Prediction Models of Corn Yield by NDVI in Function of the Spacing Arrangement

There is a need for the use of tools to estimate productive potential during corn crop development. Thus, the assistence by means of active optical sensors for the generating of vegetation indexes can provide significant information for the knowledge of the behavior and temporal relation of this index with productive parameters of the agricultural crops. It was aimed to evaluate the temporal behavior of NDVI and its relation with yield of corn in order to generate yield prediction models in plant populations (55, 60 and 65 thousand plants ha) in spacing of conventional seeding and twin rows. A factorial 2 × 3 was utilized with four replicates, with a total of 24 experimental plots of 10 m in randomized blocks, performing reading NDVI at 5 seasons (30, 45, 60, 75 and 90 days after emergence of the plants DAE). The spacing in twin rows at 30 and 90 DAE for populations of 55 and 60 thousand plants ha, respectively, allowed to generate models for the prediction of productivity based on corn NDVI, while the population of 65 thousand plants ha at 45 and 60 DAE there was no adjustment by the prediction model of yield by values close to NDVI for different productivities. In the conventional spacing generating models for the prediction of yield was possible in the populations of 55 and 60 thousand plants ha respectively at 30 and 90 DAE.


Introduction
The use of geotechnologies as remote sensing has been used extensively in Precision Agriculture (PA) in order to obtain information without physical contact with the analyzed target, through different types of sensors (active and passive) that makes it possible to calculate indices to estimate the components of agricultural crop production (Zerbato et al., 2016).
The NDVI is applied at the agricultural level in the quantification of crop response at nitrogen doses (Molin et al., 2010), chlorophyll content, leaf nitrogen content and yield (Motomiya et al., 2012) crop biomass (Kapp Junior et al., 2016) and nitrogen recommendation according to spatial variability (Amaral & Molin, 2011).
The relationship of vegetation indices and yield of agricultural crops has been widely studied by several researchers, such as studies by Zerbato et al. (2016) who remarked the yield of the peanut crop with NDVI, Povh et al. (2008) and Bredemeir et al. (2013) used the NDVI to estimate wheat yield, among other numerous examples of using remote sensing to estimate crop yield.
However, NDVI can be influenced in a number of ways, such as population and spacing between rows of plants (Barker & Sawyer, 2012), nitrogen rates (Bredemeier et al., 2013), by the phenological stage of the crop (Rissini et al., 2015) and among other factors.
Considering that the reading of the sensors for the generation of vegetation indices, can be affected by the factors mentioned above, the evaluation of the NDVI behavior in spatial arrangements can provide significant information regarding the knowledge of the behavior and temporal relation of this index with productive parameters of agricultural crops.The fertilization was applied according to technical bulletin 100 of the state of São Paulo, with 350 kg ha -1 of fertilizer formulated 08-28-16 in the sowing furrow.At the V4 stage 22 days after sowing of the crop, cover fertilization with 120 kg of KCl ha -1 and 300 kg of Urea ha -1 was carried out without incorporation, for expected yield of 10-12 t ha -1 the bulletin 100 of the state of São Paulo.
Three days after sowing, 1.2 L ha -1 of Paraquat (200 g L -1 ) were applied to eliminate germinated weeds before corn emergence, after which 2.0 L ha -1 of Atrazine was applied, corn selective herbicide for the elimination of germinated weeds after corn, the volume of syrup used for the applications were 200 L ha -1 and tip TT11002.

Used Machines
The Massey Ferguson tractor model MF 7370 with power of 125 kW (170 hp) in the engine, nominal rotation of 2000 rpm that gives 540 rpm in the power take-off, Tractor (4 × 2 TDA), working in the march L3, to realize the sowing of the crop.
For the sowing operation, the prototype Jumil seed drill, model 3070 Exacta Air with seven sowing rows, with pneumatic seed distribution system was used; tires 650 × 16E, 10 canvas; hydraulic line marker; fertilizer distributor system Fertisystem; 17" smooth cutting disc, fertilizer grooving rods, pantographic seeding units with mismatched double discs; depth controller with parallel bands; minimum spacing between 0.45m lines.
For the reduced spacing, seven spacing lines spaced 0.45m were used and for the spacing of double lines two lines were eliminated.
NDVI was evaluated in five seasons (30, 45, 60, 75 and 90 days after emergence of the plants, around 7 days after sowing), with the aid of an active terrestrial optical sensor (GreenSeeker®), emitting radiation electromagnetic in the red band at 660±12 nm and near infrared at 770±12 nm..
The apparatus was positioned approximately 1.0 m above the canopy as recommended by the manufacturer.The readings were performed manually at a velocity of approximately 1 m s -1 in any useful area of the plot, with a useful width taken by the sensor of 0.9 m, in the five collection seasons, and the collections were carried out around 10 o'clock.The sensor was coupled to the GNSS Nomad Trimble ® collector/receiver for the storage of the georeferenced data.

Sampling Procedures
For determination of yield, all spikes were collected from the useful area of 10 m 2 of each plot and tracked with the aid of mechanical stationary threshing machine.The grains were separated, weighed in a scale of 0.01 g and the values corrected to 13% moisture and extrapolated to kg ha -1 .

Research Design
The experiment was carried out using two spacings (reduced-0.45m and double lines intercalated-two lines of 0.45 m by one of 0.90 m) and three plant populations (55, 60 and 65 thousand plants ha -1 ), composing a 2 × 3 factorial with 4 replicates, making a total of 24 experimental plots of 30 × 4 m (120 m 2 ) with 10 m 2 of useful area in a randomized block design.

Estatistics
The data were submitted to the Anderson Darling normality test (p > 0.05), demonstrating that they had a normal distribution.The results were submitted to regression analysis at 5% probability by the F test, with the aid of the statistical program AgroEstat (Barbosa & Maldonato, 2014).

Results and Discussion
For the double row spacing, populations of 55 thousand plants ha -1 regression coefficient was significant, and at 90 DAE with determination coefficient (R 2 ) of 0.74.However, for a population of 65 thousand plants ha -1 , the NDVI readings were significant with yield at 45 and 60 DAE, with respective determination coefficients (R 2 ) 0.83 and 0.72 (   ----R² ----------0.00n 0.00 n 0.05 n 0.00 n 0.85* ant. e F test was sig e for the coeffi the population elationship be plants ha -1 spa ng was used an DVI (Table 1).
ons used, sinc ments that con the increase o area increases f twin lines, a rating that ND e row spacing The results obtained in this experiment serve as a subsidy for the development of algorithms for the determination of varied nitrogen doses, since Raun et al. (2005) reported that it is necessary to establish yield prediction models for the development of an algorithm for the determination of nitrogen doses that maximize crop yield.

Conclusions
(1) The recommended prediction of yield in maize by NDVI should be performed at 90 and 30 DAE in the populations of 55 and 60 thousand plants ha -1 regardless of the spacing adopted.
(2) For the prediction of yield in a population of 65 thousand plants ha -1 , the use of NDVI from proximal remote sensors is not recommended.
of 65, 60 and 55 thousand plants ha -1 respectively for double-spaced rows (two rows spaced 0.45 m and spaced 0.90 m apart for the others) if the seeding density of 5.5; 5.1; 4.7 seeds per meter to reach populations of 65, 60 and 55 thousand plants ha -1 respectively.