Advertisement

Yield Prediction Model for Potato Using Landsat Time Series Images Driven Vegetation Indices

  • Imran Hossain Newton
  • A. F. M Tariqul Islam
  • A. K. M. Saiful IslamEmail author
  • G. M. Tarekul Islam
  • Anika Tahsin
  • Sadmina Razzaque
Original Paper

Abstract

Crop growth monitoring and its yield forecasting is a pre-requisite task to ensure food security and to assess the economic return from agriculture. The application of remote sensing techniques in agricultural sector enhanced the potentiality of crop growth monitoring and estimation of crop yields. Over the last few years, potato cultivation in Bangladesh has promisingly increased. On the contrary, proper distribution capability and the storage capacity is still under the desirable condition due to lack of management system. In this context, remote sensing image analysis techniques were employed in this study to predict potato yield before harvesting to improve management system. This study used 16-day high resolution (~ 30 m) Landsat surface reflectance data for the year of 2010–2011 to detect the maximum normalized difference vegetation index (NDVI) value of a potato growing season. The NDVI will be the maximum after the 63 days of the plantation for Munshiganj District of Bangladesh. Afterward, six satellite images were selected from six growing seasons: 2010–2011, 2011–2012, 2012–2013, 2013–2014, 2014–2015, and 2015–2016 based on the day of the maximum NDVI. A total of 35 potato fields were identified from the filed investigation over the study area to extract the NDVI values from satellite images. Regression analysis was performed between NDVI values and filed level potato yields. The maximum coefficient of correlation (R2) of yield prediction equation was found to be 0.81 between the mean NDVI and potato yield. The equation was validated by using data from 2016 to 2017 growing season, and the result revealed that the difference between predicted and actual filed yield is about 10.4%. It has been found that the high-resolution Landsat images can be an effective means for early estimation of potato yield.

Keywords

Normalized difference vegetation indices Landsat images Potato yield Prediction model Remote sensing 

1 Introduction

Remote sensing technology plays an important role in the agriculture sector by providing a timely and accurate information [3]. It has the ability in the identification of crop classes, crop growth monitoring, and crop yield estimation [16]. On the other hand, potato (Solanum tuberosum L.) is one of the major crops in Bangladesh after rice and wheat, which contributes a significant part in total food supply [24], as the climate and soil condition are very favorable for potato cultivation. During the last few years, potato cultivation in Bangladesh showed an increasing trend. During the 2012–2013, total potato cultivation area and yield were 444,344 ha (0.4 million) and 8,603,000 (8.6 million) metric tons, respectively, while in 2014–2015, these figures rose to 471,054 ha and 9,254,000 (9.2 million) metric tons, respectively [6]. Recently, export of potatoes has increased from a mere volume of US$ 10,930,070 in 2008–2009 to a staggering US$ 32,221,150 in 2013–2014 [7]. These indicate that the future prospect of potato production in Bangladesh is quite bright. Henceforth, potato growth monitoring and its yield prediction have become an important fact of consideration.

Traditional field-based crop data collection becomes clumsy, costly, time-consuming, and have the possibility of error [20]. In addition to, yield computation with conventional methods are no longer beneficial for the decision makers as it takes too much time. During the last few years, many empirical models have been developed to estimate potato yield before the harvesting, but many of them have become unpractical, especially those are depending on filed data collection. As the satellite, remote sensing is one of the best tools to derive essential information about the distribution of crops and its growing conditions over large areas; it can be used for potato growth monitoring and yield estimation. During the last few years, many studies were conducted to develop correlation between crop yield and remote sensing-derived vegetation indices (VIs) for early yield prediction [1, 2, 5, 9, 13, 25]. Most of the studies have been carried out by the correlation of normalized difference vegetation index (NDVI) with yield [9, 14, 19]. In Bangladesh, Nessa [17] used NDVI for rice growth monitoring and production estimation over Bangladesh by using National Oceanic and Atmospheric Administration-Advanced Very High Resolution Radiometer (NOAA-AVHRR) satellite data with a resolution of spectral bands are greater than 1 km. Bala & Islam [4] developed potato yield prediction models by using NDVI, LAI (leaf area index), and fraction of photosynthetically active radiation (fPAR) vegetation indices for Munshiganj District of Bangladesh by using Moderate Resolution Imaging Spectroradiometer (MODIS) (with lowest resolution greater than 250 m) 8-day composite surface reflectance data. Rahman et al. [18] used NOAA-AVHRR data for estimation of potato yield in Bangladesh. However, high-resolution (~ 30 m) satellite images from Landsat is freely available since 1984. The availability of Landsat 8 images provides an ample opportunity for long-term frequent environmental monitoring [15]. Hansen et al. [10] suggested that high spatial resolution for monitoring forest is required for better assessing the rates and spatial extents of the forest change. However, very few studies have been conducted on the relationship between high resolution (~ 30 m) Landsat satellite data and potato yield in Bangladesh. Therefore, this study made an attempt to construct a potato yield prediction model based on NDVI for Munshiganj District (a potato dominant District) of Bangladesh using high-resolution Landsat 5 Thematic Mapper (TM), Landsat 7 Enhanced Thematic Mapper Plus (ETM+), and Landsat 8 Operational Land Imager (OLI) surface reflectance data. The high-resolution Landsat images have been used in this study to develop yield prediction model for the potato crop in Munshiganj District of Bangladesh.

