Detail Extended Abstract

Monitoring rice growth and predicting rice biomass by Sentinel-2 data at Giao Thuy, Nam Dinh

MeetingVIWC2026
TopicOS3: Water Environment and Mining
AuthorHoa Phan Thi Mai
OrganizationHa Noi University of Mining and Geology, North Tu Liem District, Ha Noi, Vietnam
ID2022.1670559098

Content

  1. Introduction
Agricultural crops have a causal relationship with climate change. Although production is directly affected by high temperatures, this activity can affect the GHG concentration in the atmosphere, causing some changes in climate. One key solution to mitigate climate change and boost farming in the country’s coping with Climate Change is to conserve carbon stocks.  Rice is a viable option for increasing carbon Stocks in agricultural systems in Vietnam. There are two types of remote sensing techniques: one uses optical sensors and the other uses (SAR) sensors. For example, the general time-series analysis of the vegetative and water-derived indices of these sensors can be used, such as the Vegetation Index (NDVI), the Enhanced Vegetation Index (EVI). ) and Water Index (NDWI), which can be used to estimate rice parameters (Xiao, X., 2006; Xiao, X., 2005), monitor growths of rice in rice fields (Nuarsa, I.W, 2007). Contrary to optical sensors, SAR data can work in these conditions, so it is more effective for monitoring rice fields than optical sensors. RADARSAT-1/2 were used to map and monitor rice (Li, K., 2012). Sentinel-2 data (level-1C) have been corrected for atmospheric effects and geometric distortions. In Vietnam, Pham Quoc Trung et al (2018) applied remote sensing and satellite imagery to determine the carbon stock of perennial trees in Bo Trach district, Quang Binh provinc. Tran Thi Hien et al (2013) based on image characteristics (NDVI value) that are related to changes in growth status of rice plants in space and time to help determine sowing time, and crop structure in rice growing regions. Nguyen Thanh Son et al. (2014) used MODIS image data from December 2000 to December 2012 with the EVI index to classify and manage rice along the Mekong River, Vietnam region. Therefore, this study has adopted remote sensing technology using Sentinel-2(level-1C) imagery for calculating NDVI value through a field survey; rice growth stage monitoring in order to create models for predicting rice biomass
  1. Methodology
2.1. Material The study site is located in Giao Thuy, Nam Dinh, a coastal district (lat. 20110 –180 N; long. 106220 –320 E) at the mouth of Red River in Nam Dinh Province. Among the four coastal provinces (Nam Dinh, Thai Binh, Ninh Binh, and Hai Phong) of the Red River Delta (Fig. 1), Nam Dinh has the longest coastline, and hence considered as possibly most vulnerable to salinity intrusion. Rice production is a main agricultural activity in the district. There are two crops of rice in a year, the spring rice (vu Chiem in local Vietnamese language) from January to June and the summer rice (vu Mua) from July to November. Figure 1. Study site of Giao Thuy district, Nam Dinh Province, Vietnam (blue circles indicating the locations of 15 points in the field) 2.2. Satellite data processing The remote sensing image used in the study is a Sentinel-2 data (level-1C), acquired on table 1 and corrected for atmospheric effects and geometric distortions. The data are tiled and formatted by using UTM/WGS84 reference frames. To obtain the pixel values associated with biomass, the NDVI equation was used (Rouse et al. 1973) (1) Where NDVI: Normalized Difference Vegetation Index NIR : Near Infrared Band (band 8) Red : Red Band (band 4) The purpose of NDVI image is to convert multi-spectral data into a multichannel image that shows the distribution of plants. 2.3. Field data collection To find the growing rice stage with the highest NDVI value, the study calculated the NDVI value at 6 development stages of rice (table.1). The simulated NDVI/LAI and actual NDVI/LAI values in the heading and flowering stage are less different than in others (Huyen, P., 2014). The sample plots are collected at the growth rice stage for the highest NDVI value at 2 regions, and then measured biomass in the field. The study area has 2 regions: rice cultivation area and the mixed-cropping with rice-aqua co-culture (MCAC). In the study, GPS Etrex10 is used to gather 20 sample plots. And surveys in the field were carried out with Sentinel-2 data together to calculate the rice biomass stock before the harvest stage (the heading and flowering stage – April, May 2020), and then build the correlation function between NDVI and AGB. Table 1. Sentinel-2 image selected date in this study
No Scene ID Creation date Growth phase
1 20200131T032917 31/1/2020 Seedling
2 20200225T032326 25/02/2020 Tillering
3 20200312T033026 12/3/2020 booting
4 20200421T033212 21/4/2020 Heading
5 20200520T031539 20/5/2020 Flowering
6 20200625T032832 25/6/2020 Harvesting
