on the field at the north-west side. This causedorganic matter levelsand consequently soilstructure, retention of water by organic matter, andinfiltrationto remain higher at the north-west sideof the field than at the south-east side under zerotillage with residue removal. In that way soil watercontent at the north-west side improved furthercompared to the south-east side and consequentlyplant growth and crop performance were better atthe north-west side. Over the years, this incorrectagronomic management, i.e. zero tillage withresidue removal, increased the spatial variability insoil properties and crop performance since spatial

variability was low in the field when the experimentwas started (Lopez-Noverola, 1995).The variability in soil attributes induced spatialvariability in crop performance: under zerotillage with residue removal, soil quality and cropperformance followed micro-topography withhigher values where elevation was lower. Apartfrom that, the general soil quality degradationunder zero tillage with residue removal causedstressful conditions. As stated previously, stressfulconditions increase plant-to-plant competition for

resources and this competition results in greaterplant variability (Martin et al., 2005). The zero tillagepractice with residue removal caused a non-uniformdistribution of crop performance within the field,indicating the inefficient use of available resourceswith consequent yield losses, since in some partsof the field a higher yield was achieved than inothers within the specific conditions of climate andgermplasm.

Maps of plots under zero tillage with residueretention contrasted sharply with those of plots

under zero tillage with residue removal. The soilmoisture content, infiltration (time-to-pond), soilstructure (MWD obtained through dry sieving),

and organic matter content were uniformly highunder zero tillage with residue retention (Figure 4),whereas zero tillage with residue removal showedhigher values where elevation was lower. The lack ofthe effect of topography in residue-retained systemsis a consequence of the impedance of run-off due tothe presence of a mulch layer (Govaerts et al., 2006a)and a commensurate reduction in the evaporativeloss of soil water (Scopel et al., 2004). Both increasedthe amount of water available for the crop, ensuringan even crop performance throughout the field. Thehigh soil quality reduced competition for resourcesand in that way plant variability. Values andvariability for soil attributes and crop performancewere intermediate under conventional tillage.

The foregoing shows that crop performance followedthe same pattern as soil moisture and relatedattributes, such as infiltration, soil structure, and

organic matter. Thus, soil moisture is the systemslimiting factor. To develop sustainable managementpractices for this target zone, moisture captureand storage must be optimized. The intermediatesoil quality under conventional tillage could causeproblems in dry years when soil water conservationis crucial to avoid stress.

More details on the case study can be found inVerhulst, N., B. Govaerts, K.D. Sayre, J. Deckers, L.

Dendooven. 2009. Using NDVI and soil qualityanalysis to assess influence of agronomic management

on within-plot spatial variability and factors limitingproduction. PlantandSoil 317: 41-59.

Further readingKravchenko, A.N., G.P. Robertson, K.D. Thelen, R.R.

Harwood. 2005. Management, Topographical enWeather Effects on Spatial Variability of Crop GrainYields.AgronomyJournal 97: 514-523.

Kravchenko, A.N., D.G. Bullock. 2000. Correlation of cornand soybean grain yield with topography and soilproperties. AgronomyJournal 92: 7583.

Robertson, G.P., K.L. Gross. 1994. Assessing theheterogeneity of below-ground resources: Quantifyingpattern and scale. In M.M. Caldwell, R.W. Pearcy

(eds.).PlantExploitationofEnvironmentalHeterogeneity.Academic Press, New York, New York, USA: 237-253.


15/16


16/16

Apdo. Postal 6-641, 06600 Mexico, D.F. Mexico

www.cimmyt.org

International Maize and Wheat Improvement Center

The normalized difference vegetation index (NDVI) Greenseeker(TM) handheld sensor: toward the integrated evaluation of crop management. Part A - Concepts and case studies

Documents