Validation of Near-Real Time Satellite Rainfall Products over
Different Homogeneous Regions of India and Assessment of its
Bias based on Topographical Analysis
Shruti Upadhyaya1; RAAJ Ramsankaran1; Akhil Kumar2
1IIT Bombay, 2IIT BHU
[email protected], [email protected], [email protected]
Introduction
Characteristics of Dataset
- Floods due to extreme rain events pose a major threat
not only to the human life but also have huge impact on
socioeconomic growth of agricultural based countries
like India, which has highly varied climate and
topography.
- For disaster preparedness and flood forecasting over
large river basins, there is a need for proper knowledge
on space and time distribution of rainfall in real time
basis.
- This can be achieved only based on near real time
satellite rainfall estimates (SRE) and verification of
these products are prerequisite to apply SRE for flood
prediction.
- Therefore, a study has been conducted to validate two
Near Real Time High Resolution satellite Precipitation
Products (NRT-HRPP’s) namely, Tropical Rainfall
Measuring Mission- Real Time (TRMM-3B42 RT) and
Insat Multispectral Rainfall Algorithm (IMSRA)
estimates over Indian region.
- As India has highly varied topography, another major
study has been conducted to explore the relationship
between the bias and topography of a region.
Reference Dataset : IMD Gridded daily gauge rainfall dataset developed by Pai et al., 2014 is available at 0.25° x 0.25° spatial resolution.
SRE 1: IMSRA- Insat Multispectral Rainfall Algorithm (IMSRA) is the blended IR/MW satellite rainfall algorithm developed mainly for
Indian region by (Mishra et al., 2010). The data is available at 3 hours temporal resolution and at 0.25° X 0.25° spatial resolution over
Indian region and is available through the website (www.mosdac.gov.in).
SRE 2: TRMM 3B42 RT- The TRMM-3B42 RT is a real time TRMM gridded data product, which is a merged rain product, derived using
Geostationary IR data and Microwave observations (Huffman et al., 2007). The data is available at 3 hours temporal resolution and at 0.25°
X 0.25° spatial resolution and data can be downloaded from (http://disc2.nascom.nasa.gov/Giovanni/tovas/).
The validation has been carried out for the four monsoon seasons (June, July, August, September) for the years 2010-2013for Indian region.
Results
Objectives
1-Evaluation of detection and estimation capabilities of
two SRE’s at daily scale for different climate regions of
India.
2-Assessment of Biases of two SRE’s based on
Topographical Analysis
Methodology
Statistical Validation
Extraction of IMSRA and TRMM
3B42RT for Indian land region
Statistical Validation
Separating the dataset for different
KÖppen-Geiger climate regions
Topographical Analysis
Table 3. Topographic Variables Extracted from DEM and its Loading Values:
(original variables represented by the six Principal Components (PCs) as they
explain more than 90% of the variance of the original topographic data set )
Estimation of statistics (Table 1)
Sl no
1
2
3
4
5
6
7
3B42 RT
Table 1. Statistics used for validation and its description
STATISTICS
FORMULAE
RANGE
𝐵𝑎𝑖𝑠
𝑓+ℎ
𝑚+ℎ
-∞ to
+∞
ℎ
𝑚+ℎ
𝑃𝑂𝐷
𝑧
𝑧+𝑓
𝑃𝑂𝑁𝐷
0 to 1
1
𝑛
𝐵𝑎𝑖𝑠 i
𝒏
𝒊=𝟏
1
SRE underestimates or
overestimates the number
of rainy pixels
1
Fraction of times the
reference rainy pixels are
correctly detected by the
selected index
1
0 to 1
𝐶−𝐸
𝑁−𝐸
𝐻𝑆𝑆
DESCRIPTION
0
-1 to 1
𝑛
-∞ to
+∞
𝑆𝑖 − 𝐺𝑖
(𝑮𝒊 − 𝑮)(𝑺𝒊 − 𝑺)
− 𝑮)𝟐 ×
10
North Aspect
11
East Aspect
Southeast
Aspect
-0.452
West Aspect
0.458
-
𝒏
𝒊=𝟏(𝑺𝒊
− 𝑺)𝟐
-1 to +1
Fraction of times the norain pixels are incorrectly
detected by the reference
no-rain pixels
1
HSS measures forecast
accuracy relative to that
of a random chance.
