|
|
Assessment of vegetation phenology using Sentinel-2 time series data
Danilchyk, Tatsiana ; Štych, Přemysl (vedoucí práce) ; Bohovic, Roman (oponent)
Cílem této práce je vyhodnotit detekci fenologických fází vegetace na základě fenometrických parametrů podle archivních dat Sentinel-2 ve vybraných oblastech v období 2018-2020. V první části práce je uveden literární přehled relevantních publikací, na který navazuje popis navržené metodiky. Poté jsou uvedeny výsledky s grafickými materiály a popisem pro jednotlivé sledované lokality. V závěrečné části práce jsou diskutovány výhody a nevýhody vytvořeného algoritmu, na které navazují návrhy na budoucí výzkum a zlepšení. Vyvinutý algoritmus se skládá ze 2 částí. Odmaskování oblačných pixelů a generování na časové řadě vegetačních indexů se provádí v prostředí GEE. Analýza časových řad a detekce SOS a EOS a statistická analýza se provádí v prostředí R. Studované plochy 20 x 20 m reprezentují různé druhy trvalé vegetace na celém území České republiky. Pro hodnocení detekce fenofází jsou zvoleny hodnoty NDVI, RENDVI, NDRE, NDMI a MCARI. Asymetrická Gaussova funkce a Dvojitá logistická funkce jsou aplikovany na časové řady jednotlivých vegetačních období v každé testované lokalitě, fenologické parametry jsou odvozeny na základě prahových hodnot nebo derivací. Výsledky jsou ověřeny na základě in-situ dat poskytnutých ČHMÚ. NDMI vykázal nejvyšší přesnost při detekci SOS při použití Asymetrické Gaussovy...
|
|
|
Detection of drought events using combination of satellite data and soil moisture modelling
Semerádová, Daniela ; Trnka, Miroslav ; Hlavinka, Petr ; Balek, Jan ; Bohovič, Roman ; Tadesse, T. ; Hayes, M. ; Wardlow, B. ; Žalud, Zdeněk
The use of satellite data offers a potentially well usable tool to accurate drought monitoring. The study examines the space-time possibility of agricultural drought detection using freely available data from the MODIS instrument onboard Terra and Aqua satellites that reflects vegetation condition. Vegetation greenness metrics used in this study are based on the spectral reflectance curves in the visible red and near-infrared part of the spectrum and are expressed in relation to the average for the period of 2000-2014. The results are presented in weekly time step for the whole area of the Czech Republic, and are compared to the drought monitor system, based on the SoilClim dynamic model for soil water content estimates. These data, as well as other parameters, such as soil properties and land use, are integrated at 500 meters spatial resolution.
|
|
|
Reliability of regional crop yield predictions in the Czech Republic based on remotely sensed data
Hlavinka, Petr ; Semerádová, Daniela ; Balek, Jan ; Bohovič, Roman ; Žalud, Zdeněk ; Trnka, Miroslav
Vegetation indices sensed by satellite optical sensors are valuable tools for assessing vegetation conditions including field crops. The aim of this study was to assess the reliability of regional yield predictions based on the use of the Normalized Difference Vegetation Index and the Enhanced Vegetation Index derived from the Moderate Resolution Imaging Spectroradiometer aboard the Terra satellite. Data available from the year 2000 were analysed and tested for seasonal yield predictions within selected districts of the Czech Republic. In particular, yields of spring barley, winter wheat, and oilseed winter rape during 2000–2014 were assessed. Observed yields from 14 districts were collected and thus 210 examples (15 years within 14 districts) were included. Selected districts differ considerably in soil fertility and terrain configuration and represent a transect across various agroclimatic conditions (from warm/dry to relatively cool/wet regions). Two approaches were tested: 1) using 16-day temporal composites of remotely sensed data provided by the United States Geological Survey, and 2) using daily remotely sensed data in combination with an originally developed smoothing method. Yields were predicted based on established regression models using remotely sensed data as an independent parameter. In addition to other findings, the impact of severe drought episodes within vegetation was identified and yield reductions at a district level were predicted. As a result, those periods with the best relationship between remotely sensed data and yields were identified. The impact of drought conditions as well as normal or above-normal yields of the tested field crops were predicted using the proposed method within the study region up to 30 days prior to harvest.
|
|
|
LINCOLN – an algorithm for filtering daily NDVI MODIS data and deriving the start of the season
Bohovič, R. ; Hlavinka, Petr ; Semerádová, Daniela ; Bálek, L. ; Tadesse, T. ; Hayes, M. ; Wardlow, B. ; Trnka, Miroslav
Monitoring drought has become an important tool for farmers and agriculture decision makers. This has increased efforts to create a monitoring system using satellite data that could provide an independent and current source of real information on vegetation condition. The aim of this study was to develop an algorithm for processing Normalized Difference Vegetation Index (NDVI) data from the Moderate Resolution Imaging Spectroradiometer. A software utility called LINCOLN was developed for this purpose. Its filtering output was further processed to yield a start of the season (SOS) metric. Different settings of the utility were tested and correlated to such phenological ground observations as the emergence of spring barley and the beginning of leaf sheath elongation in winter wheat. There was higher correlation observed in the case of winter wheat, probably due to its weaker dependence on crop sowing date. The matrix of coefficients of determination was applied to determine the optimal settings for the LINCOLN filter. The optimal absolute threshold NDVI value for SOS was set to 4,500.
|
|
|
Remotely sensed NDVI as a support tool for agricultural drouhgt assessment
Hlavinka, Petr ; Semerádová, Daniela ; Trnka, Miroslav ; Lukas, V. ; Bohovic, R. ; Balek, J. ; Wardlow, B. ; Hayes, M. ; Tseagaye, T. ; Žalud, Zdeněk
Th e main aim of the submitted study was to introduce how the remotely sensed NDVI (Normalized Diff erence Vegetation Index) could be used for agricultural drought assessment within the Czech Republic. Th e relationship between NDVI values and observed yields of spring barley and winter wheat was analyzed for selected districts. Moreover the ability of NDVI (at district level in the form of seasonal greenness – SG) to explain the water balance or drought occurrence and severity was tested. For this purpose a data mining technique was used. A relative form of the Palmer Drought Severity Index (rPDSI) was used as a dependent variable to indicate drought occurrence. A Standardized Precipitation Index (SPI), percentage of average SG (PASG), Start of Season Anomaly (SOSA) and district identifi cation were used as independent variables. MODIS (Moderate Resolution Imaging Spectroradiometer) observations from the Terra satellite were used as a source of NDVI. Th e situation within 6 selected districts (Olomouc, Přerov, Znojmo, Břeclav, Žďár nad Sázavou and Havlíčkův Brod) during the period from 2000 to 2012 was analyzed. Promising results were achieved, so practical use of this approach (e.g. for spatial and temporal assessment of drought stress within the vegetation) could be expected.
|
host ::
RSS.