• Tidak ada hasil yang ditemukan

Age Estimation of Paddy using Sentinel-2 Imagery: A Case Study in Java Island, Indonesia

N/A
N/A
Protected

Academic year: 2023

Membagikan "Age Estimation of Paddy using Sentinel-2 Imagery: A Case Study in Java Island, Indonesia"

Copied!
11
0
0

Teks penuh

(1)

INTRODUCTION

As the world’s third-largest rice producer and one of the biggest rice consumers, the works of monitoring and predicting the paddy crop phenomena provide a number of benefits for Indonesia. A number of studies have developed a spatial model and remote sensing-based prediction for identifying paddy crop phenomena, detail on nitrogen content (Huang et al., 2013), paddy yield (Guruprasad et al., 2019; Kanke et al., 2016; Palakuru & Yarrakula, 2019a; Palakuru & Yarrakula, 2019b; Shiu &

Chuang, 2019), phenology (dela Torre et al., 2021;

Ding et al., 2020; Dong et al., 2015; Kucuk et al., 2016; Shihua et al., 2014; Xiao et al., 2021), and paddy growth stage (Mulyono et al., 2012, Mulyono et al., 2013) . The input used for the model is remote sensing data ranging from optical (i.e., multispectral, hyperspectral) to radar microwave imagery (i.e., radar radiometer, synthetic aperture radar (SAR)).

Despite the cloudiness issue, the optical multispectral image can provide valuable object

reflectance information and an option for open accesses data. The most common open-access multispectral image used in paddy crop analysis is MODIS due to the high temporal resolution (daily) (Mulyono et al., 2012, Mulyono et al., 2013; Setiawan et al., 2014; Shihua et al., 2014).

However, MODIS data have a low spatial resolution (250 meters). As an alternative, this study tries to use another open-access multispectral image with higher spatial resolution; Sentinel-2A image with 10 meters spatial resolution and 5-7 days temporal resolution. Several studies have tested Sentinel-2 imagery for evaluating, monitoring, and predicting crop phenomena (dela Torre et al., 2021; Inoue et al., 2020; Ni et al., 2021; Wakamori & Ichikawa, 2018; Zhang et al., 2020).

The multispectral-based methods for retrieving rice plant age information have not been widely developed, limited to the availability of rice age information and the need for time series of multispectral remote sensing imagery. The vegetation index derived from multispectral images ARTICLE INFO

Keywords:

GWRJava Island Paddy age Sentinel-2 Vegetation index Article History:

Received: September 8, 2021 Accepted: April 15, 2023

*) Corresponding author:

E-mail: [email protected]

ABSTRACT

Plant age plays a crucial role in determining rice yield. The study on the prediction model of spatially specific rice plant age was still less reported, especially that based on high spatial resolution multispectral data. This study investigates the use of Geographic Weighted Regression (GWR) and extracted vegetation indexes (VI) from the Sentinel-2 multispectral image to build the prediction model based on the time-series dataset from the paddy field observation station.

The GWR result was also compared to the Linear Regression (LR) model to demonstrate the impact of including spatial attribute into the prediction model. Since the majority of paddy field observation stations are situated on Java Island, it served as the research location for this investigation. The results indicate that VI from the Sentinel-2 image shows a significant correlation with the age of the paddy, then the VI was able to use as a predictor to build the paddy age prediction model.

In the statistical evaluation of the model, the coefficient of determination values (R2) reached 0.65, and the RMSE of estimate value was 15 days.

ISSN: 0126-0537

Cite this as: Manessa, M. D. M., Supriatna, & Sidik, I. P.A. (2023). Age estimation of paddy using sentinel-2 imagery: a case study in Java Island, Indonesia. AGRIVITA Journal of Agricultural Science, 45(3), 456-466.http://doi.org/10.17503/

agrivita.v41i0.3106

Age Estimation of Paddy using Sentinel-2 Imagery: A Case Study in Java Island, Indonesia

Masita Dwi Mandini Manessa*), Supriatna, and Iqbal Putut Ash Sidik

Geography Department, Faculty of Mathematics and Natural Science, Universitas Indonesia, West Java, Indonesia

(2)

demonstrates a high association with plant phenology and has been extensively used in previous research.