1
SRE underestimates or
overestimates the rainfall
value
+/-1
Measures linear
association between SRE
and reference data
𝑖=1
𝒏
𝒊=𝟏(𝑮𝒊
9
8
12
𝑓
𝑓+ℎ
𝐹𝐴𝑅
CC
0 to 1
BEST
VALUE
Variables
Mean Slope
Range
SD
Min
Max
Mean
Sum
Northeast
Aspect
South Aspect
13
14
KÖppen-Geiger climate classification (Peel et al., 2007)
15
Region 1: Am- Tropical Monsoon
Region 2: Bwk- Arid Desert Cold
Region 3: Cwa- Temperate Dry Winter Hot Summer
Region 4: Aw- Tropical Savannah
Region 5: Bsh- Arid Steppe Hot
Region 6: Bwh- Arid Desert Hot
Region 7: Cwb- Temperate Dry Winter Warm Summer
Where: POD: Probability of Detection, POND: Probability of No-Rain Detection, FAR:
False Alarm Ratio, HSS: Heidke Skill Score, CC: Correlation Coefficient, 𝐺 : gauge
observations and 𝐺 ∶the average of gauge observations. 𝑆𝑖 and 𝑆 are satellite estimates
and their average, respectively. h, f, m and z are defined using contingency table given
in Table 2.
16
17
18
19
20
Northwest
Aspect
Flat Aspect
Southwest
Aspect
Min Relief
Max Relief
SD Relief
Mean Relief
PC1
0.274
0.329
0.325
0.275
0.338
0.318
0.318
PC2
PC3
PC4
-0.287
PC5
PC6
0.305
-0.264
-0.441
0.479
0.262
-0.466
0.249
-0.323
-0.347
-0.453
0.448
-0.346
-0.253
0.457
0.821
0.441
-0.296
0.318
0.328
0.412
0.726
Table 4. Linear Regression equation and its correlation coefficient between PCs
and Bias of SRE’s
TRMM
IMSRA
Bias=1.7756 -0.16615 PC1-0.16303 PC2+ 0.034579 PC3+ 0.089217 PC4 CC=0.031
0.19015 PC5 -0.30026 PC6
Bias=-0.5673-0.24643PC1-0.61773 PC2 -0.78816 PC3+ 0.3838 PC4
CC=0.191
+0.86262 PC5 -0.18636 PC6
Table 2. Contingency Matrix
Major Conclusions
Reference Dataset
Estimated from
Class
Yes (Rain)
No (No-Rain)
Yes (Rain)
Hits(h)
False alarms(f)
SRE
Correct
No (No-Rain)
Misses(m)
Negatives(z)
Topographical Analysis
DEM
Extract Topographic features (Table 3)
within each 0.25° X 0.25° grid
Principal Component (PC) Analysis
to reduce redundancy
Linear Regression model is employed
to interpret SRE Biases based on PC
Gao and Liu, 2013
- TRMM 3B42 RT performs comparatively better than IMSRA algorithm.
- Considering the complexity and robustness involved in the TRMM 3B42 RT algorithm w.r.t the simplicity of IMSRA, it can be said that
the accuracy of the TRMM 3B42 RT product is still poor.
- IMSRA algorithm can be further improved by including more proper input parameters which affect rain rates without making it much
complicated as TRMM 3B42 RT algorithm which involves huge data handling. One such parameter would be orography as IMSRA
shows more dependency on topographic variables than TRMM 3B42 RT
- Topographic analysis also revealed that not only elevation plays important roles in explaining biases but also aspect and local relief also
plays important role.
References
-
-
Gao, Y. C. and Liu. M. F., 2013: Evaluation of high-resolution satellite precipitation products using rain gauge observations over the Tibetan Plateau. Hydrology and Earth System Sciences 17: 837–849.
Huffman, G. J., Adler, R. F., Bolvin, D. T., Gu, G., Nelkin, E. J., Bowman, K. P., Hong, Y., Stocker, E. F., and Wolff, D. B., 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): quasi-global, multiyear, combined-sensor
precipitation estimates at fine scales, Journal of Hydrometeorology 30: 38–55.
Mishra, A., Gairola, R. M., Varma, A. K., and Agarwal, V. K., 2010: Remote sensing of Precipitation over Indian land and oceanic regions by synergistic use of multi-satellite sensors. Journal of Geophysical Research 115: D08106
Peel, M. C., Finlayson, B. L. and MacMahon, T. A., 2007: Updated world map of the KÖppen-Geiger climate classification. Hydrology and Earth System Sciences 11: 1633–1644.
Pai D.S., Latha Sridhar, Rajeevan M., Sreejith O.P., Satbhai N.S. and Mukhopadhyay B., 2014: Development of a new high spatial resolution (0.25° X 0.25°)Long period (1901-2010) daily gridded rainfall data set over India and its
comparison with existing data sets over the region; MAUSAM, 65, 1(January 2014), pp1-18.