|
Rice Crop Modeling Using Age Index Based on LANDSAT 7 ETM Data Dede Dirgahayu, Parwati Researchers, Remote Sensing Application and Development Center, LAPAN, Indonesia ABSTRACT This research is aimed to develop estimation of the age of rice crop model based on remote sensing data, especially Landsat 7 ETM, which has 30 m in spatial resolution and 16 days in temporal resolution. This research has been used Normalized Difference Vegetation Index (NDVI) method to estimate the age of rice (paddy) crop in paddy field of PT. Sang Hyang Seri (Subang, West Java) which has good plant schedule and IR 64 variety as a domain variety in the Java Island. The result found that rice crop has same value of NDVI in the different age. Based on that result a linear model can be created to estimate the age of rice crop using Age Index (AI). The age index is obtained from channels combination of green reflectance (R2), near infra red reflectance (R4), and middle infra red reflectance (R5) of Landsat 7 ETM. The formula is represented as the following equation: A (days) = AI - 27 .73 R2 = 0.91 AI = 3.212*R2 - 1.806*R4 + 7.418*R5 (t=5.93**) (t= -6.31**) (t= 16.59**) The implementation of that model can be used to predict paddy harvest area for 3 months in spatial or tabular information every District in Java Island. I. Introduction The inventory of plant area, harvest area and paddy production is conventionally conducted by statistical institution or Central Bureau Statistics (BPS) and the Department of Agriculture in regency level to province level. That information usually can be known a few months later after the agro census was done in tabular format, and not in spatial format. Beside that the conventional method has less accuracy in collecting data over a large area. Nowadays, It has been implemented some methods which are used in collecting and estimating plant area, harvest area, and paddy production by statistical institution / BPS (Maksum, 1998), Logistic Institution /BULOG (Mulyana, et al, 1998), Agriculture Departement (Napitupulu, 1998), and National Institute of Aeronautics and Space / LAPAN (Dirgahayu, 1999). Based on the method for collecting data, prediction of harvest area can be divided into 2 two ways. The first method is based on the structure of organization from village level, sub district, regency, province, and to national level. This method needs sufficient time and more executors in each ladder. Institutions which work with this method are BPS, Department of Agriculture, and BULOG. The second method more emphasizes in the use of remote sensing data for estimating paddy area and monitoring of paddy growth. LAPAN and Soil-Agro climate Research Institute (Puslittanak) usually work with this method. Department of Agriculture estimates paddy production with a few steps. They consist of tabulating harvest area and plant area dates, and measuring of yield productivity. Department of Agriculture cooperates with BPS to do this work. In order to estimate harvest area and plant area, Department of Agriculture estimates area which is done by farmers and agriculture extension agent. While Department of Agriculture uses conventional method. LAPAN has developed spatial prediction model of harvest area using Landsat data (Dirgahayu, 1999). That model is based on the age of paddy using Normalized Difference Vegetation Index (NDVI) method. The normalized difference vegetation index (NDVI) is a normalized ratio of the NIR and red bands. The NDVI is successful as a vegetation measurement, so that is sufficiently stable to permit meaningful comparisons of seasonal and inter-annual changes in vegetation growth and activity. According to estimate the age of paddy which is based on NDVI, a polynomial model has been created since 1997 using Landsat TM data. There is one thing should be considered that a crop in difference ages can have an equal of NDVI value. This condition happens because there is sigmoid growth effect in seasonal plant. Then the model has been separated in two stages of crop growth, vegetative model (8 weeks after planting) and generative model (8 weeks to 14 weeks after planting). Using the age of paddy information, harvest area and time of harvest can be described. This research is aimed to establish the age of paddy estimation model using Landsat 7 ETM+ reflectance combination. II. Method 2.1. Study Area and Data Landsat 7 ETM+ images acquired on 2001/2002 (path 122/row 64) and digital paddy field were applied in this study. The images covers the areas of PT. Sang Hyang Seri, Regency of Subang, West Java. Beside that tabular data such as planting schedule and harvest date in growing season 2001 and 2002, and topographic map (1: 25.000) were used. Paddy harvest area information is obtained from Departement of Agriculture in Subang Regency. 