Using ENVI for Agriculture Research | Optical Remote Sensing of Sugarcane Development

Introduction

Remote sensing is an effective tool for monitoring crop health and development when extensive field studies are not feasible. Using remote sensing to monitor development stages throughout the growing season (also called phenology) can help provide better estimates of productivity, yield forecasts, and biomass.

There has been a growing interest in remote sensing applications for sugarcane research in recent years. This paper provides examples of using optical remote sensing to study sugarcane phenology in Southeast Florida. It covers the following subjects:

• Sugarcane production in Southeast Florida
• Sugarcane development stages
• Spectral properties of the sugarcane plant and what vegetation indices are commonly used to assess sugarcane health and productivity
• Example of using ENVI® software to create spectral index images of sugarcane fields from RapidEye multispectral data
• Example of creating image classification maps of sugarcane development stages
• Example of using the Hotspot Analysis tool (available in the ENVI Crop Science platform) to reveal patterns of vegetation health over time

Florida Sugarcane

Florida is the largest producer of sugarcane (Saccharum officinarum) in the United States. Approximately 70 percent of commerical sugarcane is grown in Palm Beach County, south of Lake Okeechobee (Figure 1).^[1] Most sugarcane is planted from September through December and harvested from late October through mid-April. The season of active growth in the continental United States is seven to eight months. Over 80 percent of the sugarcane grown in Florida is the Canal Point (CP) variety. The total area of sugarcane production in southern Florida is approximately 410,000 acres.^[2]

Figure 1: Sentinel-2 image of sugarcane fields in Southeast Florida.

Growth Stages

Sugarcane is an annual crop with four major growth stages.^[3]

Germination and Emergence

Germination includes the time from planting new crops or harvesting previous ratoons to the germination of buds. Germination starts approximately 7 to 10 days after planting and continues for about 25 to 30 days. This is followed by the emergence of a new shoot as a primary tiller.

Key terms: A ratoon is a new sugarcane plant that grows from the stubble of a crop that has already been harvested. A tiller is a side shoot that emerges from the base of the sugarcane stem.

Tillering

After emergence, leaves develop and the primary tiller starts to develop. Tillering produces a group of upright stalks containing one primary shoot and several smaller tillers (Dillewijn, 1952). The process begins about 40 days after planting and lasts up to 120 days. Although six to eight tillers emerge from a bud, only one to two tillers remain to form canes. ^[3]

Growth

The period of active growth (also called grand growth) starts about 120 days after planting and lasts up to 270 days, for an annual crop. During this period, only 40 to 50 percent of the original tillers survive by 150 days to form canes that can be harvested. ^[3]

Maturation

The maturation period (also called ripening) lasts for about three months, starting from 270 to 360 days after planting. The sugarcane plant usually requires 10 to 12 months to reach maturity, which is characterized by the end of stalk growth (Rahman et al., 2004). Maturation is associated with decreased soil water content, temperature, and nitrogen availability, as well as increased sugar content. Begue et al. (2012) further associate maturation with decreased leaf chlorophyll and water content. During this time, sugarcane leaves are senescent and generally turn yellow or brown due to lower amounts of chlorophyll. Ripening starts at the bottom of cane stalks and moves to the top, so the bottom parts of the stalk contain more sugar than the top.

Key terms: Senescence is the aging process in plants due to chronological age or stress. It is characterized by decreased chlorophyll in leaves.

Figure 2: Young and mature sugarcane crops, Atherton Queensland. Source: Willem van Aken, January 1, 1998. Image courtesy of CSIRO [CC By 3.0 Deed], no modifications.

Harvest

Harvesting occurs when sugarcane stalks have reached peak maturity. In Florida, mechanized harvesters cut the stalks near the ground, leaving behind residue such as roots and leaves from the top of the stalk. This is called a green harvest. Some growers burn the mature sugarcane fields before cutting the stalks, which removes the green leaves as well as straw and dry leaves and makes the cutting process more efficient. This is called a burnt harvest.

