Framework for agricultural performance assessment based on MODIS multitemporal data

Abstract

We present a hierarchical classification framework for automated detection and mapping of spatial patterns of agricultural performance using satellite-based Earth observation data exemplified for the Aral Sea Basin (ASB) in Central Asia. The core element of the framework is the derivation of a composite agricultural performance index which is composed of different subindicators taking into account cropping intensity, crop diversity, crop rotations, fallow land frequency, land utilization, water use efficiency, and water availability. We derive these subindicators from net primary productivity and evapotranspiration data obtained from the MODIS sensor on board the Terra satellite during the observation period from 2000 to 2016, as well as from cropland maps created through multiannual classification of normalized difference vegetation index (NDVI). We classified pixel-based NDVI time series covering more than 8 × 10 6 ha of irrigated cropland based on a hierarchical approach concatenating unsupervised and supervised classification techniques to automatically generate and refine training labels, which are then used to train a decision fusion classifier, achieving an average overall accuracy of 78%. The results give unprecedented insights into spatial patterns of agricultural performance in the ASB. The proposed method is transferable and applicable for global-scale mapping, and the results of this remote sensing-aided assessment can provide important information for regional agricultural planning purposes. © The Authors. Published by SPIE under a Creative Commons Attribution 4.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI.

Modeling canopy water content in the assessment for rainfall induced surface and groundwater nitrate contamination: The Bilate cropland sub watershed

“…Accordingly, for the current study, monthly mean precipitation amounts from the period 1981–2020 are given for mid-March, April and May to the end of June. The dynamics of watershed characterization modeling could be greatly influenced by the selection of an agrometeorological observation station [ 86 , 75 , 66 ].…”

Characterizing Cropland Patterns Across North‐East Africa Using Time Series Vegetation Indices

“…The MOD09Q1 of 8-day NDVI composite of MODIS is believed to have a capacity to distinguish the variety of cropping patterns and is recommended for crop type mapping (Gumma et al, 2011). A recent study on agricultural performance assessment compared satellite imageries from MOD09Q1 and Landsat 8 OLI (30 m) using a hierarchical classification framework and found an overall agreement of 74% between the two products, which was verified through GPS-based in situ observations (Dimov et al, 2019).…”

Modeling canopy water content in the assessment for rainfall induced surface and groundwater nitrate contamination index from cropland N-fertilizer in Bilate downstream

“…As crop growth curve dependent on rainfall and N-application, the classified crop land area is impotent to analyze crop zonal statics mean value for each crop type pixel from MODIS 250m-EVI (Fig 6). The principle of crop growth and biomass dynamic, on crop zonal statics mean value, has applied to identify drained/managed mineral soil index (Ali et al, 2020;Dimov et al, 2019;Dong et al, 2019;Wardlow et al, 2007;G. Zhang et al, 2015).…”