Detection of paddy rice drought stress with sentinel image vegetation index and the relation with productivity in Allatengae Village, Bantimurung District, Maros Regency

Maros Regency is one of the rice storage areas in South Sulawesi. According to the Central Bureau of Statistics for Maros Regency in 2020, the harvested area of rice decreased by 10.11% or 4,849.55 Ha compared to 2018 and rice production decreased by 16.96% or 37,719.68 tons. This causes the productivity of rice also decreased. One of the factors causing the decline in rice production in Maros Regency is drought. Sentinel 2 imagery is one of the remote sensing data that can be used to detect rice drought using the drought index method. This study aims to detect dryness of paddy rice in Allatengae Village using Sentinel-2A satellite imagery based on NDDI index analysis and determine the relationship between drought index and rice productivity. The method used is the Normalized Difference Drought Index (NDDI) method to determine the dryness level of paddy fields based on NDVI and NDWI parameters and uses regression and correlation analysis methods. From the results of this study, dry rice in Allatengae Village was detected in September and October 2021. In September (42 HST), the average NDDI value was 0.0254 and was in a mild drought condition while in October (62 HST), the average NDDI value is 0.2425 and is in moderate class drought conditions. From the results of the analysis of the relationship between the drought index and rice productivity, there is a strong (negative) correlation, where the higher the NDDI index value, the lower the rice productivity value.


Introduction
Indonesia is known as one of the agrarian countries where the majority of the population works in agriculture, so agriculture is one of the most important economic sectors in Indonesia.Overall, the climate in Indonesia is a tropical climate that supports a wide variety of agricultural commodities, especially rice to grow well.However, judging from the last few years, where climate change occurs erratically and extreme global warming.This causes frequent droughts in paddy fields.
Drought is one of the problems that need to be considered when the dry season comes.Drought can cause agricultural land, especially paddy fields to become dry due to lack of water so that it cannot meet the water needs of plants [1].This is caused by the lack of rainfall in an area within a certain period of time.One of the areas that has experienced severe drought is Maros Regency.
Maros Regency is one of the rice barn areas in South Sulawesi.According to the Maros Regency Statistics Agency in 2020, the rice harvest area decreased by 10.11% or by 4,849.55Ha compared to 2018 and rice production decreased by 16.96% or by 37,719.68tons.This caused rice productivity to decline as well [2].One of the reasons for the decline in rice production in Maros Regency is drought.1230 (2023) 012147 IOP Publishing doi:10.1088/1755-1315/1230/1/012147 2 The Maros Regional Disaster Management Agency (BPBD) said that drought that occurred in the dry season occurred in almost all sub-districts.Drought can be monitored using remote sensing technology.
Remote sensing is a technique used to detect and study objects from afar without physically touching or coming into contact with them [3].Remote sensing data is obtained from satellite recordings, one example is the Sentinel-2A satellite image, where Sentinel-2A has a high spatial resolution and can help in detecting drought that occurs [4].The method that can be used to identify the level of drought, namely by using the NDDI (Normalized Difference Drought Index) index method.This index combines the greenness parameter of a plant (NDVI) and the wetness parameter of a vegetation (NDWI) [5].

Aims of study
To identify the level of drought using the NDDI index (Normalized Difference Drought Index) and its relationship with rice productivity in paddy fields.

Time and place
This research was conducted from August to December 2021 and took place in Allatengae Village, Bantimurung District, Maros Regency.

Tools and materials
The tools used in this research are laptops for data processing, cameras and QGIS software.
The materials used in this study are primary data consisting of data from interviews with farmers, field survey data and field observation documentation.Secondary data in the form of Sentinel-2A image of Maros Regency which can be accessed on the site https://earthexplorer.usgs.gov/and shp data of rice field plots in Allatengae Village.

Research procedure
The procedure in this study consists of four stages, namely the preparation stage, data collection stage, data processing stage and data analysis stage.

Preparatory stage.
This stage is carried out a literature study by looking for references related to the research title, such as journals, books and other reference materials that can support this research.

Data collection.
The data required for this research are primary data and secondary data.Primary data is in the form of field photos to see the condition of the land and plants.Secondary data in this study are Sentinel-2A images which can be downloaded freely and for free on the website http://earthexplorer.usgs.gov/with the acquisition time from August to November 2021 and shp data of rice fields in Allatengae Village.