Figure 5 shows what each development stage looks like from USGS aerial orthophotos (0.5-foot spatial resolution). The image from the germination/ emergence/tillering phases shows a combination of soil and vegetation pixels as new crops are planted and begin to grow. During the grand growth period, sugarcane leaves are green and at their maximum size, resulting in a canopy with a coarse texture. During the maturation phase, the texture is less pronounced and the leaves begin to turn brown. After a burnt harvest, the surface is nearly black.

Figure 5: Samples of high-resolution orthophotos, available from the U.S. Geological Survey, showing different development stages of sugarcane fields.

Figure 3: RapidEye satellite images of sugarcane fields being burned prior to harvest.

Figure 4: Sugarcane mechanized harvest operation without burning. Source: Mariordo Mario Roberto Duran Ortiz (Own work) [CC BY 3.0], via Wikipedia Commons, no modifications.

Spectral Characteristics of Sugarcane

Sugarcane has a unique spectral response that is influenced primarily by the presence of nutrients and water content of leaves (Gers, 2003). Water in sugarcane leaves produces absorption bands at 980 and 1205 nanometers in a spectral reflectance curve.

Physical characteristics also play a role in the spectral signature of sugarcane, including the following (Abdel-Rahman and Ahmed, 2008):

• Canopy architecture: size, shape, and distribution of leaves
• Foliage density
• Leaf pigments: chlorophyll a and b, carotene, xanthophyll, and anthocyanins
• Geometry of the detecting sensor
• Atmospheric conditions

The spatial pattern of sugarcane spectral properties will vary significantly in optical remote sensing imagery. This is due to the length of the planting and harvest seasons, as well as varying crop cycles (Begue et al, 2010). Also, it may be difficult to visually distinguish between different development stages by simply looking at true-color or color-infrared images. Spectral indices can help with this.

Spectral Indices Background

Using vegetation indices to study sugarcane is often better than using the reflectance of spectral bands alone (Simões et al., 2015). Chlorophyll strongly absorbs light in the red and blue wavelengths, and the leaf cell structure strongly reflects light in the near-infrared wavelengths. Over the past few decades, researchers have used the relationship between red and near-infrared reflectance to construct broadband spectral indices for estimating vegetation health.

Key terms: Broadband spectral indices use combinations of broad wavelength regions (blue, green, red, and near-infrared) to indicate the relative abundance of vegetation, water, burned areas, and other features. Narrowband spectral indices use combinations of specific wavelengths to target specific variables such as nitrogen concentration and canopy water content in vegetation.

Next is a summary of commonly used spectral indices for studying sugarcane health and development.

Normalized Difference Vegetation Index (NDVI)

This broadband index is popular for vegetation analysis since it is strongly correlated with chlorophyll content (Tucker, 1979). NDVI is based on the following formula:

With sugarcane, maximum NDVI occurs approximately two months before the next harvest begins. This is the best time to acquire satellite images to assess sugarcane yield.

Researchers have experimented with more specific wavelengths for NDVI. For example, Huang et al. (2005) reported that reflectance spectra showed maximum sensitivity to chlorophyll content in the leaves of sugarcane at 550 nanometers (green) and 710 nanometers (near-infrared). They also showed that NDVI and simple ratio indices using 750 nanometers (near-infrared) and 710 nanometers (near-infrared) revealed good correlation with chlorophyll content.

However, NDVI tends to saturate in images spanning the duration of actively growing vegetation. Benvenuti and Weill (2010) likewise found that NDVI was not effective at detecting multi-temporal differences on sugarcane spectral response throughout its lifecycle. They found a better performance with the Green Vegetation Index (GVI), using the formula from Crist (1985) that utilizes Landsat TM bands:

Another issue is that NDVI values in areas of leaf senescence can be difficult to distinguish from those showing post-harvest crop residue at the surface (Begue et al., 2012).

A study by Viña et al. (2004) found that the Visible Atmospherically Resistant Index (VARI) and its derivatives showed a more linear relationship to green vegetation fraction than NDVI. It was also more sensitive to the onset of senescence than NDVI.

Other spectral indices are more effective in detecting subtle variations in the spectral response of sugarcane over time. These are described next in the context of different development stages.

