TIMESAT 3.1
Software Manual
Lars Eklundh
a
and Per Jönsson
b
a
Department of Earth and Ecosystem Sciences, Lund University, S-223 62 Lund,
Sweden
b
School of Technology, Malmö University, S-205 06 Malmö, Sweden
2012-11-15
Contents
1 Introduction 5
1.1 About TIMESAT and the software manual . . . . . . . . . . . . . . . . . . . 5
1.2 TIMESAT version 3.1 vs. version 3.0 . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 TIMESAT home page . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Using and citing TIMESAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Applications of TIMESAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 About the authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Overview of data processing 10
2.1 Sequential data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Image data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Methodology 13
3.1 Least-squares fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 On the use of ancillary quality data for assigning weights . . . . . . . . . . . . 14
3.3 Pre-processing to remove spikes and outliers . . . . . . . . . . . . . . . . . . . 15
3.4 Adaption to the upper envelop e . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.5 Determination of the number of seasons . . . . . . . . . . . . . . . . . . . . . 16
3.6 Adaptive Savitzky-Golay filtering . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.7 Fits to asymmetric Gaussians and double logistic f unctions . . . . . . . . . . . 17
3.8 Separable non-linear least-squares fits . . . . . . . . . . . . . . . . . . . . . . . 19
3.9 Merging of local functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4 Extraction of seasonality parameters 21
4.1 Seasonality parameters derived from time-series spanning n years . . . . . . . 21
4.2 Extracting seasonality parameters from one year of data . . . . . . . . . . . . 21
4.3 Extracted seasonality parameters . . . . . . . . . . . . . . . . . . . . . . . . . 22
5 Aspects of processing 23
5.1 Characteristics of the processing methods . . . . . . . . . . . . . . . . . . . . 24
5.2 Controlling the process ing: input settings . . . . . . . . . . . . . . . . . . . . 24
5.3 Description of input settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
6 Output data 31
6.1 Files with time-series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6.2 File with seasonality parameters . . . . . . . . . . . . . . . . . . . . . . . . . 32
6.3 Extracting images of seasonality parameters . . . . . . . . . . . . . . . . . . . 32
6.4 Output files from ASCII data . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6.5 Index files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
7 Installation of TIMESAT and program structure 36
7.1 System requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.2 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
8 Program and processing overview 37
2
9 Getting started with TIMESAT – a quick tutorial 40
9.1 Preparing the data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
9.2 Starting the TIMESAT menu system . . . . . . . . . . . . . . . . . . . . . . . 42
9.3 TSM_IMAGEVIEW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
9.4 TSM_GUI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
9.5 TSM_settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
9.6 TSF_process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
9.7 Post-processing the results of a TSF_process run . . . . . . . . . . . . . . . . 52
9.8 Checklist for processing new vegetation index image data . . . . . . . . . . . 59
10 Reference manual 60
10.1 TSM_menu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
10.2 TSM_imageview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
10.3 TSM_GUI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
10.4 Data input for TSM_GUI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
10.5 Settings in TSM_GUI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
10.6 Output files from TSM_GUI . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
10.7 TSM_settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
10.8 TSM_process and TSF_process . . . . . . . . . . . . . . . . . . . . . . . . . 69
10.9 TSF_readheader (obsolete) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
10.10TSM_fileinfo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
10.11TSM_printseasons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
10.12TSM_viewfits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
10.13TSF_fit2time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
10.14TSF_fit2img . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
10.15TSF_seas2img . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
10.16TSF_merge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
10.17Processing from the command prompt . . . . . . . . . . . . . . . . . . . . . . 73
10.18Input files for TIM ESAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
10.19Output files for TIMESAT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
10.20Index files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
11 Acknowledgements 79
12 References 79
3
Part I
Introduction to TIMESAT
4
1 Introduction
Time-series of vegetation index derived from satellite spectral measurements can be used to
gain information on seasonal vegetation development. This information aids analyzes of the
functional and structural characteristics of the global and regional land cover and adds to our
current knowledge of global cycles of energy and matter. Long time-series of vegetation index
data can also provide information on shifts in the spatial distribution of bio-climatic zones,
indicating variations in large-scale circulation patterns or land-use changes.
Although the value of remotely sensed time-series data for monitoring vegetation seasons has
been firmly established, only a limited number of methods exist f or exploring and extracting
seasonality parameters from such data series. For this reason the Timesat program package
for extracting seasonal parameters has been developed.
1.1 About TIMESAT and the software manual
The Timesat 3.1 sof tware manual consists of three parts. Part I gives general information
together with examples of some applications of Timesat. Part II describes the algorithms
underlying the software package. Also the s ettings affecting the process ing are discussed in
detail. Part III is the sof tware user’s guide, with detailed information on how to install, run,
and handle the program package.
The Timesat program package is designed primarily for analyzing time-series of satellite
data and uses an adaptive Savitzky-Golay filtering method and methods based on upper
envelop e weighted fits to asymmetric Gaussian and double logistic model functions (Jönsson
and Eklundh, 2002, 2003, 2004). From the fitted model functions a number of seasonality
parameters, e.g. beginning and end of the growing season, can be extracted. Parameters for
a number of pixels can be merged into a map displaying seasonality on a regional or global
scale.
Timesat consists of a number of numerical and graphical routines coded in Matlab and
Fortran 2003. Timesat is normally run from Matlab. However, also users without Matlab
installed, or even without a Matlab license, can use the s oftware. Processing is then done
through a runtime engine called the Matlab Compiler Runtime (MCR), that is set up on
the users machine by executing the file MCRinstall.exe (see section 7.2). The latter file is
provided along with the Timesat program package. Fortran routines are highly vectorized
and efficient for use with large data sets. The Fortran routines are pre-compiled for Windows
and Linux.
Timesat has been developed over many years. During these years a number of new features
have been added. Below are the main features of Timesat 3.1
• Contains several smoothing methods for time-series data
• Contains several methods for detection of outliers
• Allows for weighting of data using quality information
• Allows for fitting to the upper envelope of the data
• Contains two methods for defining start and end of growing seasons
5
- 1
- 2
前往页