2 Study Area

The study was conducted in Munshiganj District, the main potato growing area of Bangladesh. The District lies between 23° 22′ to 23° 36′ N latitude and 90° 28′ to 90° 36′ E longitude (Fig. 1), with covering an area of 954.96 km2 where 59% area is cultivable land. The study area consists of six administrative units called Upazila: Munshiganj Sadar, Sirajdikhan, Sreenagar, Lohajang, Gazaria, and Tongibari. The climate condition of this area is hot and humid in the summer (May to October) and cool and dry in the winter (November to April). Munshiganj District receives an annual average rainfall of 2376 mm, where 80% rainfall occurs between May and October. The maximum and minimum mean temperatures during the winter varies between 25.82 and 12.1 °C. During summer, the maximum and minimum mean temperatures vary between 34.6 and 25.6 °C. The average monthly minimum and maximum humidity varies from 70 to 84%, where the minimum humidity observed during the dry season and the maximum humidity observed during the summer. The agricultural pattern of this area is characterized by two growing seasons, one is Rabi and the rest one is Kharif. Rabi is the main growing season, which is dominated by potato that starts in late of November or start of December and ends in April. On the other hand, Kharif is dominated by rice, which starts in June and lasts until September. Other food grains which include wheat, maize, pepper, onion, pulses, sugarcane, and mustard are also cultivated in this area. The soil condition of this area is dominated by heavy loam and the heavy clay, while clay occupies second and third position respectively.
Fig. 1

(a) Map of the study area (Munshiganj District) in Bangladesh

3 Materials and Methods

3.1 Satellite Data

In this study, satellite images of Landsat 5 (TM), Landsat 7 (ETM+), and Landsat 8 (OLI) were used for the analysis, those are available in USGS Earth Explorer website (http://earthexplorer.usgs.gov/). Landsat 5 (TM), Landsat 7 (ETM+), and Landsat 8 (OLI) are sun-synchronous satellite staying at an altitude of 705 km above the earth with a 16-day repeat cycle.

Landsat 5 (TM) has seven spectral bands, including one thermal band. Landsat 7 (ETM+) has eight spectral bands, including a pan and thermal band. Landsat 8 has two types of sensors, namely the Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS). The OLI sensor provides nine spectral bands, including a pan band, and TIRS provides two spectral bands. Spectral and technical characteristics of Landsat 5 (TM), Landsat 7 (ETM+), and Landsat 8 (OLI) are presented in Table 1. However, only Red and NIR spectral band were used for the study purposes. Total 14 satellite images of Landsat satellite were downloaded and analyzed. Among them, eight images were selected from 2010 to 2011 growing season and six images were selected from another six growing seasons: 2011–2012, 2012–2013, 2013–2014, 2014–2015, 2015–2016, and 2016–2017. Based on the image availability, Landsat 5 (TM) was used for 2010–2011 growing season, Landsat 7 (ETM+) was used for 2011–2012 and 2012–2013 growing seasons, and Landsat 8 (OLI) was used for 2013–2014, 2014–2015, 2015–2016, and 2016–2017 growing seasons. The dates of image acquisition those used for this study purposes are presented in Table 2. The potato plantation date was considered to be December 2 for each growing season for the entire study area based on the information collected from the field visits. The day of each image was calculated from the starting day of the plantation.
Table 1