Soil-Adjusted Vegetation Index (SAVI)

This index is similar to NDVI, but it contains a soil brightness correction factor (L) to reduce the impact of soil reflectance in images of vegetation, especially during the early stages of crop growth. This index is best used in areas with relatively sparse vegetation where soil is visible through the vegetation, and it can be used to study all phases of sugarcane development. Simões et al. (2005) studied the patterns of SAVI and other vegetation indices in sugarcane.

For the best accuracy, researchers often take LAI measurements in the field with a handheld instrument, but since this is not always feasible, remote sensing can be used to estimate LAI values.

ENVI calculates this index using the formula from Huete (1988), with a value of 0.5 for the correction factor:

Leaf Area Index (LAI)

This index measures one half of the total leaf area (in meters squared) of vegetation per unit area of ground (also in meters squared). It is directly related to vegetation biomass and productivity, and it is one of the most effective indicators of sugarcane biomass. The growth rate of sugarcane is a direct function of the LAI (Meki et al., 2015).

Kross et al. (2015) found that seven commonly used broadband vegetation indices were sensitive to LAI from the emergence phase to an LAI value of 8 m2/ m2. LAI also shows the best contrast between young, healthy vegetation and the period of peak growth and greenness.

For the best accuracy, researchers often take LAI measurements in the field with a handheld instrument, but since this is not always feasible, remote sensing can be used to estimate LAI values.

ENVI uses the following formula from Boegh et al. (2002) to estimate LAI from satellite imagery. The coefficients were originally derived from MODIS data:

EVI is the Enhanced Vegetation Index (EVI) value calculated from the following Non-Export Controlled Information 9 OPTICAL REMOTE SENSING OF SUGARCANE DEVELOPMENT formula from Huete et al. (2002):

Xavier and Vettorazzi (2003) mentioned a caveat to be aware of, that LAI estimates from remote sensing are difficult to accurately obtain without being able to calibrate them to LAI measurements in a field study at the same time the satellite data were captured. Also, satellite-derived LAI measurements that are based on empirical models require calibration and validation with field measurements (Ali et al., 2015).

In general, LAI values are low in the tillering phase, compared to grand growth and maturation. Maximum LAI occurs about six months after planting, then it slowly declines during maturation. LAI shows the best contrast between young, healthy vegetation and peak growth and greenness.

Burn Area Index (BAI)

No known research studies have used this index in relation to sugarcane development; however, it can be used to highlight burnt-harvest fields. Brighter pixels indicate burned areas. ENVI uses the following formula from Chuvieco, Martin, and Palacios (2002) and Martin ( 1998) to calculate BAI:

Narrowband Indices

Various narrowband spectral indices have the potential to estimate water content and senescence of sugarcane during the maturation and post-harvest stages. These indices require super-spectral or hyperspectral imagery.

Normalized Difference Water Index (NDWI)

This index highlights the ratio of bulk caretenoids to chlorophyll. An increase in PSRI indicates the onset of senescence and plant ripening. This index is calculated using the following formula from Merzylak (1999):

Normalized Difference Water Index (NDWI)

This index indicates vegetation canopy water content. It can also be used to distinguish between green-harvest fields and burnt-harvest fields by separating dry and humid surfaces. It can distinguish bare soil from the vegetation residue left after harvest. NDWI values between green and burnt harvest are significantly different, showing negative values after a burnt harvest and positive values after a green harvest. Upon burnt harvest, moisture in the soil evaporates, which presents a condition similar to drought stress in crops. Mulianga et al. (2015) proposed a NDWI threshold of 0.27 to separate burnt and green harvest classes.

ENVI calculates NDWI using the following formula from Gao (1995):

PSRI and NDWI cannot be computed from RapidEye imagery because this imagery does not have bands centered near the required wavelengths (namely, 500, 680, 750, 857, and 1241 nanometers). However, these indices are worth further investigation with WorldView-3 data, which has eight multispectral bands and eight SWIR bands.

The next section describes the data and methods used in this paper to study sugarcane development.

