Incorporating regional knowledge into global data sets: Some ideas for rice - PowerPoint Presentation

Slide1

Incorporating regional knowledge into global data sets: Some ideas for rice

in

Asia

Andy Nelson

Thursday11

th

Sep, 2014

Do we need better data? – yes!

Regional and global

modelling efforts need a

ccurate information on the

where

and

when

of staple crops for yield and yield gap estimation, area estimations, risk/exposure assessments, etc.

Commodity research centers, like IRRI, can contribute and improve multi-commodity spatial datasets (SPAM, MIRCA2000 etc.).

Here are some ideas on how regional nodes can go beyond data-sharing to improve data on the

where

and

when

of crops.

The where

of rice

Substantial literature on the use of remote sensing data to map irrigated and lowland rice. Opportunity to apply published methods to derive better information on rice extent in Asia.

Rice is a huge component of cropping systems in Asia, 144m ha. This one crop accounts for

48%

of the cereal crop area in Asia.

Improving information on rice will significantly improve other crop area estimates from SPAM, MIRCA etc.

We applied one algorithm (Xiao et al. 2006) to 13 years (2000-2012) of MODIS data for Asia. 44 tiles x 13 years x 46 tiles/yr and used this to estimate rice extent on a10km x 10km grid.

MODIS tiles – about 44 cover our area

Automated paddy mapping algorithm

Irrigated and lowland

rainfed

rice has a distinctive temporal signature that can indentified using various indices (vegetation and water indices) derived from RS data like MODIS.

Rice paddies are generally flooded at the start of the season and rice biomass increases rapidly after seeding or transplanting.

A simple rule is applied. IF the value of the water index (LSWI) is greater than either of the two vegetation indices (NDVI or EVI) and IF that is quickly followed by a rapid increase in NDVI then the pixel is labeled as rice.

Xiao, X.; Boles, S.;

Frolking

, S.; Li, C.;

Babu

, J. Y.; Salas, W.; Moore, B. Mapping paddy rice agriculture in South and Southeast Asia using multi-temporal MODIS images.

Remote Sens. Environ.

2006

,

100

, 95–113.

Automated paddy mapping algorithm

Xiao, X.; Boles, S.;

Frolking

, S.; Li, C.;

Babu

, J. Y.; Salas, W.; Moore, B. Mapping paddy rice agriculture in South and Southeast Asia using multi-temporal MODIS images.

Remote Sens. Environ.

2006

,

100

, 95–113.

3.1 The where

of rice

Rice area in Asia from SPAM

green: rice dominates the cropped area / gray: some rice in the cropped area

In general the SPAM map has the right spatial pattern but the distribution is too diffuse. There is rice in areas where there is no rice and not enough rice in the ‘rice bowls’

3.1 The where

of rice

Rice area in Asia from IRRI

green: rice dominates the cropped area / gray: some rice in the cropped area

IRRI map shows that the spatial distribution of rice is more focused, dramatically reducing

low density

rice areas and highlighting the main rice producing areas.

3.1 The where

of rice

Rice area in Asia from IRRI

green: rice dominates the cropped area / gray: some rice in the cropped area

Still some issues to address.

Is there that much rice in NW China?

Papua, probably wetland, and not rice?

Missing rice areas in Sumatra?

What is the effect of using this in SPAM?

?

?

?

The when

of rice

I believe that not enough importance is placed on temporal information in cropping systems. It is essential for initializing crop models for yield estimates and climate stress impacts. It is essential for risk management and mitigation. It leads to better rice market information. It also contributes to the improvement of global datasets on cropped area.

Again, regional nodes are best placed to improve existing crop calendars.

In this case, IRRI coordinated an effort across

GRiSP

partners to collect sub-national crop calendars for

all

rice growing countries and seasons in the world.

Around 2,100 spatial units, a 10 fold increase over previous calendars.

162m ha of rice, 113m in the main season, 44m in the second season, 5m in the third season

Jan

Feb

Mar

Apr

May

Jun

jul

Aug

Sep

Oct

Nov

Dec

The

when

of rice

– production by month

90% of production is in Asia (667mt)

Over 50% occurs in just three months

Africa and LAC increasing rapidly

The when

of rice – area by month

IRRI calendar has more rice concentrated in the wet season and much less in the dry season

Again, this new dataset shows less diffuse patterns in rice

The when

of rice – climate impacts.

Preliminary rice crop calendar developed by

GRiSP

. Storm frequency from Joint Typhoon Warning Center (2011). 2011 Annual Tropical Cyclone Report: Western Pacific (Report). United States Navy, United States Air Force.

0

20

40

60

80

100

120

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Rice area under cultivation (M ha) per month

Asia

Others

0

1

2

3

4

5

6

Tropical storm frequency (1959

-

2011)

Key Messages

When

is as least as important as where, but

where

gets most of the attention.

Regional nodes are more than just data providers, they give GEOSHARE access to a bigger network and they can lead on new ways to integrate data.

How to get these data into

Workflows

?

Fusing these data in an elegant and repeatable way will need some further work

GEOSHARE Scientific Committee (GSC)

Can GEOSHARE do more to demonstrate best practice in data sharing in regions where data is still

closed

?

GSC

Thank you

Andy Nelson

a.nelson@irri.org