Nuarsa & Nishiio (2010) first tested the performance of single time Vegetation Index from Landsat images to predict the age of paddy. This study used a linear regression model and showed a perfect model (R2 >

0.75) since there were fewer spatial differences as the study area. The use of timeseries multispectral images for paddy phenology has been tested in several locations, namely Japan (Motohka et al., 2009), China (Ding et al., 2020; Dong et al., 2015;

Shihua et al., 2014; Xiao et al., 2021) and Indonesia (Sukmono et al., 2020). Those previous studies tested the time series data in small area with low variation of data. Then these studies also tested a large dataset of paddy phenology observation of Java Island that was spatially diverse; a LR model could not fit with the prediction model. A GWR is an alternative method to nail the issue of spatial dependency factors (Brunsdon et al., 1996). The GWR technique is commonly used in geographic modeling, as well as in the paddy research. Then, the study’s primary objective was to build a model for predicting paddy age based on the Java Island dataset. By combining the Vegetation Index from Sentinel-2A imagery and the paddy growth information, a prediction model is built based on the GWR method.

MATERIALS AND METHODS

The rice production is centralized on Java Island, southern Indonesia (Fig. 1). This island has six provinces: Banten, West Java, DKI Jakarta, Central Java, DI Yogyakarta, and East Java. Based on 2020 data, half of the nation’s population are inhabitated in this Island (BPS (Indonesia Statistic Agency), 2022). All the provinces except DKI Jakarta count for the high level of rice production.

The data processing and analysis workflow applied in this study are detaily presented in the following procedures.

Sentinel-2A Imagery

At its current state, Sentinel-2 Image is one of the most interesting open-access multispectral images since it has the highest spatial and temporal resolution compared with its predecessor, the Landsat series. The spectral specification consists of 13 spectral channels in the visible/near-infrared (VNIR) and shortwave infrared spectral range (SWIR). However, this study only used the VNIR band with a 10-meter spatial resolution. The Level- 2A surface reflectance product was accessed and processed in the earth engine application (Gorelick et al., 2017).

Paddy Field Monitoring Station

Indonesia Ministry of Agriculture has developed an Integrated Planting Calendar Information System (Sistem Informasi Kalender Tanam Terpadu or SIKATAM) (Kalender Tanam Terpadu Dan Dinamik, 2020). The monitoring system provides the information of initial planting information at each sub-district level, as well as information on the area prones to drought, flood and plant pest attack, information on recommendations for varieties, seeds, fertilizers, and agricultural mechanization that need to be prepared before entering the next planting period. The observed commodities include rice and secondary crops. One of the data available in SIKATAM is the paddy field monitoring station using real-time CCTV (closed- circuit television) instruments installed across Sumatra, Java, and the Bali islands. Still, most of the stations are located on Java Island. The data are open access (http://katam.litbang.pertanian.go.id) and provided as daily, weekly, monthly, and yearly composites. Then, this study collects an everyday image of 33 CCTV stations (Fig. 1) around Java Island from 2016-2019. This data were used as the dependent variable for the prediction model. The summarised of CCTV data are presented in Table 1.

Table 1. Input data used in GWR model

Data type Number of data Description

Raw data (All time observations) 22,748 (100%) 33 CCTV station data during 2016-2019 All time observation data with

available satellite date 15,962 (70%) Preparation, planting, post harvest phases Paddy growing data with

available satellite date 9,673 (42%) Paddy planting to harvest (1-120 days). Vegetative (<50 days); Reproductive (50-80 days); Maturation (> 80 days)

(3)

Fig.1. The study sites on Java Island and distribution CCTV station

(4)

Vegetation Indices (Vl)

The vegetation indices are computed by combining reflection coefficients across several wavebands (Casanova et al., 1998). Six vegetation indices extracted from different forms of algebraic expression ratios among blue, green, red, and near-infrared bands: NDVI (Normalized Difference Vegetation Index) proposed by Rouse et al. (1974), GNDVI (Green Normalized Difference Vegetation Index) presented by Gitelson et al. (1996), SAVI (Soil Adjusted Vegetation Index) introduced by Huete (1988), EVI (Enhanced Vegetation Index) presented by Huete et al. (1999), MTVI2 (Modified Triangular Vegetation Index) by Haboudane et al. (2004), SAVI (Soil Adjusted Vegetation Index) proposed by Rouse et al. (1974), GNDI by Sripada et al. (2006).