Data processing.
At this stage, radiometric correction is carried out on the Sentinel-2A image using QGIS software in order to improve the visual quality of the image in accordance with the spectral radiance value of the object.This correction process can improve the digital number (DN) of the image in each band.This correction process is done through one of the tools in QGIS called the SemiAutomatic Classification Plugin.After the correction process is complete, the next step is cropping the image in the QGIS application.
After that, the NDVI and NDWI vegetation index data were processed.The NDVI formula can be seen in equation (1) and NDWI can be seen in equation (2).

NDVI= (NIR-RED) (NIR+RED)
(1) Where NIR = near infrared channel, Red = red channel and SWIR = shortwave infrared channel.After that, the NDDI index was processed to detect rice field drought based on NDVI and NDWI parameters using equation (3).

NDDI= (NDVI-NDWI) (NDVI+NDWI)
(3) Then classify the NDDI index values obtained in accordance with Table 1 to determine the level of drought that occurs in the research location and then create a map of rice field drought using QGIS software.
The dependent variable (y) is the value of rice productivity (tons/ha), the independent variable (x) is the value of (NDDI) and a,b are constants.From the equation can be obtained the coefficient of determination (R2) which will be used to see the magnitude of the influence of the NDDI index on rice productivity.The correlation analysis is a statistical analysis that states the degree of relationship between the independent variable and the dependent variable to determine the strength or weakness of the relationship between the variables.The magnitude of the correlation value is called the correlation coefficient expressed by the symbol r.The value of r always lies between -1 to 1.

NDDI method drought analysis
NDDI (Normalized Difference Drought Index) method drought analysis is done by first processing the NDVI (Normalized Difference Vegetation Index) and NDWI (Normalized Difference Water Index) indices.The bands used in this method are band 4 (Red), band 8 (NIR) and band 11 (SWIR) which have been radiometrically corrected.2. From the table above, it can be seen that at the age of 12 HST rice, the vegetation index is classified as low with a value of -0.0703 -0.6492 with an average value of 0.1701, this is because rice is in the early planting phase so that the rice fields are not yet fully covered by the rice canopy.According to Cahyono et al. (2019) [6], at the beginning of planting or growth the vegetation index value of rice plants will be low because it is dominated by the appearance of water.Then the vegetation index tends to increase at the age of 42 HST which has a value range of 0.1858 -0.8297 with an average value of 0.6793 and 62 HST which has a value range of 0.1510 -0.8658 with an average value of 0.7720.This occurs because at that age the highest vegetation class is a dense vegetation class, occurring due to the vegetative and generative phases of rice which is characterized by rice leaves growing thicker.In the same area at the age of 92 HST the vegetation index tends to decrease which has a value range of 0.1391 -0.3212 with an average value of 0.2786.This happens because at that age the lowest vegetation class is a dense vegetation class, occurring due to the ripening phase of rice which is characterized by yellowing of rice leaves.
The following graph of vegetation index values at each plant age can be seen in Figure 1.The results of the NDVI map can be seen in Figure 2, where the map shows that white indicators are areas with low NDVI values and green with high NDVI values.In the first map on August 27 (12 HST), it is still dominated by white color, because at that time the rice is still in the early phase of planting so that the rice fields are still not completely covered by rice.Then the second map on September 26 (42 HST) the white color began to decrease at that time the rice was in the vegetative phase but its growth had not covered the entire rice field area.Furthermore, the third map on October 16 (62 HST) changes in color patterns began to be dominated by green.At the age of 62 days, rice is in the generative phase and all parts of the rice field are covered by rice and appear green when viewed from above.However, when approaching harvest time on November 15 (92 HST) the NDVI value decreased and the green color on the map began to decrease.This is because when approaching harvest time the rice turns yellow and the chlorophyll in the plants also begins to decrease.3. values between -0.1696 to 0.5053.From the range of values, it can be seen that the rice field area in the study location is in a moderate humidity condition, low level drought to very high drought.At the age of 12 HST, the average NDWI value was 0.4185 and was in a low humidity condition.The NDWI value tends to decrease at the age of 42 HST with an average value of 0.3086 and is in low drought conditions and at the age of 62 HST has an average value of 0.2455 and is in moderate drought conditions.In the same area at the age of 92 HST has an NDWI value of 0.1998 and is in a high drought condition.All of these categorizations are based on the classification system proposed by Amalo et al. (2018) [7].Based on the data above, it can be seen that 1230 (2023) 012147 IOP Publishing doi:10.1088/1755-1315/1230/1/0121476 the lower the NDWI value of an object, the drier it is.The following graph of the wetness index value at each plant age can be seen in Figure 3 .

