Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition

G Model

ARTICLE IN PRESS

AGWAT-4407; No. of Pages 10

Agricultural Water Management xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Agricultural Water Management journal homepage: www.elsevier.com/locate/agwat

Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition Célia Toureiro, Ricardo Serralheiro ∗ , Shakib Shahidian, Adélia Sousa ICAAM, Institute of Mediterranean Agricultural and Environmental Sciences, University of Évora, Portugal

a r t i c l e

i n f o

Article history: Received 13 September 2015 Received in revised form 1 February 2016 Accepted 13 February 2016 Available online xxx Keywords: Remote sensing Vegetation Index Crop coefficients Water balance

a b s t r a c t Water use control methods and water resources planning are of high priority. In irrigated agriculture, the right way to save water is to increase water use efficiency through better management. The present work validates procedures and methodologies using remote sensing to determine the water availability in the soil at each moment, giving the opportunity for the application of the water depth strictly necessary to optimise crop growth (optimum irrigation timing and irrigation amount). The analysis is applied to the Irrigation District of Divor, Évora, using 7 experimental plots, which are areas irrigated by centrepivot systems, cultivated to maize. Data were determined from images of the cultivated surface obtained by satellite and integrated with atmosphere and crop parameters to calculate biophysical indicators and indices of water stress in the vegetation—Normalized Difference Vegetation Index (NDVI), Kc, and Kcb. Therefore, evapotranspiration (ETc) was estimated and used to calculate crop water requirement, together with the opportunity and the amount of irrigation water to allocate. Although remote sensing data available from satellite imagery presented some practical constraints, the study could contribute to the validation of a new methodology that can be used for irrigation management of a large irrigated area, easier and at lower costs than the traditional FAO recommended crop coefficients method. The remote sensing based methodology can also contribute to significant saves of irrigation water. © 2016 Elsevier B.V. All rights reserved.

1. Introduction and objectives For a sustainable irrigation management, crop evapotranspiration (ETc) should be determined, as precisely as possible. The weakest link in this weather-based approach to predict crop water use and irrigation requirement is the difficulty in reliably estimate the crop coefficient (Trout and Johnson, 2007). Crop coefficients are commonly estimated based on days since planting or (occasionally) growing degree days (Allen et al., 1998). For greater accuracy, in the place of a single Kc, dual crop coefficients may be considered as described in Allen et al. (1998): a basal Kcb, accounting for the dependence of ETc on the genetic characteristics of the crops, through transpiration; and a soil evaporation coefficient, Ke, which accounts for the degree to which the soil is covered by the crop, mainly referring to the evaporative component of ETc. A soil water balance will allow for the determination of the irrigation opportunity and crop water requirement (Allen et al., 1998). The best results with this methodology are achieved if on site determinations of soil

∗ Corresponding author. E-mail addresses: [email protected] (C. Toureiro), [email protected] (R. Serralheiro), [email protected] (S. Shahidian), [email protected] (A. Sousa).

water status are compared to the estimated values obtained from climatic data. Relationships can be defined between these types of data – weather-based and in situ determined – and biophysical parameters derived from vegetation indices (VI) that can be obtained from multispectral images, through convenient empirical equations. These relationships will incorporate the eventual influences of local factors such as crop, soil, and topography. If such empirical equations are valid and reliable for a given crop in the region, they may be used for defining the crop water balance parameters from remote sensed data, instead of using the corresponding in situ parameters, which are harder and more expensive to obtain. Research with maize has shown improvements in irrigation scheduling, due to better water-use estimation and more appropriate timing of irrigations, when Kcb estimates derived from remotely sensed multispectral vegetation indices were incorporated into irrigation-scheduling algorithms (Hunsaker et al., 2003). In the present work, remote sensing determined biophysical parameters are incorporated in the crop water balance of a maize crop in an irrigated region, in order to approach a new technology for definition of crop irrigation requirements for a wide irrigated area in that region.

http://dx.doi.org/10.1016/j.agwat.2016.02.010 0378-3774/© 2016 Elsevier B.V. All rights reserved.

Please cite this article in press as: Toureiro, C., et al., Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.02.010

G Model AGWAT-4407; No. of Pages 10

ARTICLE IN PRESS C. Toureiro et al. / Agricultural Water Management xxx (2016) xxx–xxx

2

Nomenclature CC CE CR Dr DP ETo ETc ETc* ETca fc fc* h I Kc Kcb Ke Kc* Kcb* Ke* LAI LRO NIR NDVI NDVIs P R RMSE VI ␪c ␪m NIR R

Field capacity Wilting point Capillary rise Water deficit within the soil Deep percolation Reference evapotranspiration Crop evapotranspiration, crop water requirement Spectral crop evapotranspiration (from NDVI) Actual crop evapotranspiration, crop water consumption Cover fraction Spectral cover fraction (from NDVI) Crop height Irrigation depth Crop coefficient Basal crop coefficient Evaporative soil surface coefficient Spectral crop coefficient Spectral basal crop coefficient Spectral evaporative soil surface coefficient Leaf area index Soil water content limit for optimal crop development Near infrared radiation Normalized Difference Vegetation Index Normalized Difference Vegetation Index for bare soil Precipitation Radiation in the red band Root mean squared error Vegetation Indices Soil moisture content calculated using satellite Soil moisture content measured using TDR Observed reflectance in the near infra-red band Observed reflectance in the red band