The corresponding equations are presented as follows:

Where: ρis surface reflectance for band blue, green, red, and Near InfraRed (NIR); L is coefficient for SAVI (this study uses 0.5).

Linear Regression (LR)

This work used linear regression (LR) to design a model for estimating the age of paddy based on in-situ data and VI create from Sentinel-2.

LR investigated the linear connection between

response and predictor factors as represented by the following equation:

Where: i denotes the site where the parameter estimates are acquired locally and where the data on y and x are measured, X is the VI as the explanatory variable, β is the intercept and coefficient for each explanatory variable, j is the index of VI and last is error.

To highlight the number of predictor variables, LR models are constructed using a single predictor variable or multiple predictor variables. A Simple Linear Regression (SLR) was denoted when a single predictor variable was used and Multiple Linear Regression (MLR) when two or more predictor variables were used. R software was used for all data processing and statistical modeling.

Geographic Weight Regression (GWR)

Geographic Weight Regression extends the classic regression model by allowing local variation in rates of change such that the coefficients in the model are location-specific rather than global estimates (Brunsdon et al., 1996). Since global models offer a mean estimate of the coefficient for each predictor variable throughout the research region, they implicitly presume the relationships to be spatially constant. However, this is not always the case. This assumption is relaxed by GWR, which yields a unique coefficient estimate for each location and covariable. The GWR model is an effective ensemble ensemble of LR calibrated at each point of the study region, as represented by (ui,vi) in this equation.

Where: i denotes the site where the parameter estimates are acquired locally and where the data on y and x are measured; β = value of coefficient at location, X = observed independent variable, location variable, ɛ = error. In this model, these parameter estimates are now local to location i instead of global contents.

The estimator for the parameters is then:

Where: W(i) is a matrix of weights specific to location i such that observations nearer to i are

...(1)

...(2)

...(3)

...(4)

...(5)

...(6)

...(7)

...(8)

...(9)

...(10)

...(11)

...(12)

...(13)

...(14)

(5)

given greater than observations further away, β = estimator parameter, X = independent variable, Y = dependent variable.

The matrix W(i) has the form:

Where: win is the weight given to data point n for the estimate of the local parameters at location i.

This study used the R package of spgwr:

Geographically Weighted Regression to predict the paddy age (Bivand et al., 2022). Moreover, both LR and GWR algorithm was evaluated using R2 and RMSE. The quality of model fit was assessed using the R squared (R2), while the predictive performance of the models was evaluated using root-mean- square error (RMSE).

RESULTS AND DISCUSSION

The Relation Between the Age of Paddy and Vegetation Index

Fig. 2 displays the Pearson Product Moment correlation coefficients between paddy age and the vegetation index. VARI had the lowest connection (0.10) with paddy age among the eleven vegetation indices. While the remaining VI shows no significant association (0.26-0.34). In contrast to previous research, Nuarsa & Nishio (2010) found that the correlation between VI and paddy age might range from 0.7 to 0.85. The poor relation of VI in this study might happen due to the use of a big dataset of paddy observation while the past study only used a specific location with low variety of spatial and temporal attributes.

Two pairs of VI showed high similarity (1.00), namely NDVI-IPVI and MASVI-SAVI. This finding indicates that both data display the same values and that inputting both data will not affect the model. Then, selecting one of them will provide the same consequence as entering both of VI. In the examination of the prediction model, neither IPVI

nor MASVI is considered as an input. Moreover, this statistical result that is showed almost similar correlation between the age of paddy (Biologische Bundesanstalt, Bundessortenamt und Chemische Industrie-BBCH) also implies that a single vegetation index could not be a predictor of the age of the paddy model due to less correlation. Thus, multiple predictions will increase the performance of the model predictor (MLR).

Linear Regression Model

The performance of VI in estimating the age of paddy using a LR model is presented in Table 2. Since this research utilizes a big dataset of paddy observation, the data variance are likely to be substantial, as a result of the the use of various paddy varieties, planting methods, and noises from 33 paddy station across Java Island. These various parameters in a data set results in a poor estimation of the paddy’s age, as viewed from the best model which is only able to predict with an R2 less than 1.38.

In addition, this study highlights the necessity for many variables to define the complex connections between the vegetation index and paddy age.

MLR presupposes that there is no meaningful relationship between any of the independent variables. Moreover, it requires that there is a relationship between each independent variable and the one dependent variable. By assigning a distinct regression coefficient to each independent variable, each of these correlations is weighted to guarantee that the value of the dependent variable is determined by the most significant independent factors.