2.2. Analysis 2.2.1. Geometric Correction Geometric correction of Landsat 7 ETM+ is done by using GCP (Ground Control Point) precisions, which are obtained from ground base map (1:25.000). Coordinate transformation (warping) is done by polynomial method in orde 1. We used a clear Landsat data as reference and then resampled to 25 m spatial resolution. 2.2.2. Radiometric Correction Solar zenith angle and atmospheric (absorb process by water vapor, CO2 and O3 gas, and scattering process by aerosol) effect spectral value of image. Therefore radiometric correction must be done to convet digital number to reflectance value. The effect of solar zenith angle and atmospheric become more significant in case of temporal data. In the procedure of radiometric correction, the digital number of Landsat 7 ETM must be converted to radiance value, and then convert to reflectance. The information for coming the radiometric correction (such as: solar zenith angle, acquisition date and so on) can be obtained from Landsat ETM 7 ancillary data. 2.2.3. Statistic value extraction of Landsat 7 Statistic value extraction (minimum, maximum, mean, median and standard deviation) for each channel of Landsat data was done in paddy field of PT. Sang Hyang Seri, where paddy plant is planted in different age in different block area. The age of paddy after planting can be predicted based on the difference between acqusition date of Landsat data and planting date of each block area. To calculate statistical value of Landsat 7 data, training sample must be done for every block area based on data homogenity. Training sample must be considered to composite RGB 542 image (Picture 1 in Appendix) and Index vegetation image (Picture 2 in Appendix), which the index vegetation image has been classified into several classes with 0.025 interval value for every planting block of IR 64 (name of paddy variety). Statistic value in the training sample, which is not heterogen, is shown by narrow range of difference value (Maximum value - minimum value), small standard deviation and small variation coefficient. 2.2.3. Correlation and regression analysis Correlation and regression analysis was done to get predicted model of paddy planting age. Paddy planting age (u) is used as independent variable with index vegetation (IV) and combination of reflectance value (R), as shown bellow:
(b) u = b0 + IU IU = b1*Ri + b2*Rj + b3*Rk is the result of linear regresion between age index and reflectance value of channel (i,j,k) of Landsat ETM 7 data. III. RESULT AND DISCUSSION 3.1. Paddy Crop Growth Model Growth of Paddy Crop during planting season until harvest and the bare land condition can be detected by NDVI value changes. Change of crop condition / crop growth parameter, like high accretion, wide leaves and canopy density, and also influence of weather condition (atmospheric effect) can be presented by NDVI fluctuation. To know profile of paddy crop growth based on NDVI needed Multi Temporal data during growth of paddy crop. Remote sensing data which has high spatial and temporal resolution like Landsat ETM 7 ( 16 days) is very useful to detect land condition and growth changes of paddy crop. Extraction result of NDVI value from Landsat ETM 7 data of during growth paddy crop at growing season in 2001/2002 year at research location (Sang Hyang Seri) is shown at Fig 3-1 ( Phase Vegetative) and Fig. 3-2 ( Generative Phase). Growth pattern of paddy crop at boths phase can be created in 3ed order polynom equation or spline cubic. Growth Paddy Crop Model equation can be expressed as 2 ways : (a) Vegetative Phase : y = 0.0049u3 + 0.3636u2 - 4.9462u + 104.6 R2 = 0.98 (b) Generative Phase : y = -0.0004u3 + 0.0466u2 - 1.7862u + 192.5 R2 = 0.97 ; where u = days after planting (DAP) and y = 128 + 125*NDVI  Fig 3-1. Paddy Crop Growth Profile based on Landsat 7 NDVI at Vegetative Phase  