Characteristics of Landsat TM, ETM+, and Landsat 8 sensors used in this study

Satellite

Sensors

Band types

Wavelength (μm)

Resolution (m)

Landsat 5

Thematic Mapper (TM)

Band 1 - Blue

0.45–0.52

30

Band 2 - Green

0.52–0.60

30

Band 3 - Red

0.63–0.69

30

Band 4 - Near infrared (NIR)

0.76–0.90

30

Band 5 - Shortwave infrared (SWIR) 1

1.55–1.75

30

Band 6 - Thermal

10.40–12.50

120

Band 7 - Shortwave infrared (SWIR) 2

2.08–2.35

30

Landsat 7

Enhanced Thematic Mapper Plus (ETM+)

Band 1 - Blue

0.45–0.52

30

Band 2 - Green

0.52–0.60

30

Band 3 - Red

0.63–0.69

30

Band 4 - Near infrared (NIR)

0.77–0.90

30

Band 5 - Shortwave infrared (SWIR) 1

1.55–1.75

30

Band 6 - Thermal

10.40–12.50

60

Band 7 - Shortwave infrared (SWIR) 2

2.09–2.35

30

Band 8 - Panchromatic

0.52–.90

15

Landsat 8

Operational Land Imager (OLI)

Band 1 – Coastal aerosol (deep blue)

0.43–0.45

30

Band 2 – Blue

0.45–0.51

30

Band 3 – Green

0.53–0.59

30

Band 4 – Red

0.64–0.67

30

Band 5 – NIR

0.85–0.88

30

Band 6 – Shortwave infrared (SWIR) 1

1.57–1.65

30

Band 7 – Shortwave infrared (SWIR) 2

2.11–2.29

30

Band 8 – Panchromatic

0.50–0.68

15

Band 9 – Cirrus

1.36–1.38

30

Thermal infrared sensor (TIRS)

Band 10 – TIRS1

10.60–11.19

100

Band 11 – TIRS2

11.50–12.51

100

Table 2

List of satellite images used in this study

Model validation using images for the growing season 2010–11

IAD

30/11/10

16/12/10

01/01/11

17/01/11

02/02/11

18/02/11

06/03/11

07/04/11

DAP

− 2

15

31

47

63

79

95

111

Model development using images of the season 2011–12, 2012–13, 2013–14, 2014–15, 2015–16, 2016–17

IAD

28/01/12

30/01/13

25/01/14

28/01/15

15/01/16

02/02/17

 

DAP

58

60

55

58

45

63

 

IAD Image acquisition date; DAP Days after plantation

3.2 Satellite Image Processing

Two techniques were applied to preprocess the satellite images: (1) radiometric calibration and (2) atmospheric correction. Remote sensing data acquired from satellite sensors are influenced by various factors, such as atmospheric scattering and absorption, sensor-target-illumination geometry, sensor calibration, and also by the data processing procedures [23]. Identifying the changes of multi-date data with variability changes is nearly impossible by raw satellite data. In order to identify the actual changes in the landscape, it is necessary to work with an actual surface reflectance of remotely sensed data. For that, radiometric calibration is necessary. Radiometric calibration means a set of correction techniques that are related to correction for the sensitivity of satellite sensor, topography and sun angle, atmospheric scattering, and absorption [12]. The radiometric calibration was done by converting the digital numbers (DNs) to surface reflectance through radiance conversion. Chander and Markham [8] have proposed an equation (Eq. 2) to convert the DNs to spectral radiance.

$$ {L}_{\gamma }={\mathrm{Gain}}_{\upgamma}\ast {\mathrm{DN}}_{\gamma }+{Bias}_{\gamma } $$
(2)
where, Lγ means the spectral radiance at the sensor’s aperture, Gainγ [units of W/ (m2.sr.μm)/DN] and Biasγ [units of W/ (m2.sr.μm)] are band-specific rescaling factors, those are given in the header file of satellite data. The radiance data were converted into surface reflectance by the following equation (Eq. 3).
$$ R=\frac{\pi \ast {L}_{\gamma}\ast {d}^2}{\mathrm{ESU}{\mathrm{N}}_{\upgamma}\ast \mathrm{Cos}\left({\theta}_s\right)} $$
(3)

Where π is a mathematical constant (3.141592654), R is the unitless planetary reflectance, d is the earth-sun distance in astronomical units, ESUNγ is the solar exoatmospheric irradiance [8], and θs is the solar Zenith angle in degrees (90° sun elevation).