MLR is a computation that is more specific than basic SLR. For straightforward connections, such as those modeled in a homogeneous paddy field, the basic SLR may capture the link between the two variables with relative ease. For a paddy field as complex and varied as the locations of this research, MLR is preferable. Nonetheless, if the impact of variation in the dataset is substantial, MLR will struggle to make accurate predictions.

...(15)

(6)

Fig. 2. Pearson Product Moment correlation between paddy age and vegetation index

(7)

Geographic Weighted Regression (GWR) Model The LR illustrates its failure to predict the paddy age over a large regional variation employed in this study. This research suggests the use of GWR to determine the age of plant. The geographical information of each data point is entered in order to solve the problem of data variation caused by the number of observation stations throughout Java Island. The association between the vegetation indicators and the age of the paddy was modeled using a GWR.

The descriptive statistics of the regression parameters are presented in Table 3. The parameters of various variables in the GWR model exhibited distinct geographical disparities. The DVI

is the predictor with the strongest influence on the model, shown by the highest coefficient value since the change of leaf color is related to the age of the paddy. Rice paddies covered with water and small paddy stems in the germination stage appear brown- light green in color; the light green of the paddy leaf indicates the vegetative phase; conditions without water and lush paddy leaf appear in full green color indicate the reproductive phase; and the ripening phase is indicated by the growing grains’ yellow to brown color. Fig. 3 displays the model fit; the R2 ranges from 0.52 to 0.56 and the RMSE is between 20 days. Significantly, the regression produced the following linear function at the 0.005 level. The summary of GWR was presented in Fig. 3.

Table 2. Performance of combination of VI to predict the paddy age using LR model

Rank N VI predictor rsquare adjr predrsq aic

1 9 DVI EVI NDVI GDVI GNDVI MTVI2 SAVI SR VARI 0.138 0.137 0.137 215194 2 8 DVI EVI GDVI GNDVI MTVI2 SAVI SR VARI 0.138 0.137 0.137 215195 3 7 DVI EVI NDVI GDVI GNDVI SAVI VARI 0.137 0.137 0.137 215195

4 6 DVI EVI GDVI GNDVI SAVI VARI 0.137 0.137 0.137 215196

5 7 DVI EVI GDVI GNDVI SAVI SR VARI 0.137 0.137 0.137 215197

507 2 EVI SR 0.098 0.098 0.098 216207

508 2 NDVI SR 0.083 0.083 0.083 216575

509 1 NDVI 0.082 0.082 0.082 216612

510 1 SR 0.068 0.068 0.068 216946

511 1 VARI 0.010 0.010 0.009 218343

Table 3. Descriptive statistics of the local regression parameters of the GWR model

Coefficient Minimum 1st Quadran Median 3rd Quadran Maximum Global

X.Intercept. 6.02E+03 6.02E+03 6.02E+03 6.02E+03 6.02E+03 6.02E+03

DVI 5.71E+03 5.71E+03 5.71E+03 5.71E+03 5.71E+03 5.71E+03

EVI 7.20E+01 7.20E+01 7.20E+01 7.20E+01 7.20E+01 7.20E+01

NDVI -9.60E+01 -9.60E+01 -9.60E+01 -9.60E+01 -9.60E+01 -9.60E+01 GDVI -9.51E+03 -9.51E+03 -9.51E+03 -9.51E+03 -9.51E+03 -9.51E+03

GNDVI 1.40E+03 1.40E+03 1.40E+03 1.40E+03 1.40E+03 1.40E+03

MTVI2 -1.40E+04 -1.40E+04 -1.40E+04 -1.40E+04 -1.40E+04 -1.40E+04 SAVI -3.96E+03 -3.96E+03 -3.96E+03 -3.96E+03 -3.96E+03 -3.96E+03 SR -1.57E+02 -1.57E+02 -1.57E+02 -1.57E+02 -1.57E+02 -1.57E+02

VARI 1.09E+02 1.09E+02 1.09E+02 1.09E+02 1.09E+02 1.09E+02

(8)

According to the analyses of RMSE and R2 shown in Fig. 3, the age of the paddy was not effectively predicted, especially for a period of generative (<45 days). The prediction accuracy was depended on the correlation of both inputs because the model was based on Sentinel-2 vegetation indices pixels and observed age of the paddy.

