Downscaling of vegetation indices from multi-satellite throughout-season maize

Phenomenology of the growing season The Normalized Difference Vegetative Index (NDVI) provided by satellites was employed as a replacement for quantifying the output of vegetative biomass. The MODIS sensors 250-m data have been utilized for terrestrial ecosystem modelling and monitoring. MODIS’s land surface data are credible and trustworthy because to their high temporal resolution and broad spectrum of wavelengths. Land cover and land change studies have used the spatially accurate data provided by the Landsat 30m to characterize human-scale processes. Sentinel-2 is a land surveillance satellite with innovative spectrum capabilities, extensive coverage, and excellent spatial and temporal resolutions. The primary purpose of this work is to create a downscaling vegetation indices (VI) database by combining MODIS, Landsat, and Sentinel data into 250m resolution. The most important NDVI indicates the maize growing season in April and August. MODIS, Landsat, and Sentinel 250m derived biophysical information deliver the same biophysical information for moderate-scale biological aspects. This multi-sensor inquiry also includes high-resolution Landsat data, which will be useful for local ecological investigations while keeping the full seasonal dynamic information given by MODIS.


Introduction
Food is an essential component of human life and survival [1].When everyone has consistent physical, social, and financial means to obtain sufficient and safe food that meets their dietary needs and preferences, it promotes reduced poverty, hunger, sustainable development, and overall wellbeing.A vital tool for agricultural policy development, food market standardization, and adjusting planting practices is a grain crop production evaluation model that anticipates food security issues.Early investigations concentrated on estimation production, control and management circulation, yield forecast, and assessment food security [2,3].There has been a lot of study done recently due to remote sensing's capacity to monitor crops in real time and give diverse observation data for the agricultural business with high revisit frequency and precision.Some current large-scale studies have mostly focused on analysing differences in land use [4], health watershed assessment [5], mapping and monitoring urban areas [6], fire detection [7], and soil moisture [8] utilizing GIS software and extensive coverage satellite imagery.In contrast, several extensive studies have employed highresolution satellite photos [9], in addition to theoretical crop yield forecasting methods with good accuracy [10,11].
Cropland monitoring requires high resolution, both temporally and geographically.A higher spatial resolution allows for a more detailed investigation of the terrain.Furthermore, better spatial resolution 1230 (2023) 012143 IOP Publishing doi:10.1088/1755-1315/1230/1/012143 2 can provide improved accuracy in applications such as land categorization due to a fewer number of mixed pixels [12].Higher temporal resolution allows for more precise identification of changes in land cover.This is especially important for croplands since farmland cover varies often.
The approach employs imagery with high spatial resolution, and the use of varied temporal resolutions enables improved temporal resolution.Because of its enormous coverage, high spatial and temporal resolutions, and consistency, satellite remote sensing has become a crucial instrument for analysing and tracking the dynamics of terrestrial ecosystems across vast distances.To analyse vegetation phenology at the regional and global levels, NDVI is the most often used reflectance based VI.The NDVI is used to replace green biomass, which is linked to canopy photosynthesis.The nearinfrared (NIR) and visible red wavelengths are derived to obtain the NDVI from satellite data [10,11].The photosynthetic activity from remote sensing is a plant canopy is represented by NDVI.Higher NDVI readings often indicate increased vegetative vigour and greenness [10,11].This research proposes a method novel photosynthetic downscaling vegetation index and examines the growing season for 3 different satellite imagery.This multi-sensor evaluation, in addition to employing high-resolution Landsat data, maintains the whole seasonally dynamic information captured by MODIS, which will be useful for area ecological studies.

Methodology 2.1. Study Area
The study was conducted in the maize most production, namely Tuban Regency (Coordinates: 111" 30' -112" 35' East Longitude and 6" 40' -7" 18' South Latitude), East Java province.East Java Province is the largest maize producer in Indonesia.Tuban regency have contributed more than 9% of the total national maize production from East Java Province with an average yield of around 5.3 to 6.1 tons/ha.