Removing the effects of atmospheric factors such as influences of water vapor and aerosol is a critical preprocessing step in image analysis. The actual amount of water vapor and the distribution of aerosols must be known, but measurements of these atmospheric factors are not possible in a manual way, especially in pixel basis. Open source-based Quantum Geographic Information System (QGIS) software 2.18.13 version provides a plugin, and the plugin provides a tool for atmospheric correction, which is known as dark object subtraction (DOS-1) level 1. In this study, this tool was applied in the radiometrically calibrated images to reduce atmospheric scattering effect. DOS-1 searches each pixel of a band to find the darkest value. The scattering is removed by subtracting this value from every pixel in the band. This simple technique is effective for haze correction for multispectral satellite data. Finally, the pan-sharpening process was executed for the calibrated images as the pan-sharpening technique creates high-resolution multispectral data from medium-resolution multispectral data through merging with high-resolution panchromatic data. QGIS 2.18.13 provides this pansharpening process to Landsat data.

3.3 Collection of Yield Data from Crop Fields

A total of 35 potato fields were identified for the growing seasons: 2010–2011, 2011–2012, 2012–2013, 2013–2014, 2014–2015, 2015–2016, and 2016–2017 by interviewing the farmers of different Upazilas (sub-district) of Munshiganj District (Fig. 2). Each growing season starts in December and ends in March, and most of the farmers are poor and the size of the field is small (~ 0.5 hectors). Crop information such as field locations and yield were collected from the Department of Agricultural Extension (DAE) in Munshiganj District.
Fig. 2

Location of selected 35 potato fields (green triangles) over Munshiganj

3.4 Normalized Difference Vegetation Index (NDVI)

Analysis of vegetation and prediction of its yield are related to the identification of crop types and its agronomic variables, such as density, maturity vigor, and disease. Remote sensing can provide these types of information to a great extent. There are different types of vegetation indices (VIs) derived from different spectral reflectance that are generally used to get these types of information. The NDVI is generally used extensively around the world to monitor the vegetation quality, growth, and distribution over a large area. It is a dimensionless index, which is derived from the ratio between the surface reflectance of the NIR and RED bands of the spectrum as follows (Eq. 4) [21].

$$ \mathrm{NDVI}=\frac{\left(\mathrm{NIR}-\mathrm{RED}\right)}{\left(\mathrm{NIR}+\mathrm{RED}\right)} $$
(4)

Healthy plants have a high NDVI value because they are characterized by strong absorption of red energy and strong reflectance of NIR spectrum [22]. NDVI of a crop varies with its growth and health condition with the passage of time [11].

3.5 Potato Yield Prediction Model Using Regression Analysis

NIR and the red band of the calibrated images were selected from each dataset and exported them into QGIS 2.18.13. A simple raster calculation was done by QGIS 2.18.13 using Eq. 1 to find the NDVI images. Finally, the NDVI images were masked using the shapefile of the Munshiganj District. The field points of the location were imported, and the mean NDVI values for each point were extracted from the satellite image considering a 3 × 3 matrix surrounded by each point on the image. The relationship between NDVI and potato growing period was established by plotting the respective values chronologically in terms of days from the start date of potato plantation to harvesting time. The day of the maximum NDVI was selected from the maximum and mean curve and from their relationship with crop yield. To establish this relationship, NDVI data from growing season 2010–2011 were used. Then, a total of six satellite images depending on the date of the maximum NDVI were selected from six growing seasons, namely 2010–2011, 2011–2012, 2012–2013, 2013–2014, 2014–2015, and 2015–2016 to build a relationship between the maximum and mean NDVI and field level potato yield. This relationship based on the maximum and mean NDVI was validated using the satellite image of the 2016–2017 growing season. NDVI values less than 0.25 and more than 0.95 were removed from the listed fields to eliminate the influence of reflectance of other objects like bare soil, settlements, water bodies, non-agricultural crops, and infrastructure.

4 Result and Discussion

4.1 Analysis of the Chronological Changes of NDVI

A total of eight satellite images from 2010 to 2011 growing season were selected chronologically considering the days from the start date of potato plantation to harvesting. Among these images, seven images were selected to extract NDVI values. Values of NDVI on the potato field locations were also changing with time. Spatial distribution of the NDVI values showed an increasing trend until it reached a maximum value. The value of NDVI was found maximum after 63rd days of the plantation, afterward, the NDVI values showed a decreasing trend.

