PubAg

Main content area

LIDAR and non-LIDAR-based canopy parameters to estimate the leaf area in fruit trees and vineyard

Author:
Sanz, Ricardo, Llorens, Jordi, Escolà, Alexandre, Arnó, Jaume, Planas, Santiago, Román, Carla, Rosell-Polo, Joan R.
Source:
Agricultural and forest meteorology 2018 v.260-261 pp. 229-239
ISSN:
0168-1923
Subject:
apples, canopy, crops, equations, fruit trees, humans, leaf area, lidar, pears, phenology, regression analysis, vegetation, vines, vineyards
Abstract:
This paper is based on two initial hypotheses, firstly, it is proposed that the vegetation volume obtained with a LIDAR-based system or tree row LIDAR volume (TRLV) has a high correlation with the leaf area (LA). Secondly, it is proposed that the projected outer surface or projected tree row surface (PTRS), also LIDAR-based, is linearly related with the LA. The verification of these two hypotheses corresponds to the first two objectives of this work. The third objective is to propose an alternative method, without using LIDAR sensors, simpler and more economical, for in situ LA evaluation.To achieve these objectives a total of 17 blocks of pear, 14 of apple and 26 of vine, in different phenological states, were LIDAR scanned and subsequently manually defoliated. After the field and calculation work, the TRLV and LA were compared. The logarithmic regressions obtained had high correlations. For apple and pear trees the equations are practically the same with R2 of 0.85 and 0.84, respectively. The equation corresponding to vines is somewhat different and has an R2 of 0.86. The regression without species differentiation is 3.66ln(x)+9.65 with R2 = 0.90.Based on the TRLV, the front and top projected surface areas of each block were then obtained and, using these variables, the PTRS. The linear regressions obtained between PTRS and LA have high correlations with R2 of 0.88, 0.85 and 0.80 for apple trees, pear trees, and vineyard respectively. The three crops show very similar behavior. The straight lines are very close, with very similar slopes. With no species differentiation the linear regression model is y = 1.47x – 1.18 with R2 = 0.93.The starting point of the third objective is to obtain the projected surfaces, frontal and top, without using a LIDAR sensor. These surfaces are not as precise as those obtained with LIDAR and for this reason they are referred to as “estimated” projected surfaces. To calculate the estimated PTRS without a LIDAR sensor, the height and depth of the vegetation are measured with a tape measure. It is also necessary to make a visual estimation of the frontal gap-fraction. For this, a training method with known gap-fraction pictograms is proposed. The final results with this non-LIDAR method are very similar to those obtained with LIDAR. This method, although it needs human intervention, is simple, easy, economical and precise for in situ LA estimation.
Agid:
6104228