In this context, multispectral cameras installed on satellites provide images of the earth surface that are used to estimate the crop coefficient Kc and other crop parameters such as the fraction of land cover (fc) and the leaf area index (LAI). These crop parameters can be estimated from vegetation indices defined on the multispectral and thermal images obtained with cameras installed aboard the satellites or other vehicles. Vegetation indices (VI), computed as differences, ratios, or linear combinations of reflected light in the visible (blue, green, or red) and near infrared (NIR) spectra have been found to be closely related to several crop growth parameters (Heilman et al., 1982; Jackson and Huete, 1991; Moran et al., 1994). Satellite imagery from Landsat 5 was of broad utility in the present work, although spatial and temporal resolutions present severe limitations. Alternative satellites have had similar problems, not providing great help in this context. However, the next future is much more promising, as some countries have recently launched or are planning to launch a new generation of satellites that can overcome most of such limitations by frequently providing quality images highly suitable for precision agriculture, inclusive the irrigation scheduling (Mulla, 2013). Special relief is due to the NASA Sentinel 2 satellite, with multispectral high resolution images (up to 10 m), with a relatively short recurrence interval (5 days), quite suitable for irrigation management purposes. Several authors (Neale et al., 1989; Choudhury et al., 1994; Bausch, 1995; González-Piqueras, 2006) have observed that NDVI is well related to the water use and transpiration of the plants.

Therefore, the definition of reliable relationships between NDVI and Kcb and Kc became of main concern. Reginato et al. (1985), Neale et al. (1989, 1996, 2003, 2005), Jackson et al. (1980), Heilman et al. (1982), Bausch and Neale (1987), Michael and Bastiaanssen (2000), Jochum et al. (2002), Anderson et al. (2007), Hunsaker et al. (2005), Zhang and Wegehenkel (2006), Gonzalez-Dugo et al. (2009), Droogers et al. (2010), and Allen et al. (2011a,b) are some of the authors that have defined relationships for deriving crop coefficients from NDVI and other indices based on the reflectance of the crop surface, determined from multispectral images. It is clear that the authors are looking for the most reliable possible relation between the indices and the crop parameters. Correlation equations have been defined to relate biophysical crop parameters, such as land cover fraction fc and leaf area index LAI, with NDVI and other vegetation indices (Bausch and Neale, 1987; Neale et al., 1989; González-Piqueras, 2006; Calera-Belmonte et al., 2005; González-Dugo and Mateos, 2008). The present work is a first step in a research program with the general objective of testing and validating procedures for obtaining information on crop water status and growth stages, estimating crop coefficients, evapotranspiration and crop irrigation requirement using satellite images. This information can be extremely useful to elaborate, over short term intervals, GIS maps of crop water status and irrigation requirements in any large irrigation area, in order to serve as basis for an irrigation advisory system. In the present case, the study was applied to seven experimental plots in the irrigated area of Divor, close to Évora, south Portugal. The objective is to get relations between NDVI and crop parameters (basal Kcb and global Kc crop coefficients, land cover fraction fc, leaf area index LAI) that could be used in the referred regional irrigation management system. A first attempt is done to validate equations in Table 1 against locally determined data, looking for the possibility of adapting them for use in the regional context. The specific objectives of this work can therefore be enumerated as: • Use of Landsat 5 satellite multispectral images for determination of crop water requirements pertinent to the maize crop in the specific Mediterranean context, through the determination of vegetation indices and estimation of crop coefficients. • Validation of equations to calculate basal (Kcb) and global (Kc) crop coefficients from a vegetation index (Normalized Difference Vegetation Index, NDVI). • Validation of the empirical relation between Kcb and the vegetation index, through the determination of biophysical parameters leaf area index (LAI) and soil cover fraction (fc). • Validation of all estimations from remote sensing with on-site monitoring of soil water content, water applied with irrigation, and water balance. 2. Methods and experimental conditions 2.1. Experimental design The experiment area, the irrigation district of Divor, Évora, Portugal (38◦ 44 N; 7◦ 56 W, 309 m), has about 500 ha of irrigated land (Fig. 1). In 2007, seven experimental plots were prepared, cultivated to maize, and irrigated with centre pivot systems. Maize is the main irrigated crop of the region, with more than 50% of the total irrigated area (245 ha in the seven experiment plots). The climate is Mediterranean, with 50% or more of the 550 mm average annual precipitation falling during winter (November to February) and next to zero during summer (mid-June to midSeptember). Average monthly temperatures vary from 10 ◦ C in January to 24 ◦ C in August. Sunshine hours reach 2800 h a year,

Please cite this article in press as: Toureiro, C., et al., Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.02.010

G Model AGWAT-4407; No. of Pages 10

ARTICLE IN PRESS C. Toureiro et al. / Agricultural Water Management xxx (2016) xxx–xxx

3

Table 1 Relations between NDVI and Kc*, Kcb*, fc*, and LAI* as presented in reference works. Parameter to estimate

Relation

References

Crop coefficients Kc* and Kcb*

Kcb* = 1.36 NDVI − 0.06 Kcb* = 1.81 NDVI + 0.026 Kc* = 1.15 NDVI + 0.17 Kcb* = 1.46 NDVI − 0.19 Kc* = 1.25 NDVI + 0.2 Kcb* = 1.56 NDVI − 0.1

Bausch and Neale (1987) Neale et al. (1989) González-Piqueras (2006)

Land covering fraction fc*

fc* = 1.19 NDVI − 0.16 fc* = 1.35 NDVI − 0.2011

González-Piqueras (2006) Calera-Belmonte et al. (2005)

Leaf area index LAI*

LAI* = 1/0.8 ln (0.91 − NDVI/0.91–0.14) LAI* = 0.0287 e(5.081NDVI)

