Erik Nelson: How does he care about the data?

1) Brief description of your project

Notes from paper:
- Change in crop mix is one of the most important contributions to the gains in yield.
- Both analyses show that investment in irrigation land, and machinery and equipment, and the quality of cropped soil have had little effect on yield change.  
- One of the concern: if nighttime temperature gets high, crops do not have the chance to rest. 
- Relationship between yield and temperature: downard parabolic.
- Precipitation: can be augmented by irrgation. 

2) His goals from what we understand:

- Changes in yield, in productivity across countries. 
- Split screen, one is for yield, the other one is for temperature.
- Recover absolute value of yield from annual percentage change.

- Possible goal:
+ The relative importance of agricultural inputs to the growth in global and regional crop yields between 1975 and the mid-2000s.
+ How does fertilization affect expected yield curve? 
+ Anything else?

- Some kind of models that we can understand how a country should change crops, with temperature as endogenous variables, to best suit its new condition. 
- A model for suitability measure for a country’s decomposition to its current situation.

3) Data features:

- This is a panel dataset - some observation is repeated again over time.
- How many tons of food a country produced, and divide by the number of hectars that are used in the country.
- Focus on kcal (more relevant than tons):
	+ Include all the ingredients.
- Cannot deduce consumption, since this is domestic production.  
- Mostly numbers
- Categorical by country and climate region.
- What do you think are the most important attributes in this data?
- Where do we obtain data about country's ID according to UNFAO? Maybe "faostat".

- Data that has fertilizer, only lasts until 2002.

** Missing things:
- Amount of sunlight.

Where can we get the decision trees? Ask for supplementary materials.
	+ 12 decision trees were created.
	+ Predict the annual change in a country's yield.
	+ Highlight traversal in each tree with the highest number of records. These traversals indicate the annual changes in agricultural inputs that are most common across space and time.  

Mention about the annual changes dataset (transformed). Can we have those?

4) What are some of our potential ideas:
- Time-lapse visualization with a geomap. Help visualize how crops yield/preferences change over time globally.
	+ Time-lapse is good for time-series, matches conceptual model of the user.
	+ Show changes, which is what we think we want to know.  
- Size of the country change according to crop yield, also with a geo-map.
	+ Gives a better quantitative sense of how the yields are distributed across the globe.
- Another geo-map: The user has control over climate temperature input (night, day, precipitation). The countries that fall into the category would lighten up, and a pie/radar chart will pop up to show the decomposition of yields by averaging them across relevant countries. 
	+ Cool and complicated yay.
	+ Maybe allow the option to show individual countries in the selection with dropdown menu.
- Another geo-map: Filtering timeframe, and light up country by shades. Show average rate of change for certain types of crop.
- Globe idea? 3D.
- Slider.

- Which country are high yield in, which country are low yield in, change the most over time?

Search on yield, crop-type: top 10 oil crop producers, etc. (don't want to do it by area).

*** CHANGE, CHANGE, CHANGE, CHANGE...! That's what driving the whole project.

- Time-lapsed scatter plot. (plot yield versus temperature, with size is land, color is region). hans moving population/income. Cherry and top!

- Show temperature increase but precipitation decreases (drought regions).

- An option to either include fertilizer/or not. (nitrogen fertilizer)

Any other notes?

** ACTUAL QUESTIONS WE WOULD LIKE TO TACKLE:

1) Change in crop mix over a user-selected time range. Possible visualization: maps with chropleth.
2) Change of crop yield over time with respect to temperature. X - temp, Y - Crop yield (kcal/hectar), radius - precipitation (?), time - moving parts, color - region (tropical/temperate). Some kind of animated scatter plot. We can select between night time or day time averages.
3) The big picture: line graphs, but more interactive. Answer the big picture questions: how does different crop mixes change over time. Interactivity: create a grid underneath, pops up, show tooltips.
4)