Unlike the previous study (Nuarsa & Nishio, 2010), the vegetation indices did not wholly agree with the observed age of paddy (poorer with an increase of in age). Nuarsa & Nishiio (2010) and Sukmono et al. (2020) demonstrated a strong association between vegetation indicators and the observed

time of paddy phenology; however, this study analysis used only 10-30 observations. The present study used 9,673 observations of paddy phenology gathered over the time period of three years at 33 distinct sites. Furthermore, the contribution of image noise (atmospheric noise effect) still severely impacted the Sentinel-2 vegetation indices, which contributes to the poor performance of this research in predicting the paddy ages. The discrepancy between the Sentinel-2 vegetation indices and the actual beginning age of the paddy explains the forecast mistakes.

Fig. 3. Scatter plots of Sentinel-2A estimated age of paddy versus in situ rice growth observation during 2016-2019 in different phenology phase using the GWR

(9)

CONCLUSION

Information on rice’s phenology and growing phase are critical for paddy phenomenon investigations. However, accurate information on paddy age with sufficient precision is not yet accessible at the regional level. This research established the ability of vegetation indices obtained from multitemporal Sentinel-2 optical multispectral data to forecast the age of paddy. Sentinel-2’s multi- temporal vegetation indices were sensitive enough to detect changes in paddy growth patterns. The prediction model might be used to map the paddy age distribution. This knowledge is essential for agricultural policymakers. Due to the necessity for a daily noise-free imagery of Sentinel-2 imagery, the current research does not depict the geographical distribution of the age of paddy across Java Island.

This research assessed a number of the most effective techniques for rectifying the huge imaging database of Java Island. In addition, evaluating the non-linear regression model on a resilient variant of the dataset may further improve prediction accuracy. The investigation for a solution to this issue is already begun.

ACKNOWLEDGEMENT

This research was funded by Universitas Indonesia under research grant PUTI Q3 2020 with grant contact number NKB-4492/UN2.RST/

HKP.05.00/2020.

REFERENCES

Bivand, R., Yu, D., Nakaya, T., & Garcia-Lopez, M-A.

(2022). Geographically Weighted Regression [R package spgwr version 0.6-35]. https://CRAN.R- project.org/package=spgwr

BPS (Indonesia Statistic Agency). (2022). Indonesia da- lam angka 2022. https://www.bps.go.id/publica- tion/2022/02/25/0a2afea4fab72a5d052cb315/

statistik-indonesia-2022.html

Brunsdon, C., Fotheringham, A. S., & Charlton, M. E.

(1996). Geographically Weighted Regression:

A method for exploring spatial nonstationarity.

Geographical Analysis, 28(4), 281–298. https://

doi.org/10.1111/J.1538-4632.1996.TB00936.X Casanova, D., Epema, G. F., & Goudriaan, J. (1998).

Monitoring rice reflectance at field level for estimating biomass and LAI. Field Crops Research, 55(1–2), 83–92. https://doi.

org/10.1016/S0378-4290(97)00064-6

dela Torre, D. M. G., Gao, J., Macinnis-Ng, C., & Shi, Y.

(2021). Phenology-based delineation of irrigated and rain-fed paddy fields with Sentinel-2 imagery in Google Earth Engine. Geo-Spatial Information Science, 24(4), 695–710. https://doi.org/10.1080 /10095020.2021.1984183

Ding, M., Guan, Q., Li, L., Zhang, H., Liu, C., & Zhang, L. (2020). Phenology-based rice paddy mapping using multi-source satellite imagery and a fusion algorithm applied to the Poyang Lake plain, Southern China. Remote Sensing, 12(6), 1022.

https://doi.org/10.3390/rs12061022

Dong, J., Xiao, X., Kou, W., Qin, Y., Zhang, G., Li, L., Jin, C., Zhou, Y., Wang, J., Biradar, C., Liu, J., & Moore, B. (2015). Tracking the dynamics of paddy rice planting area in 1986–2010 through time series Landsat images and phenology-based algorithms. Remote Sensing of Environment, 160, 99–113. https://doi.

org/10.1016/j.rse.2015.01.004

Gitelson, A. A., Merzlyak, M. N., & Lichtenthaler, H. K.