González-Piqueras (2006)

annual average reference evapotranspiration reaching 1225 mm. Average relative humidity varies from about 70 to 75% and average wind speed from 1.5 to 2.0 m s−1 . Most soils (81%) are Luvisol type, the other (19%) are Fluvisol soils. Luvisol soils are very representative of pedogenic processes under Mediterranean climate. Fig. 1 represents the experiment field. In order to determine in situ soil and crop parameters (procedure 1), six monitoring stations were prepared in each of the seven experimental plots. At each station, an access tube was installed for a time domain reflectometry (TDR) probe (TRIME-FM3), inserted 60 cm into the soil, allowing for the determination of volumetric soil water content (SWC) during the irrigation season. Observations were carried out at each station on a weekly basis. Soil water content was determined by integrating data obtained with the TDR probe within the 60 cm depth of the root zone (Toureiro et al., 2007). 2.2. Crop and irrigation parameters Crop development was monitored regularly through observation of maize growth stages. Crop height (h) was directly measured. Leaf area index (LAI) was obtained from leaf length L and width W measurements, using Eq. (1) proposed by Cavaco (2007) for maize crop: LAI = c ×



(W × L)

(1)

c being an empirical adjusting factor. Cavaco (2007) found a 0.76 value for c of a maize crop in Divor region. The irrigation depth was measured in each station using a calibrated precipitation gauge. Actual evapotranspiration or crop water consumption (ETca) is a fundamental term of a soil water balance. 2.3. ETc with FAO methodology ETc was determined using the FAO recommended crop coefficients methodology (Allen et al., 1998), a procedure that starts with the calculation of the reference evapotranspiration (ETo) by the Penman–Monteith method, and then determines ETc by replacing the crop coefficient Kc in Eq. (1) with FAO recommended values of Kcb and Ke, according to the described FAO dual crop coefficient method. Reference evapotranspiration ETo is usually used, combined with agronomic information (type of crop and its growth stage) which define a crop coefficient Kc that allows for the calculation of ETc as: ETc = KcETo

(2)

ETc is assumed to be the evapotranspiration of a full developing crop, not withstanding any water stress, therefore indicating crop water requirement. A particular case of ETc is actual evapotranspiration ETca, which may account for some eventual restriction

Calera-Belmonte et al. (2005)

in water availability for the crop. Therefore, ETca represents the (effective) crop water consumption. 2.4. In situ water balance The daily values of ETca were obtained through a water balance:

␪mi = ␪mi−1 + Pi + Ii + CRi − ROi − DPi − ETcai

(3)

where the index i is the day in the calculation procedure, ␪m is water content within the soil root depth, P is total rainfall (precipitation), I is the effective depth of water applied with irrigation (infiltrated depth), CR is capillary rise from deep (free) water table, RO is run off, and DP is deep percolation (vertical water movement from the soil to the subsoil). All terms in the water balance are in mm. The terms determined by monitoring the soil in a weekly basis were converted into daily values assuming a linear variation during the week. Soil water terms (␪m and DP) can be calculated by integrating the TDR probe measurements. Runoff and precipitation are measured and capillary rise estimated or neglected. Therefore, ETca is the only unknown term in Eq. (3), and can thus be obtained. This value is assumed as ETc if ␪m is greater than the soil moisture limit for optimal crop development LRO. Daily precipitation Pi can be ignored if it is less than 0.2 ETo, because it will evaporate directly from the soil surface. CR can normally be assumed to be zero when the water Table is more than about 1 m below the bottom of the root zone (Allen et al., 1998). 2.5. Agronomic parameters and ETc with data from multispectral images The multispectral images used in this work were obtained by Landsat 5 satellite, during the every 16 days. The multispectral camera installed aboard this satellite produces images of the ground, covering the entire electromagnetic spectrum. Probably the weakest feature of these images is the 16 day interval between them. The spatial resolution of these data is 30 m in the shortwave bands. This satellite provides all multispectral bands (blue, green, red, near infrared (NIR)). As often happens with satellites, some images of the 2007 irrigation campaign were covered with clouds, and thus the radiometric determinations were not possible, and these images could not be used (29 June, 15 July, 31 July, 16 August and 01 September). The preliminary processing of the images – geometrical and atmospheric corrections – was carried out by the DEMETER project team (Calera-Belmonte et al., 2005), with the following processing steps: • Used software: ERDAS Imagine 9.1; • Geometrical correction and geo-reference: control point uniformly distributed on the images;

Please cite this article in press as: Toureiro, C., et al., Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.02.010

G Model AGWAT-4407; No. of Pages 10 4

ARTICLE IN PRESS C. Toureiro et al. / Agricultural Water Management xxx (2016) xxx–xxx

Fig. 1. Map of the experiment area, the irrigation district of Divor, with the seven experiment plots.

• Radiometric correction: the radiometric data received were transformed in radiance values; • Atmospheric correction: Absolute correction.