Spatial distribution of the NDVI with the passage of time is presented in Fig. 3. From the each NDVI images, the maximum and mean values of the NDVI were extracted for 35 fields. The chronological changes of maximum and mean values of the NDVI are presented in Fig. 4. The maximum and mean values of the NDVI for potato crop reached on the 63rd days after the plantation.
Fig. 3

Spatial distribution of the NDVI over Munshiganj District a 15th days after plantation, b 31st days after plantation, c 47th days after plantation, d 63rd days after plantation, e 79th days after plantation, f 95th days after plantation, and g 111th days after plantation during the potato growing period of 2010–2011

Fig. 4

Chronological changes of the maximum and mean NDVI extracted from 35 selected fields from Munshiganj district during the potato growing period of 2010–2011

4.2 Regression Analysis of the NDVI Values over the Filed Locations

A total of six satellite images from six growing seasons during the 2010–2016 were selected. Based on available images, those showing the maximum NDVI in each growing season were found 63rd, 58th, 60th, 55th, 58th, and 45th days of the plantation for 2010–2011, 2011–2012, 2012–2013, 2013–2014, 2014–2015, and 2015–2016 growing seasons, respectively. The spatial distribution of the NDVI varies from year to year. During the 2014–2015 growing period, the NDVI distribution was the maximum and during the 2010–2011 growing period, the NDVI distribution was the minimum (Fig. 5). From the 35 field points, NDVI values were extracted and their distribution is presented in Fig. 6. The minimum values of the NDVI are 0.32, 0.40, 0.28, 0.31, 0.40, and 0.42 for 2010–2011, 2011–2012, 2012–2013, 2013–2014, 2014–2015, and 2015–2016 growing season, respectively. The maximum values of the NDVI are 0.64, 0.67, 0.62, 0.88, 0.92, and 0.83 for 2010–2011, 2011–2012, 2012–2013, 2013–2014, 2014–2015, and 2015–2016 growing season, respectively.
Fig. 5

Spatial distribution of the NDVI over the study area for selected images against different growing seasons. a 2010–2011 (63rd days after plantation). b 2011–2012 (58th days after plantation). c 2012–2013 (60th days after plantation). d 2013–2014 (55th days after plantation). e 2014–2015 (58th days after plantation). f 2015–2016 (45th days after plantation)

Fig. 6

NDVI values of 35 selected potato fields for different growing seasons

Regression analysis among filed yield against the maximum and mean NDVI were performed and graphically presented (Fig. 7). The parameters of the regression analysis estimated from yield vs NDVI relationship, together with their correlation coefficients (R2) are presented in Table 3. The relationship between mean NDVI and yield provides higher correlation coefficients (R2 = 0.81) compared to the maximum NDVI vs yield.
Fig. 7

Yield prediction model established from regression analysis between yield data collected from 35 fields and the maximum NDVI (a) mean NDVI (b)

Table 3

Regression parameters and the coefficients of determination (R2)

Regression parameter for maximum value

Regression parameter for mean value

a

b

R 2

a

b

R 2

25.535

8.359

0.35

23.421

11.547

0.81

4.3 Development and Validation of the Yield Prediction Model

The yield prediction model based on the regression analysis was developed based on the yield data collected from 2010 to 2011, 2011–2012, 2012–2013, 2013–2014, 2014–2015, and 2015–2016 potato growing season. To evaluate performance of the model validation is necessary. Based on the deviation from the observed and model prediction, model performance can be determined. Hence, the model has further validated using yield data of 2016–2017 potato growing season. After 63rd days of the plantation, an NDVI image was selected from 2016 to 2017 growing season (Fig. 8) and the mean NDVI value was extracted from the 35 fields. As the coefficient of determination (R2) was found high from the mean value relationship, the validation was done using mean value equation. The regression equation from the mean NDVI is presented below (Eq. 5).
Fig. 8

NDVI distribution over Munshiganj District after the 63rd days of plantation during the 2016–2017 growing season. NDVI analysis was done by Landsat 8 (OLI) data

$$ \mathrm{Yield}=11.547\ast {\mathrm{NDVI}}_{\mathrm{mean}}+23.421 $$
(5)

The actual filed yield was 32.71 ton/ha during the 2016–2017 growing season whereas from the regression equation, it was found 29.31 ton/ha. The deviation was found to be 3.4 ton/ha and the error of yield between predicted and estimated is about 10.4%. According to Bala and Islam [4], the mean error of potato yield prediction equation by using MODIS data was 15%. As the spatial resolution of Landsat is better than MODIS, less error has been found in the yield prediction using Landsat images.