Data and Research Methods

The research objectives of this paper are as follows, using ENVI version 5.4:

• Construct a time series of vegetation index images for sugarcane fields throughout a typical growing season. Plotting the vegetation index values over time reveals patterns related to sugarcane health and development.
• Create an image classification map of sugarcane fields with different development stages.
• Show how the Hotspot Analysis tool can reveal subtle variations in crop health over time.

Accomplishing these objectives requires high-resolution imagery with a frequent revisit time for the selected study area. RapidEye multispectral images were chosen for this study for the following reasons:

• Frequency of revisit time for a given area: Off-nadir scenes for a given area are usually available on a daily basis, while at-nadir images are available every 5.5 days.
• High spatial resolution: Orthorectified resolution is 5 meters, with a ground sampling distance of 6.5 meters
• Ease of availability: Images can be ordered using a search-and-order interface from Planet.com.
• Radiometric corrections: Conversion to absolute radiometric values has already been applied.
• Red-edge band: Having a band in the red-edge wavelength region shows more variations in chlorophyll content and the leaf structure.

RapidEye analytic ortho tiles have the following bands:^[4]

RadidEye provides reflectance measurements in the red-edge wavelength region, from 690 to 730 nanometers. In this part of the spectrum, the reflectance curve of healthy vegetation drastically increases from the red wavelengths toward the near-infrared plateau.^[5]

Figure 6: ENVI spectral profile of a healthy vegetation pixel from RapidEye imagery.

Seventeen RapidEye orthorectified tiles from January 24, 2016, through March 20, 2017, were selected for analysis:

January 26, 2016
February 12, 2016
February 16, 2016
March 15, 2016
May 7, 2016
June 14, 2016

July 1, 2016
July 30, 2016
September 17, 2016
October 22, 2016
November 20, 2016

December 29, 2016
January 8, 2016
January 20, 2017
February 26, 2017
March 15, 2017
March 20, 2017

More scenes were available during this time period, but they had too much cloud cover to be considered for analysis. No images were available in April and August of 2016.

Pre-Processing Steps

The following steps were taken to prepare the images for spectral analysis, classification, and hotspot analysis.

1. Select a geographic area of interest within the Florida sugarcane region. This area is shown in Figure 7.
2. For each RapidEye tile, create a spatial subset of the common area of interest. Since the images are already orthorectified in the same spatial reference (UTM Zone 17N, WGS-84), no further reprojection or image registration is necessary.
3. Correct each image to apparent surface reflectance, using the QUick Atmospheric Correction (QUAC) tool in the ENVI Atmospheric Correction Module.
4. Use Spectral Math to divide the reflectance pixel values (from QUAC) by 10,000 in each image. Pixel values still represent reflectance from 0 to 100%, but now they are scaled from 0.0 to 1.0.

Figure 7: RapidEye true-color image from January 24, 2016, near Canal Point, Florida. This image shows the area of interest from Steps 1-2.

Time Series of Spectral Index Images

LAI images were created with the Spectral Indices tool for each date. Then the Build Raster Series tool was used to create a time series of the LAI images. The Series Manager was used to animate through the series.

The next step was to create a time series of image that utilized the RapidEye Red Edge band.

Figure 8: Example of using the Series Manager to animate through the LAI images.

Ali et al. (2015) discussed how this band is more sensitive to the presence of healthy vegetation than the red band. They presented an idea to replace the red band with the red edge band in the SAVI formula. Their intent was to show more subtle variations in crop health throughout a growing season, compared to traditional broadband indices.

We chose to modify this approach by using the Optimized Soil Adjusted Vegetation Index (OSAVI) formula from Rondeaux et al. (1996), while also substituting the red band with the red edge band. The result is a Soil Adjusted Red-Edge Index (SAREI):

Substituting the NIR and red edge terms with the appropriate RapidEye bands results in the following formula:

Since this custom index is not available in the ENVI Spectral Indices tool, it can be defined using the Band Math tool.

The following figure shows an example of a SAREI image from the time series:

Figure 9: Example of using the Series Manager to animate through the SAREI images.

In the SAREI images, fields with emerging and tillering sugarcane plants showed a speckled pattern. Young sugarcane plants had high SAREI values against a background of dark soil pixels.