One common vegetation index is NDVI (Normalized Difference Vegetation Index): NDVI =

NIR − R NIR + R

(4)

Please cite this article in press as: Toureiro, C., et al., Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.02.010

G Model AGWAT-4407; No. of Pages 10

ARTICLE IN PRESS C. Toureiro et al. / Agricultural Water Management xxx (2016) xxx–xxx

5

2,5

1,4 1,2

2,0 1,5

0,8

h (m)

NDVI

1,0

0,6

1,0

0,4 0,5

0,2 0,0 0 10 20 30 40 50 60 70 80 90 100 110 120

0,0 0 10 20 30 40 50 60 70 80 90 100 110 120

Days aer sowing

Days aer sowing

Fig. 2. Relations between crop height and NDVI for two separate plots (P1 and P2).

where NIR and R are the observed values of reflectance in the near infra-red and red bands, respectively. NDVI can vary from −1 to 1. Values close to 1 occur with dense vegetation, which has high reflectance values in the NIR band and low in the R band. This is due to the high absorption of radiation in the red band by the photosynthetic pigments, and the low absorption in the near infrared band by the cell structures of the leaves. Where vegetation is less dense, NDVI values approach zero. Negative values occur with free water surfaces and clouds, which have high reflectance in the visible band of the spectrum and low reflectance in the NIR band. The ETc* values were calculated using spectral global Kc* and basal Kcb* crop coefficients obtained from NDVI values. Jackson et al. (1980) observed the similarity between the temporal evolution of NDVI and crop coefficients. Bausch and Neale (1987) and Neale et al. (1989) defined a linear relation between Ke* and NDVI for maize. Wright (1982) determined Ke* combining land cover fraction, fc, with the leaf area index LAI. This procedure was also followed in the present work. Looking at the objectives stated in former section, relationships in Table 1 should be validated for the regional context, incorporating crop parameters determined by three comparative procedures:

Table 2 Correlation equations for agronomic and biophysical parameters obtained from remote sensing for the maize crop in the experiment field of Divor.

1) In situ soil water balance based on field measurements of soil moisture content. This is the reference procedure, requiring actual experiment field determinations. 2) ETc determination based on FAO recommended crop coefficients (Allen et al., 1998). 3) ETc* determination based on crop parameters determined from relationships with NDVI.

LAI = −

Parameters

Correlation equations

R2

Land cover fraction fc and NDVI Leaf area index and NDVI

fc* = 0.136 e2.12NDVI LAI* = 0.268 e3.43NDVI

0.89 0.87

Global crop coefficient and NDVI Basal crop coefficient and NDVI

Kc* = 0.918 NDVI + 0.303 Kcb* = 1.464 NDVI − 0.253

0.82 0.86

less than 80%, it can be related to NDVI (González-Piqueras, 2006) as: fc∗ = 1.19(NDVI − NDVIs)

(5)

where NDVIs, corresponding to bare soil, can be taken as 0.14 (González-Piqueras, 2006). Leaf area index, LAI, corresponds to the ratio of the total area of green leaves to the area of the soil that they cover. It represents the photosynthetic capacity of the crop. Values of LAI between 3 and 5 are common for full crop development stage. Baret et al. (1989) and Gilabert et al. (1996) defined a mathematical relation between NDVI and LAI that can be expressed by Eq. (6): 1 ln C

 NDVI

− NDVI NDVImax − NDVIs max



(6)

where NDVImax is the limiting value of vegetation index at large LAI values, NDVIs is the vegetation index for bare soil, and C is related with the extinction of radiation through the canopy. The following values can be used as typical with multispectral cameras (GonzálezPiqueras, 2006): NDVImax = 0.91; NDVIs = 0.14; C = 0.8. 3. Results and discussion

Some of these equations are listed in Table 1. The * on the name of the parameters indicates that it is derived from NDVI, therefore being based on remote sensing multispectral data. This distinction in the nomenclature of the parameters is also adopted for the rest of this article. In the present work, NDVI (Eq. (2)) was used to establish fundamental correlations with agronomic and hydrological parameters. Actual crop evapotranspiration depends on fractional cover, fc, which is the fraction of area of vertical projection of the vegetation surface on the soil surface. Johnson and Trout (2012) observed that NDVI was strongly related (R2 = 0.96) to green fractional cover across a broad variety of annual and perennial crop types. If fc is

In this section two sets of results will be presented and discussed: those referring to crop parameters or “agronomic characteristics”, and those referring to evapotranspiration and crop irrigation requirements, or “hydrological characteristics”. 3.1. Crop parameters Fig. 2 shows crop height, when compared with NDVI evolution. It can be seen that the index increases as the crop develops, peaking some 80 days after sowing, and then decreasing, thus indicating the senescence of the crop at the end of the growth season. This

Please cite this article in press as: Toureiro, C., et al., Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.02.010

G Model

ARTICLE IN PRESS

AGWAT-4407; No. of Pages 10

C. Toureiro et al. / Agricultural Water Management xxx (2016) xxx–xxx

0.9

1.0

0.8

0.9

0.7

0.8

0.6

0.7

0.5

0.6

fc FAO

fc*

6

0.4

0.5 0.4

0.3

0.3

0.2

0.2

0.1

0.1

0.0 0.0

0.2

0.6

0.4

0.8

0.0

1.0

0.0

0.2

0.4

fc in situ

0.6

0.8

1.0

fc in situ

Fig. 3. Linear relations of fc* and fcFAO with fc in situ: fc* = 0.78 (fc in situ) + 0.099 with R2 = 0.84; fcFAO = 0.93 (fc in situ) + 0.117 with R2 = 0.89.

6 P1

P2

P3

P4

P5

P6

P7

5

LAI

4

3

2

1

0 0

0.2

0.4

0.6

0.8

1

NDVI

1.4

1.4

1.2

1.2

1.0

1.0

0.8

0.8

0.6

0.6

Kcb*

Kc*

Fig. 4. Correlation between NDVI obtained from satellite images and LAI obtained by in situ determinations at the monitoring stations: LAI = 0.268 e3.439 NDVI , with R2 = 0.87.