The chronological changes of the NDVI data of potato crop for Munshiganj District are analyzed for the cropping season 2010–2011. A relationship between the NDVI and potato yield from the field has been established. It has been noted that the influences of reflectance by other objects, such as soil, water areas, settlements, and infrastructures should be avoided. The influence of different elements rather than potato filed in the computation of the NDVI was carried out by eliminating and/or rejecting the abnormal values. From the NDVI curves of potato, three phonological stages have been observed: the beginning of the growing period, active growing period, and the decaying period. Hence, the prediction of potato yield for any potato cultivated area is quite possible with higher accuracy before the start of the harvesting period.

5 Conclusions

This study made an attempt to investigate the early yield prediction capacity using remote sensing data for potato crop in the Munshiganj District of Bangladesh which is known for a potato dominating district. Landsat 5 (TM), Landsat 7 (ETM+), and Landsat 8 (OLI) data were used in this research. Chronological changes of NDVI with the passage of time has been established to determine the maximum field NDVI for a growing period (2010–2011) and found that NDVI will be maximum after 63rd days of the plantation. A total of six images from six growing seasons based on the day of the maximum NDVI were selected to develop the yield prediction model. The yield prediction model was established form the regression analysis between the NDVI and potato yield for the Munshiganj District. The maximum correlation coefficients (R2) between the potato yield and mean NDVI was found to be 0.81 and the maximum error is about 10.4%. It was found that NDVI data extracted from Landsat satellite images can be successfully used to predict the potato yield with an appreciable accuracy.

Notes

Acknowledgments

The authors would like to give their sincere thanks to the people of the Department of Agricultural Extension (DAE) at Munshiganj Sadar of Bangladesh for their support, cooperation, and help during the collection of potato field related data for this research work.