Although a visual analysis of LAI and SAREI images can show patterns in sugarcane development, a plot of their mean values will prove even more useful. To do this, a group of adjacent sugarcane fields with the same planting/ harvesting schedule was selected. A region of interest (ROI) was drawn to define this area from which to calculate image statistics (Figure 10).

Figure 10: A region of interest (indicated by the red box) defines the area from which statistics will be calculated.

Next, mean LAI and SAREI values were calculated from the ROI, for each date. A plot of the SAREI mean values over time shows the growing and harvesting periods (Figure 11). Sometime between February 16, 2016, and May 5, 2016, the selected field was burned and harvested. Following that are periods of germination/emergence (not colored in the plot), tillering, and active growth. Toward the end of 2016 is the maturation period, followed by another burned harvest in February of 2017:

Figure 11: Plot of sugarcane SAREI mean values over a 14-month period. The images below the plot are examples of each development stage from RapidEye true-color images.

For this analysis, the “harvest” period refers to recently burned fields and the immediate time afterward, when leaves and straw remain on the surface. Next is a plot of the LAI mean values over the same time period:

Figure 12: Plot of sugarcane LAI mean values over a 14-month period.

Compared to the SAREI plot, the LAI plot shows a more stable pattern during the tillering, grand growth, and maturation periods. As Rahman et al. (2014) mentioned, maximum LAI occurs about six months from planting and then slowly declines. LAI values should be relatively high during the maturation period. This plot reflects those observations.

The next section shows how image classification can categorize different stages of sugarcane development.

Classification Maps of Development Stages

The classification framework was introduced in ENVI version 5.4. It provides the ability to classify data using a Softmax Regression or Support Vector Machine (SVM) algorithm (also called a classifier). The classification framework is available through the ENVI application programming interface (API). Example classification code is available in the ENVI Help for users who may be new to ENVI API programming. See the ENVI Classification topic in the Harris Geospatial Solutions Documentation Center.

The classification framework can create and train a classifier on one dataset and apply it to a similar dataset.

Key terms: A trainer is an algorithm that iteratively trains a classifier in order to minimize its error. It tries to adjust the classifier’s parameters until the error (also called loss) converges on a minimum value. ENVI provides iterative and gradient descent trainers.

Here are the steps used to create classification maps of sugarcane development:

Create an Attribute Image

In most cases, only spectral bands are used for classifying features of interest. However, adding different types of data may improve the classification accuracy. Examples include elevation data, texture, shape, and spectral indices. Use the Layer Stacking tool to create an attribute image, where each band represents a different attribute. Classification will occur on the attribute image. Even if you use a multispectral image with no additional attributes, that is still considered an attribute image. It just consists of the different multispectral bands.

In this example, an attribute image was created with seven layers (bands), shown in Figure 13. The layers include the five RapidEye bands, BAI, and SAREI. LALI was excluded from the attribute image because the SAREI image already provided a good representation of sugarcane health. Having too many attributes can negatively affect the classification results. Only one attribute image is needed, not one for every date in the time series. This example is from February 16, 2016: Roads and buildings were masked out prior to creating the attribute image in order to exclude them from classification.

Next is the process of identifying areas in an image that are known to belong to certain categories.

Figure 13: Attribute image with layer (band) names listed in the Data Manager.

Collect Training Samples

Supervised classification methods such as Softmax Regression or Support Vector Machine require the collection of training samples for each class. These represent ground truth in classification. The informational classes must be known before collecting training data. This example identifies four classes for each major phase of sugarcane development:

• Emergence and Tillering
• Grand Growth
• Maturation
• Burnt Harvest

ROIs can be drawn around pixels that presumably represent these classes. A preferred method is to draw ROIs over a high-resolution orthophoto (if available) that is co-registered with the image you will classify. Or, ground-truth samples can be collected during a field study, then converted to point, line, or polygon ROIs in ENVI. For this example, ROIs were drawn on a true-color RapidEye image from February 16, 2016. Four sets of ROIs were created, one for each development stage (Figure 14).