2.4. Data analysis The correlations were calculated by the linear regression equation; y = a + bx. the values of a and b can be calculated using the following formula (Walpole. 1992): (2)   (3) Where x = NDVI value y = AGB value And then, the correlation analysis is calculated using the following Pearson correlation test equation: (4) In which: Yi and Y ̅_i are the estimated variables and their average value Xi and X ̅_i are the measurement variables and their mean n is the number of samples in the data set. If r2 = 1 or r2 = –1, the relationship of x and y is determined; that is, for any value of x we can determine the value of y. If r = 0, the variables x and y are completely independent and are not related. The r-value is classified as follows: 0.1 ≤ r2 <0.3  have little if any (linear) correlation, 0.3 ≤ r2 <0.5 indicate variables which have a low correlation, 0.5 ≤r2 indicate variables which can be considered highly correlated.
  1. Results
3.1. Relationship between vegetative indices NDVI and growth of rice plants The variation in NDVI value of the rice planted area at the growth stages is very important. It is the basis for monitoring the crop, estimating the rice growth period, the post-harvest period, and estimating the maximum biomass value during the growing rice. There are 15 NDVI values at rice cultivation areas and MCAC in 6 stages presented in Figure 3. The NDVI results were evaluated based on the value curve at 15 sample plots for 6 growing periods of rice plants in Figure 2. The results showed that NDVI value started to increase gradually when the rice appeared leaves, reaching maximum at the time before heading and flowering, then decreasing gradually when harvesting. After this period, NDVI also began to decline. Rice with a short growth time is shorter than that of rice with a longer growth period. This affects the growth rate of NDVI. During the sowing period, the field was flooded with water so the NDVI value ran from -0.2 to -0.1, In the growth period, the rice plant develops through three stages including the growth phase, the reproductive stage and the main rice stage, therefore, the value of NDVI will reach the highest value in both the growth process of plants with NDVI in the range from 0.5 and 0.75. In the post-harvest period, the field is bare and arid, the value of NDVI decreases gradually, between 0.05 – 0.15. Through the survey results and statistics, the chart showed the fluctuation of NDVI index over time in Figure 3. Simultaneously compared with the growth of rice, the variable of NDVI value is suitable for the rice growing season. NDVI index of vu Chiem in Giao Thuy district, Nam Dinh province in 2020 changes from 0.2 to 0.75, following the rule of low from the beginning of the season, increasing gradually and peaking at the time when the rice plant is well developed in the post-tillering period. Then it decreases gradually in the ripening stage, and decreases to the lowest level at the end of the season (in the post-harvesting stage). Figure 2. The development of rice in the winter-spring season through NDVI index   The high temporal access to NDVI data provides more details regarding the agriculture season. 3.2. Correlation between AGB and NDVI Biomass is relevant with rice production and normally biomass classified into wet and dry biomass[1] The relationship between rice biomass and NDVI has shown the highest determination coefficient (R2 = 0.885) with the equation Y = 346.45X - 156.98 (Eq.5),  where y and x are the rice biomass (rice AGB) and NDVI, respectively in figure 4. Figure 4. Correlation graph between NDVI and rice biomass (AGB) Retest reliability of the biomass value for five other plots is based on Eq.5 and the field data measurement. Quantitative comparison of rice plants between analysis results and reference data showed a linear relationship with R² = 0.6685, where y was the rice biomass of reference data, and x is the rice biomass of the analysis results of the Sentinel-2. The standard error of this estimation is 30.2 ha. The ratio of the sampling error to the measurement error is about 23.69% (Tab. 2); or the biomass estimation on Sentinel-2 images can give an accuracy up to 76.3%. Sentinel-2 has good capabilities to monitor rice plants and map rice biomass. Figure 3 also shows the analysis and reference data ​​scattered in ascending direction and centered near the Linear. Linear line showed the rice biomass of reference data and the rice biomass of the analysis results ​​with concentration and upward trend in the rise of NDVI. However, there are still simulated and actual AGB points that do not match. These simulated values ​​are largely higher than the measured values. Simulated results depend on the growing time of rice and weather data. The Winter-Spring (vu Chiem) crop in 2020 falls in the dry season of Giao Thuy\'s year. This is also a cause of higher biomass simulated results. With the same weather data, the simulation results of NDVI, AGB at different points are different. This difference is due to different growing conditions in terms of cultivation techniques, soil conditions, and optimum water–fertilizer management, and the time of sowing. Due to the different planting date, the influence of weather on the growth and development of rice at these points is also different. (a) (b) Figure 5. Biomass map of Giao Thuy, Nam Dinh on April (a) and May (b), 2020  