(1996). Detection of red edge position and chlorophyll content by reflectance measurements near 700 nm. Journal of Plant Physiology, 148(3–4), 501–508. https://doi.org/10.1016/

S0176-1617(96)80285-9

Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., & Moore, R. (2017). Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sensing of Environment, 202, 18–27. https://doi.org/10.1016/j.rse.2017.06.031 Guruprasad, R. B., Saurav, K., & Randhawa, S. (2019).

Machine Learning Methodologies for Paddy Yield Estimation in India: a Case Study. IGARSS 2019 - 2019 IEEE International Geoscience and Remote Sensing Symposium, 7254–7257.

https://doi.org/10.1109/IGARSS.2019.8900339 Haboudane, D., Miller, J. R., Pattey, E., Zarco-Tejada,

P. J., & Strachan, I. B. (2004). Hyperspectral vegetation indices and novel algorithms for predicting green LAI of crop canopies: Modeling and validation in the context of precision agriculture. Remote Sensing of Environment, 90(3), 337–352. https://doi.org/10.1016/j.

rse.2003.12.013

Huang, C.-W., Huang, C.-C., Yang, C.-K., Wu, T.-H., Tsai, Y.-Z., & Miao, P.-L. (2003). Determination of Nitrogen Content in Rice Crop Using Multi- Spectral Imaging. 2003 ASAE Annual Meeting, 031132. https://doi.org/10.13031/2013.13741 Huete, A. R. (1988). A soil-adjusted vegetation index

(SAVI). Remote Sensing of Environment,

(10)

25(3), 295–309. https://doi.org/10.1016/0034- 4257(88)90106-X

Huete, A. R., Didan, K., Huete, A., Didan, K., Leeuwen, W. Van, Jacobson, A., Solanos, R., & Laing, T. (1999). Modis Vegetation Index (MOD 13) Algorithm Theoretical Basis Document Principal Investigators Development Team MODIS Product ID: MOD13. https://modis.gsfc.nasa.

gov/data/atbd/atbd_mod13.pdf

Inoue, S., Ito, A., & Yonezawa, C. (2020). Mapping Paddy Fields in Japan by Using a Sentinel-1 SAR Time Series Supplemented by Sentinel-2 Images on Google Earth Engine. Remote Sensing, 12(10), 1622. https://doi.org/10.3390/rs12101622 Kalender Tanam Terpadu dan Dinamik. (2020). Retrieved

18 October 2022, from http://katam.litbang.

pertanian.go.id/main.aspx

Kanke, Y., Tubaña, B., Dalen, M., & Harrell, D. (2016).

Evaluation of red and red-edge reflectance- based vegetation indices for rice biomass and grain yield prediction models in paddy fields.

Precision Agriculture, 17(5), 507–530. https://

doi.org/10.1007/s11119-016-9433-1

Kucuk, C., Taskin, G., & Erten, E. (2016). Paddy-Rice Phenology Classification Based on Machine- Learning Methods Using Multitemporal Co-Polar X-Band SAR Images. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 9(6), 2509–2519. https://doi.

org/10.1109/JSTARS.2016.2547843

Motohka, T., Nasahara, K. N., Miyata, A., Mano, M.,

& Tsuchida, S. (2009). Evaluation of optical satellite remote sensing for rice paddy phenology in monsoon Asia using a continuous in situ dataset. International Journal of Remote Sensing, 30(17), 4343–4357. https://doi.

org/10.1080/01431160802549369

Mulyono, S., Fanany, M. I., & Basaruddin, T. (2012). A paddy growth stages classification using MODIS remote sensing images with balanced branches support vector machines. 2012 International Conference on Advanced Computer Science and Information Systems, ICACSIS 2012 - Proceedings, 203–206. https://ieeexplore.ieee.

org/document/6468764

Mulyono, S., Pianto, T. A., Fanany, M. I., & Basaruddin, T. (2013). An ensemble incremental approach of Extreme Learning Machine (ELM) for paddy growth stages classification using MODIS remote sensing images. 2013 International Conference on Advanced Computer Science and Information

Systems, ICACSIS 2013, 309–314. https://doi.

org/10.1109/ICACSIS.2013.6761594

Ni, R., Tian, J., Li, X., Yin, D., Li, J., Gong, H., Zhang, J., Zhu, L., & Wu, D. (2021). An enhanced pixel- based phenological feature for accurate paddy rice mapping with Sentinel-2 imagery in Google Earth Engine. ISPRS Journal of Photogrammetry and Remote Sensing, 178, 282–296. https://doi.

