Dalezios_FROST.pps

Frost Risk Mapping
Using Satellite Data
C. Domenikiotis1, M. Spiliotopoulos2, E. Kanelou2 and
N. R. Dalezios1
1Department of Agriculture Animal Production and Aquatic Environment
2Department of Management of Environment and Natural Resources
University of Thessaly
Volos,
GREECE
Aim
Frost risk mapping
Objectives
Examination of cases with radiation frost.
Comparison of satellite derived LST and air
temperature as recorded at the meteorological
stations of this area.
Classification of Thessaly region according to
the temperature pattern of meteorological
stations.
Region of study (Total area ~14.000 Km2)
TYRNAVOS
KARDITSA
AGIA
VOLOS
AGHIALOS
ZAGORA
Dataset
Air Temperature Data, from six meteorological
stations in Thessaly region, for the years 1999,
2000, 2001.
Satellite Data from NOAA/AVHRR for the
years 1999,2000,2001.
Meteorological maps (850hPa and 500hPa).
Methodology
Steps
 Processing of temperature data.
 Preprocessing of satellite data.
 Correlation between satellite and meteorological
data.
 Classification of the study area.
 Spatiotemporal expansion of data.
 Validation.
 Frost risk mapping.
Processing of temperature data
Selection of minimum air temperature (06:00
for summer time or 07:00 for winter time).
Satellite images are georeferenced, and values
of brightness temperature are retrieved.
Comparison of satellite and in situ data.
Image Processing
Utilization
of sixty-six (66) non-cloud
night images from NOAA/AVHRR, where
radiation frost is appearing.
Examination of the synoptic conditions of
the 66 selected days.
16 night images were rejected, where cold
or warm advections are observed.
Finally fifty (50) images with normal
conditions are utilized.
Selection of “clear” (non cloud) images,
(50 images).
Extreme cold advection
(Example)
Normal conditions
(Example)
Finally selected images
Α/Α
Date
Time
Α/Α
Date
Time
Α/Α
Date
Time
1
31-01-2000
6:00
18
24-03-2000
6:00
35
07-02-2001
6:00
2
01-02-2000
6:00
19
25-03-2000
6:00
36
08-02-2001
6:00
3
02-02-2000
6:00
20
07-04-2000
7:00
37
09-02-2001
6:00
4
03-02-2000
6:00
21
09-04-2000
7:00
38
13-02-2001
6:00
5
05-02-2000
6:00
22
05-05-2000
7:00
39
14-02-2001
6:00
6
07-02-2000
6:00
23
06-01-2001
6:00
40
16-02-2001
6:00
7
08-02-2000
6:00
24
07-01-2001
6:00
41
17-02-2001
6:00
8
25-02-2000
6:00
25
10-01-2001
6:00
42
18-02-2001
6:00
9
29-02-2000
6:00
26
11-01-2001
6:00
43
21-02-2001
6:00
10
01-03-2000
6:00
27
12-01-2001
6:00
44
22-02-2001
6:00
11
02-03-2000
6:00
28
13-01-2001
6:00
45
23-02-2001
6:00
12
07-03-2000
6:00
29
15-01-2001
6:00
46
16-03-2001
6:00
13
08-03-2000
6:00
30
16-01-2001
6:00
47
03-04-2001
7:00
14
09-03-2000
6:00
31
19-01-2001
6:00
48
04-04-2001
7:00
15
12-03-2000
6:00
32
03-02-2001
6:00
49
07-04-2001
7:00
16
15-03-2000
6:00
33
05-02-2001
6:00
50
25-04-2001
7:00
17
23-03-2000
6:00
34
06-02-2001
6:00
Correlations Between Ts and Tmin
Station
Relations
r
R2
Fytoko
Ts= 0.6297Tmin-5.5553
0.85
0.72
Zagora
Ts =0.9108Tmin+3.5965
0.96
0.93
Aghialos
Ts =0.8455Tmin+0.6077
0.87
0.75
Agia
Ts =0.7365Tmin+2.5588
0.85
0.73
Karditsa
Ts =0.9949Tmin+1.6338
0.95
0.88
Tyrnavos
Ts =0.9715Tmin+2.3536
0.98
0.96
Classification of the study area
Correlation between the LST corresponding to every
station and any pixel of the whole Thessaly region.