0.4 0.2

0.4 0.2

0.0

0.0 0

0.2

0.4

0.6

0.8

1

0

NDVI

0.2

0.4

0.6

0.8

1

NDVI Fig. 5. Linear relations of Kc and Kcb with NDVI.

ability of NDVI to detect the senescence of the crop is important for irrigation management. In Fig. 2 a lack of information can be observed in the values of NDVI during the fastest growing phase. This is due to a gap (about 30 days) in the information from satellite imagery, due to the presence of clouds in multispectral images.

In general, the duration of the growing phase was longer than the values encountered in FAO works, while phase 3, flowering and grain formation, was around 20 days shorter than FAO reference values. This may be due to a late planting date, with relatively high air temperatures, increasing the crop growth rate. Fig. 3 compares land cover fraction during the growing stages as determined from in situ observation (fc in situ), satellite images

Please cite this article in press as: Toureiro, C., et al., Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.02.010

G Model AGWAT-4407; No. of Pages 10

ARTICLE IN PRESS C. Toureiro et al. / Agricultural Water Management xxx (2016) xxx–xxx

fc∗ = 0.136e2.123 NDVI

(7)

Fig. 4 shows the relation between NDVI and LAI for maize crop in the experiment plots. It can be observed that NDVI increases rapidly until a LAI of approximately 3, and then increases slowly, possibly due to the soil surface being almost completely covered with leaves. These results confirm those obtained in other works (Carlson and Ripley, 1997; Gamon et al., 1992). Trout and Johnson (2007) observed that NDVI increased with LAI to about 0.8, but did not increase further with increasing LAI. They attributed this finding to NDVI levelling off at high vegetation biomass. The best fit between LAI and NDVI is obtained by an exponential curve (Eq. (8)) with an R2 of 0.87: LAI∗ = 0.268e

3.4396NDVI

(8)

In Fig. 5 linear relations can be observed between NDVI determined using Landsat 5 images, and the kc and Kcb coefficients calculated using water balance and in situ observed data. Kc can be correlated to NDVI through Eq. (9): Kc = 0.918 NDVI+0.303

(9)

with R2 = 0.82. Kcb can also be correlated to NDVI through Eq. (10): Kcb = 1.464 NDVI - 0.253

(10)

R2

with = 0.85. The equations in Table 1 proposed by several authors for Kc* and Kcb* were fitted to the same set of field data in the present work, and an ANOVA (simple factor) analysis was performed. The results show that with the equations of González-Piqueras (2006) the correlation coefficients are lower than with Eqs. (9) and (10), but the differences are not significant. With the equations proposed by other authors the correlation coefficients are lower than with Eqs. (9) and (10) and the differences are statistically significant. The conclusion is that Eqs. (9) and (10) can be used for obtaining Kc* and Kcb* crop parameters from NDVI for regional conditions. The relation between Kc* and NDVI can sometimes vary and show lesser determination coefficients, due to evaporation from the soil, especially during the initial growing stages, when fc is small. Therefore, special care is required if using these data to calculate crop water requirements. On the other hand, Kcb* may be less stable than Kc*. In these cases Kc* can be used and then calculated on a weekly basis, attenuating daily fluctuations (Calera-Belmonte et al., 2005). Table 2 shows the correlation equations obtained in this work for relating biophysical indicators and agronomic parameters. 3.2. Evapotranspiration and crop irrigation requirements The values obtained for actual crop evapotranspiration (ETca, based on in situ measured data) and spectral crop evapotranspiration (ETc*, based on vegetation indices from satellite images) are characteristic to the vegetation cover at the moment the determination is carried out, with all associated climatic, soil, and vegetation conditions, influencing the values of calculated Kc*. In Fig. 6 actual measured soil moisture content, ␪m, is compared with soil moisture content calculated through a water balance using satellite data, ␪c. The results show a good correlation, with R2 = 0.78 and a RMSE = 7.83 mm. Fig. 7 shows the evolution of soil water content obtained using three types of data – in situ measurements (actual values), water

130

Water Balance (Satélite) (mm Soil Water Content)

(fc*), and the FAO methodology (Allen et al., 1998) (fc FAO). The relation between NDVI and fc* is often referred to as linear (e.g., González-Dugo et al., 2006; Calera-Belmonte et al., 2005). However, in the present work the best fit was obtained with an exponential function, with a coefficient of determination of 0.89 (Eq. (7)).

7

110 90 70 50 30 10

10

40 70 100 Soil Water Content in mm (TDR)

130

Fig. 6. Correlation between soil water balance calculated using satellite data and in situ measurement of soil moisture content: (␪c = 0.918 ␪m + 4.77).