Figure 3. Graph of NDWI index values.
The results of the NDWI map can be seen in Figure 4, where the map shows that the white color indicator is an area with a low NDWI value and blue color with a high NDWI value.In the first map dated August 27 (12 HST), the blue color is dominated, indicating a moderate level of humidity.Furthermore, on the second map on September 26 (42 HST) the blue color began to decrease and turned into white in several rice fields in the north and south and was dominated by mild drought levels.Meanwhile, on the third and fourth maps on October 16 (62 HST) and November 15 (92 HST), the NDWI value is decreasing marked by the dominance of white color which indicates moderate to high drought levels.The lower the spectral value of the wetness index result of an object, the drier the object, conversely the higher the level of spectral value of the wetness index result of an object, the wetter the object.

Drought analysis using the NDDI method
The NDDI (Normalized Difference Drought Index) algorithm in this research is used to determine drought on agricultural land.This NDDI processing uses a combination of NDVI and NDWI formulas.The processing results in each month display the NDDI drought index value from August to November.4, the NDDI index experienced clear changes from August to November where the NDDI index increased.The processing results show that the NDDI value varies in each month.In August when the age of rice was 12 HST the average NDDI value was -0.0516 and was in normal condition.The NDDI value tends to rise in September at the age of 42 HST with an average value of 0.0254 and is in mild drought conditions and in October at the age of 62 HST has an average value of 0.2425 and is in moderate drought conditions.In the same area in November at the age of 92 HST has an NDDI value of 0.7904 and is in a high class drought condition.All these categorizations are based on the classification system proposed by Renza et al. (2010) [8].Based on the data above, it can be seen that the higher the NDDI value of an object, the drier it will be.The following graph of NDDI index values at each plant age can be seen in Figure 5. Based on the data analysis carried out, the results of the drought value with NDDI variations in the Allatengae Village area consist of several levels of drought.Classification of vegetation drought levels in the study area can be seen in Figure 6.On the first map on August 27 (12 HST), the condition of the research location is in the normal category marked with green color.This is because rice is still in the early phase of planting and rice fields are still dominated by the appearance of water.On the second and third maps on September 26 (42 HST) and October 16 (62 HST), the condition of the research location has experienced drought in several northern and southern rice fields marked by green color changes.This is because irrigation water in the research location is not able to meet the water needs of the northern and southern areas.Then on the fourth map on November 15 (92 HST) drought (very heavy) dominates the research location which is marked in red on the map.This is because the condition of rice has been in the ripening phase and does not really need water.According to Rahman et al. (2017) [9], generally to determine the occurrence of drought in irrigated rice fields, it is determined in the vegetative and generative phases.Where in rice plants the vegetative phase starts from the age of 0-55 days while the generative phase starts from the age of 55-80 days.For this reason, in this study the determination of drought was seen on August 27 (12 HST), September 26 (42 HST) and October 16 (62 HST).From the explanation above, it can be seen on the map that drought has occurred in September to October, precisely in the northern and southern parts of the research location.While on August 27 (12 HST) there was no drought.This is because in the early planting phase, the land conditions were still dominated by water.

Correlation test between drought index (NDDI) and rice productivity
The correlation test used in this research is correlation analysis (r) and regression analysis.At this stage using 15 productivity samples taken at the research location.By correlating the relationship between NDDI and rice productivity, the equation can be seen in Figure 7.  From the results of the nonlinear regression analysis of the exponential model above results in the equation y = 6.615e -0.589x with an R square (R2) value of 0.8275, it can be seen that the coefficient of determination (R2) obtained from the regression of rice productivity with the NDDI index is quite high.The coefficient of determination explains the influence between the independent variable and the dependent variable.The coefficient of determination is converted to percent, so it can be said that the NDDI variable has an influence of 82.75% on the rice productivity variable.While the rest is influenced by other factors that are not included in the regression analysis, such as pests, planting systems and fertilization.According to Ariani et al. (2020) [10], other factors that can affect rice productivity include the level of rice health, irrigation system, fertilization, type of rice, pests and rice planting system.The processed data obtained a correlation coefficient value of -0.90 which means that the correlation that occurs between NDDI values and rice productivity is a negative correlation, which means that the higher the NDDI value, the lower the productivity.The strength of the relationship between NDDI and rice productivity is in the strong category.This is in line with the opinion of Guilford (1980) in Noer (2008) [11] who interpreted the correlation value between -0.70 to -0.90 as a strong correlation relationship.