Selection of the highest correlation for each pixel
 Assignment of each pixel to one of the stations
Classification of the whole area, based on
meteorological stations.
Mapping of Thessaly in six sub- regions.
Result of the Classification
Spatiotemporal extension of the
air temperature data
Combination of two regression equations:
(i. between air temperature and surface temperature and
ii. between pixel corresponding to the meteorological station and other pixels
of the region)
Tmin (x,y) = a΄ Tmin (xi,yi) – a΄ b + ab΄ + b
where:
a΄: slopes from the regression (i)
b΄: intercepts from the regression (i)
a: slopes from the regression (ii)
b: intercepts from the regression (ii)
Tmin (xi,yi): minimum temperature at station’s location.
Validation of the method
April model
(1994-1995-1997-2001)
Έτη
Stations
1994
Obser.
1994
Estim.
1995
Obser.
1995
Estim
1997
Obser.
1997
Estim
2001
Obser.
2001
Estim
Volos
4,56
3,98
5,75
5,07
1,08
1,82
-3,83
-3,67
Zagora
7,35
4,83
1,84
5,84
1,22
1,65
3,76
-2,76
Aghialos
3,84
3,74
1,69
1,78
-0,18
0,08
2,86
2,85
Agia
2,81
2,57
-1,80
-0,60
-1,10
-0,09
1,31
1,55
Karditsa
3,33
3,30
-2,67
-1,50
0,60
1,11
4,34
5,04
Tirnavos
7,30
7,32
2,07
2,38
0,95
1,32
-
-
Comparison between observed and
estimated values.
Correlation between observed and estimated values for April
Correlation between observed and estimated values for March
8
4
0
-9 -8 -7 -6 -5 -4 -3 -2 -1 0
-4
1
2
3
4
5
6
7
8
-8
-12
Observed values
Estimated values
Estimated values
8
4
0
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
7
8
-4
-8
y = 0,863x + 0,4959
R2 = 0,9591
Observed values
y = 0,8973x + 0,4004
R2 = 0,9753
Frost risk mapping
 Definition of surface temperature thresholds
(0ο C, -1o C, -2o C).
 Utilization of 18 images of spatial extension (9
per month) and the classification map.
 Frost probability (%) division to ten (10)
classes for the whole Thessaly region.
Frost risk map
(March - temperature threshold -1o C)
Frost risk map
(April - temperature threshold -1o C)
Frost risk mapping results (April)
Threshold 0ο C
Threshold -1ο C
Threshold -2ο C
Frost
Risk
%
No of
pixels
Percentage
%
No of
pixels
Percentage
%
No of
pixels
Percentage
%
0-10
1943
13,89
3659
26,17
6070
43,41
11-20
1923
13,75
3213
22,98
3283
23,48
21-30
1212
8,67
1094
7,82
351
2,51
31-40
1525
10,91
1582
11,31
1643
11,75
41-50
1140
8,15
688
4,92
448
3,20
51-60
3024
21,62
1499
10,72
599
4,28
61-70
1221
8,73
864
6,18
688
4,92
71-80
59
0,42
24
0,17
41
0,29
81-90
345
2,47
319
2,28
313
2,24
91-100
1592
11,38
1042
7,45
548
3,92
Results
 High correlation between conventional and


satellite data.
Satisfactory pixel by pixel classification of
Thessaly region, according to the temperature
characteristics of the sub-regions.
Satisfactory spatial and temporal extension of
data with average deviation 0.5ο C.
Conclusions
The described procedure:
Identifies the areas with common temperature
characteristics.
Could be a useful tool for the estimation of
minimum air or surface temperature for each
1x1 km pixel.
Could provide accurate information about
frost impact in agriculture.
Recommendations
 Dense network of meteorological stations as well as more
representative stations is required.
 Utilization of minimum correlation threshold for the pixel by
pixel classification (e.g. R2>70%).
 Application of the method to a more satisfactory data series.
 Application of the method to agriculture, crop yielding, as
well traffic protection.
 Extension of the method to whole Greece.