balance using satellite imagery, and water balance using FAO recommended coefficients – during the growing season in two experimental plots. Soil water limits (field capacity, CC, wilting point, CE, and the recommended lower limit for optimal yield of the maize crop, LRO) are also represented. It can be observed that in both plots the farmers allowed the soil to dry to below the recommended levels at the early stages of the crop growth. In plot P5, the farmer was later able to restore the soil moisture level to recommended values, while in P6, the soil moisture content remained essentially bellow the recommended levels for most of the growing season. The results indicate that both the satellite image and the FAO methods provided results that were close to those obtained by direct soil measurement. The statistical – ANOVA – analysis reveals that the values do not have significant differences. Fig. 8 shows crop evapotranspiration as calculated by the three soil water balance procedures: (a) actual ETc determined from soil water content and irrigation data measured in situ at the monitoring stations; (b) spectral ETc* determined from Landsat 5 satellite imagery; (c) ETc (FAO) determined using the crop coefficients Kc recommended in Allen et al. (1998). The results show that the spectral ETc* follows closely the actual ETca during most of the growth season, while the FAO ETc tends to overestimate the crop water needs during almost all the growth season. Table 3 presents the values of crop evapotranspiration (ETca, ETc*, and ETc), precipitation and irrigation depth, applied to the four main stages of crop growth (1—initial development; 2—active growing; 3—flowering and grain formation; 4—maturity), for the 7 experimental plots. The total values of ETc calculated with FAO crop coefficients are higher than the values determined from experimental data, whether they were obtained by in situ monitoring or by satellite information. This tendency is accentuated for the growth stages with higher water requirements (stages 2 and 3). Table 4 presents the corresponding net crop irrigation requirements for each of the seven experimental plots. It can be observed that the net irrigation requirements varied from 334 mm in plot 1 and 434 mm in plot 5. These results indicate that even fields with such similar conditions can have very diverse irrigation needs, requiring site specific irrigation management. Analysis of net crop irrigation requirements shows that irrigation management based on FAO methodology (Allen et al., 1998) can lead to an overestimation of up to 20% of the irrigation needs, if compared to in situ or remote sensing determinations of net crop irrigation requirements. A statistical – ANOVA – analysis shows that these differences between results with the FAO methodology and the other two methods of soil water balance are statistically significant. The reasons for such a large difference may be easy to understand: FAO methodology is universal, for broad and generic application, and might need calibration for local crop varieties and conditions. Therefore, the environmental and economic benefits that may be

Please cite this article in press as: Toureiro, C., et al., Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.02.010

G Model

ARTICLE IN PRESS

AGWAT-4407; No. of Pages 10

C. Toureiro et al. / Agricultural Water Management xxx (2016) xxx–xxx

8

120

P5

Soil Water Content (mm)

100 80 60 40 20 0 38

47

55

62

68

76

84

88

97

102

95

102

111

Days aer sowing 160

P6

Soil Water Content (mm)

140 120 100 80 60 40 20 0

45

54

62

70

77

84

90

Days aer sowing CC TDR

CE Balance NDVI

LRO Balance FAO

CC - Field capacity; CE - Wilng point; LRO - Soil water content limit for opmal crop development Fig. 7. Soil water balance in experiment plots P5 and P6 during the growing season, showing results obtained with in situ, FAO, and satellite information.

ETc (FAO)

ETca

ETc*

20

ETcFAO, ETca, ETc* (mm/day)

18 16 14 12 10 8 6 4 2 0 8

13

18

23

28

33

38

43 48 53 58 Days aer sowing

63

68

73

78

83

88

Fig. 8. Crop evapotranspiration determined by the three methods of soil water balance: in situ determinations (actual Kc), FAO recommended coefficients (FAO Kc), and satellite imagery derived data (Kc*).

Please cite this article in press as: Toureiro, C., et al., Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.02.010

G Model

ARTICLE IN PRESS

AGWAT-4407; No. of Pages 10

C. Toureiro et al. / Agricultural Water Management xxx (2016) xxx–xxx

9

Table 3 Crop evapotranspiration (ETca, ETc*, and ETc) obtained from in situ (Kca), satellite imagery (Kc*), and FAO recommended (Kc) crop coefficients. Plot

Growth stages

Crop evapotranspiration (mm)

Crop water input (mm)

ETca (in situ)

ETc* (satellite)

ETc (FAO)

Effective rain

Irrigation

Total input

1

1 2 3 4 Total

19 158 185 18 380

20 163 186 19 388

33 195 215 22 464

41 3 2 0 46

24 118 165 5 312

65 121 167 5 358

2

1 2 3 4 Total

22 175 184 28 409

21 163 180 29 393

32 187 216 36 471

16 29 2 0 47

48 190 170 21 429

64 219 172 21 476

3

1 2 3 4 Total

22 215 145 30 412

21 195 133 29 378

27 198 156 36 417

16 29 0 0 45

48 190 141 32 411

64 219 141 32 456

4

1 2 3 4 Total

31 170 198 53 452

27 145 188 52 412

41 183 219 63 506

1 43 1 2 47

3 180 214 54 451

4 223 215 56 498

5

1 2 3 4 Total

32 187 207 55 481

34 184 199 54 471

41 183 219 63 506

1 43 1 2 47

42 179 215 67 503

43 222 216 69 550

6

1 2 3 4 Total

24 212 136 83 455

25 219 145 89 478

34 235 165 107 541

42 43 1 2 88

44 148 109 58 359

86 191 110 60 447

7

1 2 3 4 Total

28 158 190 50 426

28 159 191 53 431

41 183 219 63 506

1 43 1 2 47

36 143 198 30 407

37 186 199 32 454

Growth stages: 1—initial development, 2—development stage, 3—flowing and grain formation and 4—maturation.