org/10.1016/J.ISPRSJPRS.2021.06.018

Nuarsa, I. W., & Nishio, F. (2010). Relationships between rice growth parameters and remote sensing data. International Journal of Remote Sensing and Earth Sciences (IJReSES), 4(1). https://doi.

org/10.30536/j.ijreses.2007.v4.a1221

Palakuru, M., & Yarrakula, K. (2019a). Study on paddy phenomics eco-system and yield estimation us- ing multi-temporal remote sensing approach. In- dian Journal of Ecology, 46(2), 293–297. https://

www.indianjournals.com/ijor.aspx?target=ijor:i- je1&volume=46&issue=2&article=011

Palakuru, M., & Yarrakula, K. (2019b). Study on paddy phenomics ecosystem and yield estimation using space-borne multi sensor remote sensing data. Journal of Agrometeorology, 21(2), 171–

175. https://www.scopus.com/inward/record.

uri?eid=2-s2.0-85074379509&partnerID=40&m d5=5dfab58963139835c3b148b25306f11a Rouse, J. W., Jr., Haas, R. H., Deering, D. W., Schell, J.

A., & Harlan, J. C. (1974). Monitoring the Vernal Advancement and Retrogradation (Green Wave Effect) of Natural Vegetation (Contractor Report (CR) E75-10354 NASA-CR-144661 RSC-1978–

4; pp. 1–390). Texas A&M University, Remote Sensing Center, College Station, Texas. https://

ntrs.nasa.gov/citations/19750020419

Setiawan, Y., Rustiadi, E., Yoshino, K., Liyantono, &

Effendi, H. (2014). Assessing the Seasonal Dynamics of the Java’s Paddy Field Using MODIS Satellite Images. ISPRS International Journal of Geo-Information, 3(1), 110–129.

https://doi.org/10.3390/ijgi3010110

Shihua, L., Jingtao, X., Ping, N., Jing, Z., Hongshu, W.,

& Jingxian, W. (2014). Monitoring paddy rice phenology using time series MODIS data over Jiangxi Province, China. International Journal of Agricultural and Biological Engineering, 7(6), 28–36. https://doi.org/10.25165/IJABE.

V7I6.1433

Shiu, Y. S., & Chuang, Y. C. (2019). Yield estimation of paddy rice based on satellite imagery:

Comparison of global and local regression

(11)

models. Remote Sensing, 11(2). https://doi.

org/10.3390/rs11020111

Sripada, R. P., Heiniger, R. W., White, J. G., & Meijer, A. D. (2006). Aerial Color Infrared Photography for Determining Early In-Season Nitrogen Requirements in Corn. Agronomy Journal, 98(4), 968–977. https://doi.org/10.2134/

AGRONJ2005.0200

Sukmono, A., Nugraha, A. L., Ariwahid, A. N., & Shabrina, N. (2020). Growth models and age estimation of rice using multitemporal vegetation index on landsat 8 imagery. Advances in Science, Technology and Engineering Systems, 5(5), 506–511. https://doi.org/10.25046/AJ050563 Wakamori, K., & Ichikawa, D. (2018). The Combined

Use of Sentinel-1, Sentinel-2 and Landsat 7&8

Data for Estimating Heading Date of Paddy Rice. IGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium, 2018-July, 7715–7718. https://doi.org/10.1109/

IGARSS.2018.8518709

Xiao, W., Xu, S., & He, T. (2021). Mapping Paddy Rice with Sentinel-1/2 and Phenology-, Object-Based Algorithm—A Implementation in Hangjiahu Plain in China Using GEE Platform. Remote Sensing, 13(5), 990. https://doi.org/10.3390/rs13050990 Zhang, W., Liu, H., Wu, W., Zhan, L., & Wei, J. (2020).

Mapping Rice Paddy Based on Machine Learning with Sentinel-2 Multi-Temporal Data:

Model Comparison and Transferability. Remote Sensing, 12(10), 1620. https://doi.org/10.3390/

rs12101620

Referensi

Dokumen terkait

Using Sentinel-2 multispectral images to map the occurrence of the cossid moth Coryphodema tristis in Eucalyptus nitens plantations of Mpumalanga, South Africa..