Figure 14: Training sample ROIs overlaid on a RapidEye true-color image. The Layer Manager shows the corresponding classes that each ROI represents.

Since field observations were not available to validate the actual development stages, a combination of different spectral indices were used to assist in the selection of training samples. These include LAI, BAI, and SARE.

Figure 15: Multi-view display in ENVI showing the training sample dataset and three co-registered spectral index images.

The following criteria were used to select training samples for each development stage:

Emergence and Tillering

• Fields that are nearly black in the true-color image
• Speckled appearance in the SARE image
• Medium grey in the LAI image

• Fields with a prominently green color in the true-color image, and:
  • White in the SARE index image
  • White in the LAI image

• Fields that are brown or greenish-brown in the true-color image
• Medium grey in the SARE and LAI images

• Fields that are dark brown (some with white spots) in the true-color image
• White in the BAI image
• Significantly dark in the SARE and LAI images

Run the Classification

A Support Vector Machine (SVM) algorithm was used to classify the attribute image. Annotations were added in ENVI to create a map of the classification result. The annotations include a classification legend, grid lines, scale bar, and descriptive text.

After evaluating the performance of the classifier, the reported overall accuracy was 92%.

Figure 16: Map of sugarcane classification.

Figure 17: Confusion matrix.

Once the classifier has been trained on one dataset, it can be applied to a similar dataset from a different date or even a different study area. The second dataset must have the same attributes and data type as the original attribute image. Figure 18 shows an example of running the classifier on a sugarcane image from a different date:

Figure 18: Map of sugarcane classification from June 14, 2016.

Agronomists and farmers can use the classification map as an overview of different development stages, then use the Hotspot Analysis tool for more detailed analysis.

Hotspot Analysis

The ENVI Crop Science platform includes a Hotspot Analysis tool to identify areas in an image that are relatively different than the rest of the image. This tool calculates Getis-Ord Gi* local statistics (Getis and Ord, 1992) to group neighboring pixels of similar values into clusters. The algorithm evaluates each pixel and its surrounding pixels within a specified distance to classify the pixel as:

• Hot (green): Statistically significant clusters of high values
• Low (red): Statistically significant clusters of low values
• Neutral (yellow): Not statistically significant

The resulting hotspot image can help with crop management and anomaly detection; for example, areas affected by weeds, fungus, or water line breaks. Hotspot analysis operates on a single-band raster, most commonly, a vegetation index. In this case, it can be used to look for subtle variations in crop health throughout an area.

Figure 19: Soil Adjusted Red Edge Index images of a group of sugarcane fields during the grand growth period. A 2% linear stretch was applied to both images.

However, Hotspot Analysis reveals differences in sugarcane health over the two-week period:

Figure 20: Hotspot analysis of sugarcane health, two weeks apart, based on Soil Adjusted RedEdge Index.

The “neutral” category represents areas with no spatial clustering. Red areas do not necessarily represent senescent crops; the entire area is undergoing active growth and shows relatively high SARE values. Instead, the red areas indicate crops that are not as green and healthy, relative to the nearby crops.

Conclusion

To accurately assess the health and development of sugarcane requires knowledge of many different variables. Examples include nitrogen content, leaf water content, height, foliage density, canopy structure, and others. Since field studies are only feasible for small study areas, remote sensing can be used as a tool to estimate the health and development of sugarcane crops. Remote sensing also provides the ability to measure these variables at discrete intervals throughout the growing season. This paper presented some examples of using ENVI software to accomplish these objectives.

References

Abdel-Rahman, E. M., and F. B. Ahmed. “The Application of Remote Sensing Techniques to Sugarcane (Saccharum spp. hybrid) Production: A Review of the Literature.” International Journal of Remote Sensing 29, No. 13 (2008): 3753-3767.

Ali, M., C. Montzka, A. Stadler, G. Menz, F. Thonfeld, and H. Vereecken. “Estimation and Validation of RapidEye-Based Time-Series of Leaf Area Index for Winter Wheat in the Rur Catchment (Germany).” Remote Sensing 2015, No. 7 (2015): 2808-2831.