obtained with a regionally based irrigation management can compensate for the extra effort needed to carry it out. 4. Conclusions Remote sensing techniques based on the analysis of satellite images can be used to improve and validate procedures and methods that can contribute to an integrated irrigation management system for large irrigated areas. In the present work, three different methodologies for irrigation management were compared for maize crop irrigated by centre pivot systems, each using data obtained from a different source: (a) remote sensed data; (b) calculated data based on FAO recommendations for irrigation water management; and (c) actual experimental values measured in the experiment field. Agronomic parameters and biophysical indicators were determined using the three methodologies, and significant correlations were obtained between the parameters. Among the biophysical indicators, the “Normalized Difference Vegetation Index”, NDVI, assumed special relevance, being suitable for correlation with the other agronomic and biophysical parameters. This conclusion confirms results of several authors using similar vegetation indices based on the analysis of multispectral images of the vegetation surface. Correlation equations were determined with the information for the irrigated area of Divor. The duration of the crop growth stages, when determined by remote sensing, were similar to the actual observed values, but

were longer when determined using the parameters provided by the FAO methodology. Actual crop water requirements were 20% less than those obtained from the FAO methodology. Probably, this smaller water requirements determined with in situ or remote sensing criteria result from the combination of the shorter growth stages with the lower values of the crop coefficients Kc and Kcb, when compared to FAO data. In summary, the results of the present work show that remote sensing based on multispectral images may be used, with a high degree of accuracy and spatial representation, to calculate crop water and irrigation requirement. However, it should be noted that, for an irrigation management advisory system to be interesting and sustainable, it requires the frequent mapping of soil water status and crop irrigation requirements of the whole irrigation scheme. This is probably the weakest point of the satellite imagery methodology used in the present work: temporal representation, once each 16 days with the Landsat 5 satellite, is not enough for irrigation management. This constraint is aggravated when, as happened with the present work, some images are not usable due to the presence of clouds. Nowadays, this limitation of satellite image use is probably overcome, as accurate (10 m resolution) and frequent (5 days recurrence) data from Sentinel 2 satellite is available and can be combined with data from other satellites to provide more frequent images of the land surface.

Please cite this article in press as: Toureiro, C., et al., Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.02.010

G Model

ARTICLE IN PRESS

AGWAT-4407; No. of Pages 10

C. Toureiro et al. / Agricultural Water Management xxx (2016) xxx–xxx

10

Table 4 Net irrigation requirements determined using crop coefficients obtained in situ (Kca ), from remote sensing (Kc*), and FAO recommended (Kc). Plot

Growth stages

Net irrigation requirements (mm) Kca

Kc*

Kc

1

1 2 3 4 Total

22 155 183 18 334

21 160 184 19 342

8 192 213 22 418

2

1 2 3 4 Total

6 146 182 28 362

5 134 178 29 346

16 158 214 36 424

3

1 2 3 4 Total

6 186 145 30 367

5 166 133 29 333

11 169 156 36 372

4

1 2 3 4 Total

30 127 197 51 405

26 102 187 50 365

40 140 218 61 459

5

1 2 3 4 Total

31 144 206 53 434

33 141 198 52 424

40 140 218 61 459

6

1 2 3 4 Total

−18 169 135 81 367

−17 176 144 87 390

−8 192 164 105 453

7

1 2 3 4 Total

27 115 189 48 379

27 116 190 51 384

40 140 218 61 459

Growth stages: 1—initial development, 2—development stage, 3—flowing and grain formation and 4—maturation.

Acknowledgements The present research was supported by a grant from the Portuguese Foundation for Science and Technology (FCT) and Institute of Mediterranean Agricultural and Environmental Sciences (ICAAM). References Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration. Guidelines for computing crop water requirements. In: FAO Irrigation and Drainage Paper 56. FAO, Roma. Allen, R.G., Pereira, L.S., Howell, T.A., Jensen, M.E., 2011a. Evapotranspiration information reporting: I factors governing measurement accuracy. Agric. Water Manag. 98, 899–920. Allen, R.G., Pereira, L.S., Howell, T.A., Jensen, M.E., 2011b. Evapotranspiration information reporting: II recommended documentation. Agric. Water Manag. 98, 921–929. Anderson, M.C., Kustas, W.P., Norman, J.M., 2007. Upscaling flux observations from local to continental scales using thermal remote sensing. Agron. J. 99, 240–254. Bausch, W.C., 1995. Remote sensing of crop coefficients for improving the irrigation scheduling of corn. Agric. Water Manag. 27 (l), 55–68. Bausch, W.C., Neale, C.U., 1987. Crop coefficients derived from reflected canopy radiation: a concept. Trans. Am. Soc. Agric. Eng. 30 (3), 703–709. Baret, F., Guyot, G., Major, D.J., 1989. TSAVI: a vegetation index which minimises soil brightness effects on LAI and APAR estimation. Canada In: Proceedings of 12th Canadian Symposium on Remote Sensing, vol. 30, pp. 1355–1358. Calera-Belmonte, A., Jochum, A., Cuesta Garcia, A., Montoro Rodriguez, A., Lopez fuster, P., 2005. Irrigation management from space: towards user-friendly products. Irrig. Drain. Syst. 19 (3–4), 337–353.

