Colloquium

Deriving alerting services from crop monitoring time-series in precision agriculture

Organisator Laboratory of Geo-information Science and Remote Sensing
Datum

di 26 mei 2015 09:30 tot 10:00

Locatie Gaia, building number 101
Droevendaalsesteeg 3
101
6708 PB Wageningen
+31 317 48 16 00
Zaal/kamer 2

by Ratih Nurhayati (Indonesia)

Abstract

Tracking down the crop nutrient status over the growing season is the most principle step for monitoring crop health status in precision agriculture. As a measure of the crop response to nitrogen application, the nutrient status is also related to chlorophyll content and an indicator of photosynthetic activity. To detect the growth problem within the crop fields, the crop N status monitoring can be used as a solution. The main objective of this study was to derive an alerting service for potato growth development from crop monitoring time-series in the precision agriculture. A potato field located in the South of the Netherlands was used as a case study to answer the research objectives. The first step in this research was comparing the three different N status measurement methods over the temporal development. After that, the relationships between crop biophysical and biochemical parameters over the growing season were analysed. The crop biophysical indicator that is highly related to the potato N status was identified from the regression analysis between the eight well-known VIs (NDVI, EVI, WDVI, REP, MCARI/OSAVI, TCARI/OSAVI, CIgreen and CIred-edge) and the chlorophyll content (on the leaf and canopy level). The results showed that the chlorophyll ratio index TCARI/OSAVI has the strongest relationship with the leaf chlorophyll among the eight VIs, even though the coefficient of determination value was relatively low (R2=0.517). With a high coefficient of determination value (R2=0.858), CIred-edge has the strongest relationship with the chlorophyll canopy. Next, the TCARI/OSAVI and CIred-edge time series data over the growing season were used in the time series similarity measures. The time series similarity measures based on the distance measures (Manhattan distance, Euclidean distance and Root Mean Square distance) and the correlation measures were used to calculate the crop growth deviation from each experimental plot or subplot towards two selected reference plots: maximum yield and mean curve approach. The results showed that both Euclidean distance and RMSD were the best similarity measures in characterizing the growth status of each experimental plot and subplot over the growing season. However, the Euclidean distance was used in this research to derive the alerting service using the Control Chart theory. The alerting services were available in the 30x30 m experimental plot level and 13x30 m experimental subplot level. In the plot level, the alerting services were acquired using the two different reference plot approach. The results showed that the alerting services were able to give alerts to specific plot or subplot with considering the changes in the plot or subplot condition): changes from “in control” state (green) to alert state (yellow) or “out-of-control” state (red). The alerting services were validated using fused satellite and UAV imagery dataset (STRS dataset) and twelve subplots validation set from the Cropscan dataset. The validation results showed that the alerting services were working properly for both dataset.

Keywords: nutrient status; nitrogen; maximum yield approach; mean curve approach; time series similarity measures; remote sensing; control chart; alerting service.