Fig 3-2. Paddy Crop Growth Profile based on Landsat 7 NDVI at Generative Phase Based on both figures show that Maximum NDVI during paddy growth reaches at paddy's age 40 - 55 DAP. Irrigated domination happen until at paddy's age age 20-25 DAP, assessment of NDVI value come to zero. 3.2. Paddy Age Estimation Paddy Age Estimation can be conducted by using both growth crop models as above mentioned, but arise problem if only used one data, because overlap condition happened. There are more than 1 ages values occure in the same NDVI interval. So, it needs 2 data at least data minimize of multi temporal Landsat ETM 7 and should be determined land condition and phase of paddy growth (bare land, water, vegetative, and generative). Estimation of paddy age crop by using both above models are really complicated enough, caused by some step of process which must be conducted and also implementation of model which have the reverse. Therefore, required to be made an index which have the positive correlation to with change of paddy age crop, so that is not happened by prediction values of age which each other the overlap. The index or hereinafter called as Age Index ( AI) can be created based on linear combination of 2 or more channels from Landsat ETM 7 data. Channels which used are have sensitive respond to water object, vegetation and bare land. Based on stepwise regression, we have found that 3 reflectance channels, that are green reflectance (R2), near infra red reflectance (R4), and middle infra red reflectance (R5) of Landsat 7 ETM better than other channels combination. The combination is having Determination coefficient and t-computed of Regression coefficient larger than others and have smaller Standard Error. Mean value reflectances of R2,R4, and R5 that chosen (n = 58) at age interval of paddy's age presented at Tables 3-1. The next step is made regression equation to estimate age of paddy crop (A) by using AI. AI is produced from reflectance combination of R2, R4, and R5 Landsat 7 ETM reflectance. The formula is represented as the following equation : A (days) = AI - 27 .73 R2 = 0.91 AI = 3.212*R2 - 1.806*R4 + 7.418*R5 (t=5.93**) (t= -6.31**) (t= 16.59**) We can create age of paddy mapping only use one Landsat 7 ETM (Fig 3-3) data based that model. The map is very useful to detect bare land, water condition and vegetative, generative phases of paddy crop growth. Time of when harvest occurred and paddy harvest area also can be predict using Paddy Age Map (see Fig 3-4) 
3.4. Model Verification Based on the standard error value (6 %) which is given from model application of paddy growth, we verify that model especially in paddy area of Shang Hyang Seri, Ltd. Then we validated that model using ratio value of harvest area prediction with actual value of harvest area which is obtained from Agriculture Department of Karawang and Subang, for the second growing season in Mei - Agustus, 2002. The result shown that the difference between prediction value and actual value is about 9.5 % in Karawang and 10.3 % in Subang. Validation in other area need to do for operational prediction and monitoring of harvest paddy area.  Fig 3-3. Landsat ETM 7 RGB (5,4,2) Composite at Karawang, and Subang District, West Java  Fig 3-4. Paddy Age Map at Bekasi, Karawang, and Subang District, West Java IV. CONCLUSION Estimation of paddy crop age can be carried out by using paddy crop model, which is based on NDVI value from Landsat 7 ETM, in vegetative and generative phase. Monitoring of paddy crop age using multitemporal remote sensing data need conversion of remote sensing digital number into reflectance to eliminate the atmospheric effect and temporal radiometric difference influence. The best combination channels of Landsat 7 ETM for calculating Age Index (AI) of paddy crop is green reflectance (R2), near infra red reflectance (R4), and middle infra red reflectance (R5) to estimate age of paddy crop. The equation is shown as: A (days) = AI - 27 .73 R2 = 0.91 AI = 3.212*R2 - 1.806*R4 + 7.418*R5 (t=5.93**) (t= -6.31**) (t= 16.59**) We sugest for next research need to use other remote sensing data such as SPOT or ASTER to replace the SLC-off Landsat data. Furthermore, MODIS (Moderate Resolution Imaging Spectroradiometer) data can be used for monitoring with higher temporal resolution. REFERENCE
Page 1 of 1
|