Carlson, T.N., Ripley, A.J., 1997. On the relationship between fractional vegetation cover: leaf area index and NDVI. Remote Sens. Environ. 62, 241–252. Cavaco, L., 2007. Gestão da Água de Rega: utilizac¸ão Remota de um Sensor de Infravermelhos, na cultura do milho. In: Trabalho Final de Curso da Licenciatura em Engenharia de Recursos Hídricos. Universidade de Évora. Choudhury, B.J., Ahmed, N.U., Idso, S.B., Reginato, R.J., Daughtry, C.S., 1994. Relations between evaporation coefficients and vegetation indices studied by model simulations. Remote Sens. Environ. 50, 1–17. Droogers, P., Immerzeel, W.W., Lorite, I.J., 2010. Estimating actual irrigation application by remotely sensed evapotranspiration observations. Agric. Water Manag. 97, 1351–1359. ˜ Gamon, J.A., Penuelas, J., Field, C.B., 1992. A narrow-wave band spectral index that track diurnal changes in photosynthetic efficiency. Remote Sens. Environ. 41, 35–44. Gilabert, M.A., Gandia, S., Melia, J., 1996. Analyses of spectral biophysical relationships for a corn canopy. Remote Sens. Environ. 55 (1), 11–20. González-Dugo, M.P., Mateos, L., 2008. Spectral vegetation indices for benchmarking water productivity of irrigated cotton and sugar beet crops. Agric. Water Manag. 95, 48–58. González-Dugo, M.P., Moran, M.S., Mateos, L., Bryant, R., 2006. Canopy temperature variability as an indicator of crop water stress severity. Irrig. Sci. 24 (4), 233. Gonzalez-Dugo, M.P., Neale, C.M., Mateos, L., Kustas, W.P., Prueger, J.H., Anderson, M.C., LI, F., 2009. A comparison of operational remote sensing-based models for estimating crop evapotranspiration. Agric. Water Manag. 149, 1843–1853. González-Piqueras J., 2006. Evapotranspiration de la Cubierta Vegetal mediante la Determinación del Coeficiente de Cultivo por Teledetección. Extensión a Escala Regional: Aquífero 08.29 Mancha Oriental. Tese Doctoral apresentada na Universitat de Valencia. Heilman, J.L., Heilman, W.E., Moore, D.G., 1982. Evaluating the crop coefficient using spectral reflectance. Agron. J. 74, 967–971. Hunsaker, D.J., Pinter, J.R., Barnes, E.M., Kimball, B.A., 2003. Estimating cotton evapotranspiration crop coefficients with a multispectral vegetation index. Irrig. Sci. 22 (2), 95–104. Hunsaker, D.J., Barnes, E.M., Clarke, T.R., Fitzgerald, G.J., Pinter, P.J., 2005. Cotton irrigation scheduling using remotely sensed and FAO-56 basal crop coefficients. Trans. ASAE 48, 1395–1407. Jackson, R.D., Huete, A.R., 1991. Interpreting vegetation indexes. Prev. Vet. Med. 11 (3–4), 185:200. Jackson, R.D., Idso, S.B., Reginato, R.J., Pinter, P.J., 1980. Remotely sensed crop temperatures and reflectances as inputs to irrigation scheduling. ASCE, Boise, N.Y., USA In: Irrigation and Drainage Special Conference Proceedings, 1980, pp. 390–397. Jochum A.M., Belmonte A.C., Cuesta A., 2002. Tecnologías de observación de la tierra en los servicios de asesoramiento de riegos. III Congreso sobre Gestión y Planificación del Agua, Sevilla, pp. 1–10. Johnson, L.F., Trout, T.J., 2012. Satellite NDVI assisted monitoring of vegetable crop evapotranspiration in California’s San Joaquin valley. Remote Sens. 4 (2), 439–455. Michael, M.G., Bastiaanssen, W.G., 2000. A new simple method to determine crop coefficients for water allocation planning from satellites: results from Kenya? Irrig. Drain. Syst. 14 (3), 237–256. Moran, M.S., Clarke, T.R., Inoue, Y., Vidal, A., 1994. Estimating crop water deficit using the relation between surface-air temperature and spectral vegetation index. Remote Sens. Environ. 49 (2), 246–263. Mulla, D.J., 2013. Twenty five years of remote sensing in precision agriculture: key advances and remaining knowledge gaps. Biosyst. Eng., Elsevier 114, 358–371. Neale, C.M., Bausch, W.C., Heerman, D.F., 1989. Development of reflectance-based crop coefficients for corn. Trans. ASAE 32 (6), 1891–1899. Neale C.M., Ahmed R.H, Moran M.S., Pinter P.J., QI J., Clarke T.R., 1996. Estimating seasonal cotton evapotranspiration using canopy reflectance. Evapotranspiration and irrigation scheduling: Proceedings of the International Conference, November 3 to 6, San Antonio, Texas, USA, In: Camp C.R., Sadler E.J., and Yoder R.E. (Eds.), 173–181. St. Joseph, Michigan, USA: American Society of Agricultural Engineers. Neale C.M., Jayanthi H., Wright J.L., 2003. Crop and irrigation water management using high resolution airborne remote sensing. Proceedings of 54th IEC meeting of the international commission on irrigation and drainage (ICID) workshop remote sensing of ET for large regions, 17 September (2003) Montpellier, France. CD–ROM. New Delhi, India. Neale, C., Jayanthi, H., Wright, J., 2005. Irrigation water management using high resolution airborne remote sensing. Irrig. Drain. Syst. 19 (3–4), 321–336. Reginato, R.J., Jackson, R.D., Pinter, P.J., 1985. Evapotranspiration calculated from remote multispectral and ground station meteorological data. Remote Sens. Environ. 18, 75–89. Toureiro, C., Serralheiro, R., Oliveira, M.R., 2007. Maximizac¸ão da Economia da Água em Regadio: Rega Deficitária Controlada. Revista Tecnologia da Água. Trout, T.J., Johnson, L.F., 2007. Estimating crop water use from remotely sensed NDVI, crop models, and reference ET. In: Proc. Fourth Int. Conf. Irrig. Drain. USCID, Denver, CO, pp. 275–285. Wright, J.L., 1982. New evapotranspiration crop coefficients. J. Irrig. Drain. 108, 57–74. Zhang, Y., Wegehenkel, M., 2006. Integration of MODIS data into a simple model for the spatial distributed simulation of soil water content and evapotranspiration. Remote Sens. Environ. 104 (4), 393–408.

Please cite this article in press as: Toureiro, C., et al., Irrigation management with remote sensing: Evaluating irrigation requirement for maize under Mediterranean climate condition. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.02.010