Methods
Three separate satellite images, including MODIS 250m, LANDSAT 8 OLI, and Sentinel-2A, were utilized to examine the image using satellite remote sensing.Each imagery were obtain from earth explorer usgs https://earthexplorer.usgs.gov .Sentinel-2A has a resolution of 10 meters and a temporal return of 10 days, while Landsat satellite OLI has a resolution of 30 meters and a temporal return of 16 days.The resolution of MODIS Terra Vegetation Types is 250 meters.The MOD13Q1 package includes two fundamental vegetation layers.The first is the current NDVI assessed by the National Oceanic and Atmospheric Administration, often known as the consistency index to NDVI.To remove extra atmospheric noise, the NDVI time series were smoothed using a Moving Average technique.The Enhanced Vegetation Index (EVI) is the second vegetation layer, with improved precision in increased crop zones.The algorithm chooses the best value by each pixel it can get from all observations made throughout the 16-day period.The NDVI

Downscaling imagery
The imagery from Landsat 8 OLI and Sentinel-2A were processed for downscaling imagery.The LANDSAT 8 OLI and Sentinel-2A imagery were downscaled from 30 m and 10 m into 250 m resolution using ENVI®4.4.One important method for boosting the temporal resolution of low imagery is downscaling.Include indicators besides the vegetation cover, including such non-vegetated surface indices, to improve downscaling in dry non-vegetated situations alongside surface reflection [13][14][15].This technique could be benefits to know the vegetation and non-vegetation.Also, the methods of downscaling are helpful for showing regional climatic features [16].The downscaling imagery from LANDSAT 8 OLI and Sentinel-2A were resample to 250 m as a downscaling imagery (Figure 1).

NDVI
NDVI from the satellite data is simply the difference in reflection between visible red and nearinfrared (NIR) wavelengths [17,18].The NDVI scale illustrates the chlorophyll capability of a vegetation.Higher NDVI levels often indicate better vegetation vigour and greenness.
The pixel parameters and topographic elements are determined to be major dynamic drivers in NDVI spatial distribution and scale conversion build for NDVI downscaling model to identify sensitive factors and determine the vertical conversion and establish function [19].

Temporal Vegetation NDVI Pattern
The temporal vegetation NDVI were processed for five years from the year 2013-2017.Every 16 days, this pattern was extracted temporally.As the pattern was extracted, we saw that the growing season based on the days after planting (DAP) started on day 108 and ends on day 232 or approximately 96-112 days and 112-128 days for maize production.Finally, time series NDVI data were utilized to construct the growth cycle estimated NDVI, with the season beginning in early April (Julian date 108) and running until August (Julian date 240) (Figure 2).

Days After Planting downscaling imagery
After the temporal vegetation NDVI pattern were made, we could gather the time series for each year to find the pattern (Figure 3).We collect the Days After Planting (DAP) from the NDVI pattern from the satellite imagery MODIS, Landsat, and Sentinel 2.

Downscaling 3 satellite
Using the downscaling method from the 3 satellite (Figure 4).From the figure 4, we saw in 2016 that the greeness using MODIS are higher than LANDSAT 8 OLI and Sentinel-2A.The MODIS reaches 0.8 greeness, LANDSAT 0.5, and SENTINEL 0.6.In 2017, MODIS reaches 0.7, LANDSAT reaches 0.6, and SENTINEL reaches 0.4.We saw clearly the pattern of DAP and has the same pattern peak at 50-60 days compared in 2017.

Conclusion
The geographically focused research in this study was conducted in Tuban Regency, where East Java Province's maize output peaked from 2013 to 2017.The research was processed from knowing the temporal pattern NDVI, DAP to know the days for the growing crop, and knowing the pattern of the greness NDVI from 3 different satellite imagery by using the Downscaling method.When paired with multi-satellite datasets, the location of maize was identified using the NDVI.Using the Multi-Satellite Vegetation Indices, downscaling was performed throughout the corn harvest season.The downscaling technique calculates the growth season for planting maize using varied and numerous satellite revisit timings.With the aid of the downscaling method, many satellite records may be integrated into a single route.The NDVI time series trajectories depict the growing season, which lasts around 96-112 days each year while cultivating maize from the DAP method.In the Downscaling method, we saw that the year 2016 has a similar pattern compare with the year 2017.In figure 4, we saw that the 3 satellite were compared each other and the value of NDVI from MODIS were highly than Landsat 8 OLI and Sentinel 2. Also, in the year 2017 saw the similar pattern with the year 2016 but the MODIS has high value of NDVI.From this information help for the information for the farmers to grow their crops from the information of NDVI to know the planting system and help the stakeholders on decision planting their crops from this study.The NDVI trajectory might be used to validate maize yield and productivity forecasts.The projection of the strategy also provides regional food security and safeguards against declining output.