References

  1. 1.
    Akhand K, Nizamuddin M, Roytman L, Kogan F (2016) Using remote sensing satellite data and artificial neural network for prediction of potato yield in Bangladesh. In Remote Sensing and Modeling of Ecosystems for Sustainability XIII (Vol. 9975, p. 997508). International Society for Optics and PhotonicsGoogle Scholar
  2. 2.
    Al-Gaadi KA, Hassaballa AA, Tola E, Kayad AG, Madugundu R, Alblewi B et al (2016) Prediction of potato crop yield using precision agriculture techniques. PLoS One 11(9):e0162219CrossRefGoogle Scholar
  3. 3.
    Atzberger C (2013) Advances in remote sensing of agriculture: context description, existing operational monitoring systems and major information needs. Remote Sens 5(2):949–981CrossRefGoogle Scholar
  4. 4.
    Bala SK, Islam AS (2009) Correlation between potato yield and MODIS-derived vegetation indices. Int J Remote Sens 30(10):2491–2507CrossRefGoogle Scholar
  5. 5.
    Baruth B, Royer A, Klisch A, Genovese G (2008) The use of remote sensing within the MARS crop yield monitoring system of the European Commission. Int Arch Photogramm Remote Sens Spat Inf Sci 37:935–940Google Scholar
  6. 6.
    BBS (2016) Yearbook of Agricultural Statistics−2015, 27th ser Statistics and Information Division, Ministry of Planning, Dhaka, BangladeshGoogle Scholar
  7. 7.
    BFTI (2016) Analysing export readiness of the vegetables sector of Bangladesh. Bangladesh Foreign Trade Institute. http://www.bfti.org.bd/index.php/research-publication/concluded-research. Accessed 26 Jan 2017
  8. 8.
    Chander G, Markham B (2003) Revised Landsat-5 TM Radiometrie calibration procedures and Postcalibration dynamic ranges. IEEE trans Geosci. Remote Sens 41(11):2674–2677CrossRefGoogle Scholar
  9. 9.
    Groten SME (1993) NDVI-crop monitoring and early yield assessment of Burkina Faso. Int J Remote Sens 14(8):1495–1515CrossRefGoogle Scholar
  10. 10.
    Hansen MC, Roy DP, Lindquist E, Adusei B, Justice CO, Altstatt A (2008) A method for integrating MODIS and Landsat data for systematic monitoring of forest cover and change in the Congo Basin. Remote Sens Environ 112(5):2495–2513CrossRefGoogle Scholar
  11. 11.
    Islam AS, Bala SK (2008) Assessment of potato phenological characteristics using MODIS-derived NDVI and LAI information. GIScience Remote Sens 45(4):454–470CrossRefGoogle Scholar
  12. 12.
    Kim HH, Elman GC (1990) Normalization of satellite imagery. Int J Remote Sens 11(8):1331–1347CrossRefGoogle Scholar
  13. 13.
    Kouadio L, Newlands NK, Davidson A et al (2014) Assessing the performance of MODIS NDVI and EVI for seasonal crop yield forecasting at the ecodistrict scale. Remote Sens 6(10):10193–10214CrossRefGoogle Scholar
  14. 14.
    Liu WT, Kogan F (2002) Monitoring Brazilian soybean production using NOAA/AVHRR based vegetation condition indices. Int J Remote Sens 23(6):1161–1179CrossRefGoogle Scholar
  15. 15.
    Mandanici E, Bitelli G (2016) Preliminary comparison of Sentinel-2 and Landsat 8 imagery for a combined use. Remote Sens 8(12):1014CrossRefGoogle Scholar
  16. 16.
    Mohd MIS, Ahmad S, Abdullah A (1994) Agriculture application of remote sensing: paddy yield estimation from Landsat-5 thematic mapper data. Internet publication. http://www.gisdevelopment.net/aars/acrs/1994/ts1/ts1003.shtml, published GIS Development 1994. Accessed date 26 Aug 2017
  17. 17.
    Nessa M (2005) Monitoring of rice growth and production in Bangladesh using NOAA Satellite Data. Dissertation, Bangladesh University of Engineering and Technology, Dhaka, BangladeshGoogle Scholar
  18. 18.
    Rahman A, Khan K, Krakauer NY, Roytman L, Kogan F (2012) Using AVHRR-based vegetation health indices for estimation of potato yield in Bangladesh. J Civil Environ Eng 2:111Google Scholar
  19. 19.
    Rasmussen MS (1997) Operational yield forecast using AVHRR NDVI data: reduction of environmental and inter-annual variability. Int J Remote Sens 18(5):1059–1077.  https://doi.org/10.1080/014311697218575 CrossRefGoogle Scholar
  20. 20.
    Reynolds CA, Yitayew M, Slack DC et al (2000) Estimating crop yields and production by integrating the FAO crop specific water balance model with real-time satellite data and ground-based ancillary data. Int J Remote Sens 21(18):3487–3508CrossRefGoogle Scholar
  21. 21.
    Rouse JW, Haas RH, Schell JA, Deering DW (1974) Monitoring vegetation systems in the Great Plains with ERTS. In: Proceedings of the ERTS-1 3rd symposium, vol 1. NASA SP-351. NASA, Washington, pp 309–317Google Scholar
  22. 22.
    Taylor JC, Wood GA, Thomas G (1997) Mapping yield potential with remote sensing. In: Stafford JV (ed) Proceedings of the first European conference on precision agriculture, vol 2. SCI, London, pp 713–772Google Scholar
  23. 23.
    Teillet PM (1986) Image correction for radiometric effects in remote sensing. Int J Remote Sens 7(12):1637–1651CrossRefGoogle Scholar
  24. 24.
    Wennergren EB, Antholt CH, Whitaker MD (1984) Agricultural development in Bangladesh. Westview Press, BoulderGoogle Scholar
  25. 25.
    Zhang P, Anderson B, Tan B, et al (2010) Monitoring crop yield in USA using a satellite-based climate-variability impact index. In: Geoscience and remote sensing symposium (IGARSS), 2010 IEEE international. IEEE, pp 1815–1818Google Scholar

Copyright information

© Springer Nature Switzerland AG 2018

Authors and Affiliations

  1. 1.Institute of Water and Flood Management (IWFM)Bangladesh University of Engineering and Technology (BUET)DhakaBangladesh
  2. 2.Agro-Environmental Remote Sensing and Modeling (ARSaM) Lab, ASICT DivisionBangladesh Agricultural Research Institute (BARI)GazipurBangladesh

Personalised recommendations