March 3, 2020

Remote sensing imagery helps to adjust irrigation for several almond varieties according to their water requirements

Marie Marion
Agronomist and Project Manager


Objective: Analyse the spectral signature of different varieties to determine whether a correlation can be established and can help to better manage the irrigation of valves A and B. 

Context: 4 varieties are planted on the orchard. The irrigation system is installed according to these varieties, separated into two valves A and B. A waters all the Nonpareil variety and B the 3 other varieties. 

Method

Location: Hillston, NSW, Australia

Data source: Aerial imagery from the 5th of Dec 2019

Resolution: 0.45 m/px

Layer calculated: NDVI and CCCI The values from NDVI 0 to 0.2 (spaces between plots/streets) were masked, so they wouldn’t interfere with the index calculation. 

Figure 1: Map of the orchard and the field analysed in this study. 

Boundaries: 100 boundaries have been drawn - one per each tree row - with the objective of analyzing them individually.  

Zonal statistics were calculated for each one of the rows. Average NDVI and CCCI have been generated and the values were used to colour the rows’ boundaries.

Varieties: The almond orchard has 4 different almond varieties planted. 

Figure 2: Map of the irrigation system and varieties on the orchard. 

Inter - variety analysis

Figure 3: Price variety
Figure 4: Peerless variety
Figure 5: Carmel variety
Figure 6: Nonpariel variety
Figure 7: All varieties merged, analyzed and compared

Results

CCCI increases as NDVI decreases between rows along a west-east gradient, suggesting that something other than the variety, influences the spectral signature of the orchard.   

Results show a strong effect of row location within the block with a correlation of R2 = 0.6 between row location and NDVI and R2 = 0.5 between row location and CCCI. 

Under these conditions, it difficult to identify a correlation between the varieties and the vegetation indices due to an important spatial variability. 

However, we can identify a pattern in the distribution of NDVI of the red varieties and blue varieties. The NDVI spectrum of the red varieties ranges from 0.78 to 0.85, with a peak at 0.78 and 0.82. While the NDVI spectrum of the bleu varieties is more diffuse ranging from 0.78 to 0.86 with 2 peaks at 0.8 and 0.86. 

Figure 8: Distribution of the NDVI for each variety. The vertical axis represents the number of rows detected per range of NDVI
Figure 9: Distribution of the CCCI for each variety. The vertical axis represents the number of rows detected per range of CCCI

Conclusions

There seems that be an important gradient West-East on this block, due to a parameter independent of the variety. This variability could be due to different parameters such as soil type or location of the irrigation pumps (closer to the pump - higher water pressure - higher NDVI?). 

However, a spectral signature has been identified between red and blue varieties and could be used to differentiate varieties and ultimately monitor separately the irrigation. 

In order to conclude this analysis and validate the possibility to use remote sensing imagery to monitor varieties, this study would need to be extended to other dates and location to get rid of the spatial effect. 

Continue reading:

Podcasts
June 2, 2023

Regrow CEO Shares Insights on Hire Learning Podcast

What are the secrets behind hiring a diverse and innovative team?

Regrow
Team
Blog
May 30, 2023

Reflections on the UN Climate Change Conference

Regrow Director of Climate Policy, Dr. Jeff Seale, reflects on progress made since he spoke at the UN Climate Conference.

Jeff Seale, PhD
Director of Environmental Strategy and Climate Policy
Events
May 31, 2023

Future of Food USA | May 31-June 1 | Minneapolis

Join Regrow at Innovation Forum’s Future of Food USA event in Minneapolis, MN.

Regrow
Team

Keep up with Regrow!

There is always a lot going on at Regrow and you are not going to want to miss a thing.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.