Begue, A., V. Lebourgeois, E. Bappel, P. Todoroff, A. Pellegrino, F. Baillarin, and B. Siegmund. “Spatiotemporal Variability of Sugarcane Fields and Recommendations for Yield Forecast using NDVI.” International Journal of Remote Sensing 31, No. 20 (2010): 5391-5407.

Benvenuti, F., and M. Weill. “Relationship Between Multi-spectral Data and Sugarcane Crop Yield.” Proceedings of the 19th World Congress of Soil Science, Soil Solutions for a Changing World, Brisbane, Australia (2010): 33-36.

Boegh, E., H. Soegaard, N. Broge, C. Hasager, N. Jensen, K. Schelde, and A. Thomsen. “Airborne Multispectral Data for Quantifying Leaf Area Index, Nitrogen Concentration and Photosynthetic Efficiency in Agriculture.” Remote Sensing of Environment 81, No. 2-3 (2002): 179-193.

Canata, T., J. Molin, A. Colaço, R. Trevisan, M. Martello, and P. Fiorio. “Measuring Height of Sugarcane Plants Through LiDAR Technology.” Proceedings of the 13th International Conference on Precision Agriculture (2016): 1-13.

Crist, E.P. “TM Tasseled Cap Equivalent Transformation for Reflectance Factor Data.” Remote Sensing of Environment 17 (1985): 301-306.

Dillewign, C. Botany of Sugarcane. Waltham, MA: Chronica Botanica (1952), 371 pp. Gao, B. “Normalized Difference Water Index for Remote Sensing of Vegetation Liquid Water From Space.” Proceedings of SPIE 2480 (1995): 225-236.

Gers, C. J. “Remotely Sensed Sugarcane Phenological Characteristics at Umfolozi South Africa.” In Proceedings of the IEEE International Geoscience & Remote Sensing Symposium (IGARSS’03), Toulouse, France (2003): 1010-1012.

Getis, A., and J. Ord. “The Analysis of Spatial Association by Use of Distance Statistics.” Geographic Analysis 24, No. 3 (1992): 189-206.

Gunnula, W., M. Kosittrakun, T. L. Righetti, P. Weerathaworn, and M. Prabpan. “Evaluating Sugarcane Growth and Maturity Using Ground-Based Measurements and Remote Sensing Data.” Thai Journal of Agricultural Science 45, No. 1 (2012): 17-28.

Hatfield, J., and J. Prueger. “Value of Using Different Vegetative Indices to Quantify Agricultural Crop Characteristics at Different Growth Stages under Varying Management Practices.” Remote Sensing 2010, No. 2 (2010): 562-578.

Huang, W-D., M-H Hsu, Z-W Yang, J-C Chen, Y-Z Tsai, S-S Chang, and C-M Yang. “Mimicking Satellite Remote Sensing of Chlorophyll Content in Sugarcane (Saccharum officinarum) Leaves.” Crop, Environment & Bioinformatics 2 (2005): 137-147.

Huete, A., K. Didan, T. Miura, E. P. Rodriguez, X. Gao, and L. G. Ferreira. “Overview of the Radiometric and Biophysical Performance of the MODIS Vegetation Indices.” Remote Sensing of Environment 83, No. 1-2 (2002): 195-213.

Huete, A. “A Soil-Adjusted Vegetation Index (SAVI).” Remote Sensing of Environment 25 (1988): 295- 309.

Kross, A., H. McNairn, D. Lapen, M. Sunohara, and C. Champagne. “Assessment of RapidEye Vegetation Indices for Estimation of Leaf Area Index and Biomass in Corn and Soybean Crops.” International Journal of Applied Earth Observation and Geoinformation 34 (2015): 235-248.

Lebourgeois, V., A. Begue, P. Degenne, and E. Bappel. “Improving Harvest and Planting Monitoring for Smallholders with Geospatial Technology: The Reunion Island Experience.” International Sugar Journal 1298 (2010): 109-119.

Li, H., J. Chen, S. Liang, and Q. Li. “Sugarcane Mapping in Tillering Period by Quad-Polarization TerraSAR-X Data.” IEEE Geoscience and Remote Sensing Letters 12, No. 5 (2015): 993-997.