Accuracy test of rice productivity model
The accuracy test of the rice paddy productivity model was conducted by applying the vegetation index regression formula that had been generated by the regression analysis conducted earlier.The process of regression analysis of vegetation index values and field rice productivity produces a regression formula that can be used to predict the value of (y).In this study, the calculated (y) value is the value of rice productivity obtained from the pixel value of the vegetation index image.The regression formula used is y = 6.615e-0.589(NDDI).The samples used in this model test are 6 different samples from the 15 samples used for regression analysis.The following rice productivity based on estimation and measurement can be seen in Table 5.In Table 5, it can be seen that there is a deviation between the estimation results based on the modeling results and the conditions in the field.The average deviation is around 0.46 tons/ha or 9.20%.The deviation value shows how much the data value deviates from the estimated results and the actual data productivity, if the deviation value is smaller then the value is said to be good or in accordance with the situation in the field.From the table above, it can be seen that the estimated productivity value on plot 49 is 4.57 tons/ha and plot 214 is 4.76 tons/ha, while from the measurement results, the productivity value on plot 49 is 4.03 tons/ha and plot 214 is 5.66 tons/ha.This shows that the estimation results have different values, some are smaller and some are greater than the productivity value obtained from the measurement results.According to Nafi (2017) [12], more or less the estimated value with the value in the field is influenced by differences in the age of rice plants, health levels, differences in the types of 1230 (2023) 012147 IOP Publishing doi:10.1088/1755-1315/1230/1/01214710 plants planted and cloud disturbances in the process of taking satellite images.The following graph of the relationship between estimated productivity and measurement results can be seen in Figure 8.In Figure 8, it can be seen that the pattern of the distribution of points is close to a linear line, indicating that rice productivity based on estimation and measurement results shows a fairly accurate relationship.However, there are some data that are underestimated, which means that the estimated productivity is lower than the measured productivity shown in the points below the linear line.

Conclusion
From the results of the study, the following conclusions were obtained: a. Paddy rice in Allatengae Village detected drought in September and October 2021.In September (42 HST), the average NDDI value was 0.0254 and was in a mild class of drought conditions while in October (62 HST), the average NDDI value was 0.2425 and was in a moderate class of drought conditions.b.The results of the correlation analysis (r) show that the relationship between the vegetation index and rice productivity has a strong (negative) correlation, where the higher the NDDI index value, the lower the rice productivity value.c.The results of the accuracy test of the rice productivity model obtained the average deviation of the estimated productivity with measurements of 9.20% or 0.46 tons/ha.Estimation (ton/ha) Measurement (ton/ha)

References
4.1.1.Vegetation Index Analysis.NDVI values are obtained by calculating Near Infrared with Red reflected by plants.The NDVI value calculation uses band 4 (Red) and band 8 (NIR) of the Sentinel-2A image.The following vegetation index result values can be seen in Table

Figure 7 .
Figure 7. Relationship between NDDI index and rice productivity in 62 day after planting.

Figure 8 .
Figure 8. Relationship graph between estimated productivity and measured results.

[ 1 ]
Mishra A K and Singh V P 2010 A review of drought concepts Journal of hydrology.391 202-216.[2] Central Bureau of Statistics 2020 Booklet Statistics Agriculture 2020 [3] Gupta R P 2017 Remote Sensing Geology Third Edition Springer [4] Berger M, Moreno J, Johannessen J A, Levelt P F and Hanssen R F 2012 ESA's Sentinel missions in support of earth system science Remote Sensing of Environment.120 84-90 [5] Gu Y, Jesslyn F B, James P V and Brian W 2007 A five-year analysis of MODIS NDVI and NDWI for grassland drought assessment over the central great plains of the United States Geophysical Research Letters.34 L06407 [6] Cahyono B E, Nugroho A T and Arifilla A 2019 Rice plant age analysis based on NDVI values

Table 1 .
NDDI classification.The analysis used is regression and correlation analysis.The regression analysis used is non-linear regression of the exponential model.Nonlinear regression is a method of regression analysis to obtain a nonlinear model used to determine the relationship between the independent variable and the dependent variable.This analysis is intended to determine the relationship between the NDDI index and rice productivity.The exponential model regression equation is written in equation 4.

Table 3 .
NDWI spectral results NDWI index from August to November decreased.NDWI values on rice fields in Allatengae Village in August to November have a range of

Table 5 .
Rice productivity based on estimation and measurement.