Figure 3. NDVI spatial distribution in each period for 6 times Table 2. Detailed statistics of multiple regression between observed biomass and predicted biomass in Giao thuy
No UTM X UTM Y Observed Biomass (t/ha) Predicted Biomass (t/ha) Value difference Percentage (%)
1 665474 2237743 68.29 100.1 -31.81 68.29
2 664915 2237937 28.75 32.77 -4.02 28.75
3 667420 2246300 35.67 42.99 -7.32 35.67
4 654845 2243523 44.10 23.34 20.76 2.58
5 659631 2244795 25.67 39.57 -13.90 25.67
Average 47.75    
Min 23.34    
Max 100.1    
Std.e 30.20   23.69
  Conclusions Modeling the Relationship between rice field AGB and (Sentinel-2) is a method of determining rice biomass without destroying the crop. AGB values predicted by remote sensing images are correlated with NDVI. It is possible to monitor the changes of NDVI each day during the life cycle of a rice plant. This result can be used in the prediction of rice biomass. The NDVI value in the area of winter – spring rice cultivation in Giao Thuy was from 0.41 to 0.78. Giao Chau and Giao Thanh communes have the potential to exploit rice and the  agricultural by-products with the highest biomass. Biomass values of Giao Chau and Giao Thanh are 23.17 and 43.59 tons / ha respectively. Therefore, the exponential equation form of NDVI was used for rice biomass mapping with the following equation: y = 365.92x – 170.95; where y, and x are the rice AGB, and the NDVI, respectively. Quantitative comparison of rice biomass between analysis results and reference data showed a linear relationship with R2 = 0.6685. The standard error of this estimation is 30.2 ha. To improve the accuracy and increase the applicability of the research into practice, a number of measures can be taken such as: using satellite images with a higher resolution; sample plot size is also larger than.   References   Campos-Taberner, M.; García-Haro, F.J.; Camps-Valls, G.; Grau-Muedra, G.; Nutini, F.; Busetto, L.; Katsantonis, D.; Stavrakoudis, D.; Minakou, C.; Gatti, L.; et al (2017), Exploitation of SAR and Optical Sentinel Data to Detect Rice Crop and Estimate Seasonal Dynamics of Leaf Area Index. Remote Sens. 9, 248; Drusch, M.; Del Bello, U.; Carlier, S.; Colin, O.; Fernandez, V.; Gascon, F.; Hoersch, B.; Isola, C.; Laberinti, P.; Martimort, P.; et al. (2012), Sentinel-2: ESA\'s Optical High-Resolution Mission for GMES Operational Services. Remote Sens. 120, 25-36. Ferrant, S.; Selles, A.; Le Page, M.; Herrault, P.A.; Pelletier, C.; Al-Bitar, A.; Mermoz, S.; Gascoin, S.; Bouvet, A.; Saqalli, M.; et al. (2017), Detection of irrigated crops from Sentinel-1 and Sentinel-2 data to estimate seasonal groundwater use in South India. Remote Sens. 9, 11-19. Li, K.; Brisco, B.; Yun, S.; Touzi, R. (2012), Polarimetric decomposition with RADARSAT-2 for rice mapping and monitoring. Can. J. Remote Sens. 38, 169–179. Mutanga, O.; Adam, E.; Cho, M.A. (2012), High density biomass estimation for wetland vegetation using WorldView-2 imagery and random forest regression algorithm. Int. J. Appl. Earth Obs. Geoinf. 18, 399–406. Nguyen  Thanh Son, etc (2014), A Phenology-Based Classification of Time-Series MODIS Data for Rice Crop Monitoring in Mekong Delta, Vietnam. RS ISSN 2072 – 4292. Phạm Quốc Trung, Nguyễn Hoàng Khánh Linh, Huỳnh Văn Chương, Nguyễn Văn Tiến (2018), Sử dụng ảnh vệ tinh để xác định trữ lượng cacbon của cây lâu năm ở huyện Bố Trạch, tỉnh Quảng Bình. Tạp chí Khoa học Đại học Huế: Nông nghiệp và Phát triển nông thôn; ISSN 2588–1191. Tập 127, Số 3A, 2018, Tr. 49–66. Trần Thị Hiền , Võ Quang Minh , Huỳnh Thị Thu Hương , Trần Thanh Dân, Hồ Văn Chiến , Nguyễn Hữu An4 và Nguyễn Phước Thành, (2013), Theo dõi hiện trạng trà lúa phục vụ cảnh báo dịch hại lúa trên cơ sở sử dụng công nghệ viễn thám và hệ thống thông tin địa lý GIS. Tạp chí Khoa học trường Đại học Cần thơ. 143-151. Wang X., Shi X. and Ling F. (2010), Images difference of ASAR data for rice crop mapping in Google Scholar. Xiao, X.; Boles, S.; Frolking, S.; Li, C.; Babu, J.Y.; Salas, W.; Moore, B. (2006), Mapping paddy rice agriculture in South and Southeast Asia using multi-temporal MODIS images. Remote Sens. Environ. 100, 95–113. Xiao, X.; Boles, S.; Liu, J.; Zhuang, D.; Frolking, S.; Li, C.; Salas, W.; Moore, B. (2005), Mapping paddy rice agriculture in southern China using multi-temporal MODIS images. Remote Sens. Environ. 95, 480–492.     [1] Marshall, M., & Thenkabail, P. (2015). Developing in situ non-destructive estimates of crop biomass to address issues of scale in remote sensing. Remote Sensing, 7(1), 808-835

Key word: NDVI, AGB, the growth of rice, Giao Thuy