Lin, H., J. Chen, Z. Pei, S. Zhang, and X. Hu. “Monitoring Sugarcane Growth Using ENVISAT ASAR Data.”

IEEE Transactions on Geoscience and Remote Sensing 47, No. 8 (2009): 2572-2579. Meki, M., J. Kiniry, A. Youkhana, S. Crow, R. Ogoshi, M. Nakahata, R. Tirado-Corbalá, R. Anderson, J. Osorio, and J. Jeong. “Two-Year Growth Cycle Sugarcane Crop Parameter Attributes and Their Application in Modeling.” Agronomy Journal 107 (2015): 1310-1320.

Merzlyak, J., A. Gitelson, O. Chivkunova, and V. Y. Rakitin. “Non-destructive Optical Detection of Pigment Changes During Leaf Senescence and Fruit Ripening.” Physiologia Plantarum 106 (1999): 135- 141.

Mulianga, B., B. Agnes, P. Clouvel, and P. Todoroff. “Mapping Cropping Practices of a Sugarcane-Based Cropping System in Kenya Using Remote Sensing.” Remote Sensing 2015, No. 7: 14428-14444.

Rahman, R., A. Hedayutul Islam, and A. Rahman. “NDVI Derived Sugarcane Area Identification and Crop Condition Assessment.” Plan Plus 1, No. 2 (2004): 1-12.

Rondeaux, G., M. Steven, and F. Baret. “Optimization of Soil-Adjusted Vegetation Indices.” Remote Sensing of Environment 55 (1996): 95-107.

Sang, H., J. Zhang, L. Zhai, C. Qiu, and X. Sun. “Analysis of RapidEye Imagery for Agricultural Land Cover and Land Use Mapping.” Third International Workshop on Earth Observation and Remote Sensing Classification (2014): 366-369.

Simões, M. S., J. V. Rocha, and R. A. C. Lamparelli. “Spectral Variables, Growth Analysis and Yield of Sugarcane.” Scientia Agricola 62, No. 3 (2005): 199-207.

Tucker, C. “Red and Photographic Infrared Linear Combinations for Monitoring Vegetation.” Remote Sensing of Environment 8 (1979): 127-150.

Villareal, M., and M. Burce. “Mapping of Sugar Cane Crops Using Light Detection and Ranging (LIDAR) Data.” Research paper, University of San Carlos Talamban Campus (2017): 1-10.

Viña, A., A. Gitelson, D. Rundquist, G. Keydan, B. Leavitt, and J. Schepers. Monitoring Maize (Zea mays L.) Phenology with Remote Sensing. Agronomy Journal 96 (2004): 1139-1147.

Xavier, A., and C. Vettorazzi. “Leaf Area Index of Ground Covers in a Subtropical Watershed.” Scientia Agricola 60, No. 3 (2003): 425-431.

Web resources:

1 Baucum, L., and R. Rice. “An Overview of Florida Sugarcane,” University of Florida, Institute of Food and Agricultural Sciences. Accessed April 2017. https://ufdc.ufl.edu/IR00003414/00001.

2 Sandhu, H., M. Singh, R. Gilbert, J. Shine Jr., R. Rice, and D. Odero. “Maturity Curves and Harvest Schedule Recommendations for CP Sugarcane Varieties.” University of Florida, Institute of Food and Agricultural Sciences. Last modified January 2016. Accessed March 2017. https://edis.ifas.ufl.edu/sc069.

3 “Crop Growth Phases,” Netafim. http://www.sugarcanecrops.com/crop_growth_phases/. Accessed April 2017.

4 “Planet Imagery Product Specifications,” Planet Labs. Last updated December 2023. https://assets.planet.com/docs/Planet_Combined_Imagery_Product_Specs_letter_screen.pdf.

5 “The RapidEye Red Edge Band,” Planet Labs/BlackBridge whitepaper. Publication date unknown. Accessed April 2017. https://www.scribd.com/document/195521834/Red-Edge-White-Paper