Spectral neural modeling of multi-stage/indices data for accurate prediction of cotton yield under arid and semi-arid conditions

Abstract

   Artificial intelligence has achieved an advanced step in the field of agricultural crop productivity estimation programs using remote sensing technology that records growth data spectrally during all developmental stages until harvest under conditions of all geographical ranges. This research aimed to apply artificial intelligence to multi-indices/multi-stage spectral sensor data. To estimate production and then predict crop productivity from early growth stages under the influence levels of water and nutrient supply stress for the cotton crop under growing conditions in arid and semi-arid areas, as well as in order to determine the best predictive neural models with a smaller number of data for input spectral indices and specific growth stages that It gives the highest accuracy of forecasting crop productivity. Productivity was estimated and predicted according to three neural models: the first: the estimated/comprehensive model: which estimates productivity based on 231 factors of data for all spectral indices (21 indices) during all stages of development (11 stages), the second: the predictive/comprehensive model: which predicts productivity based on 63 factors for the data of all spectral indices /21/ from (first 6 stages of 11). Third: The predictive/abbreviated model: which predicts productivity based on 30 factors for the data of a number of spectral indices /17/ from early stages (first 6 stages of 11). The average productivity estimated by the three neural models was: 167.1, 168.4, and 169.3 kg/1000m2, respectively, compared to the actual productivity of 168.4 kg/1000m2for the cotton crop (as an average under all stress conditions). The predictive spectral model designed with artificial intelligence according to a certain number of specific spectral indices and with specific growth stages (the predictive/abbreviated model) gave high accuracy in predicting crop productivity compared to both the estimated and comprehensive predictive models under the influence of growth factors and water and/or nutritional stresses versus the recorded productivity. Field conditions in dry and semi-arid areas.