the gap between the data of harvest projections versus field reality is that the code makes a hole

dua minggu yang lalu

it's been two consecutive seasons, the number of crop projections from the service, and from the models I used to miss myself far from actual. One miss is 20% lower than projection, one times 30% higher. Two different directions in two consecutive seasons.

This is a serious matter of planning. My shipping contract to the buy is based on that projection and the end should either renegotiate or find the volume of another farmer to close the shortfall.

BPS 2023 note the production of Indonesian coffee 795,000 tons but field people know this number is different from the actual international buyer because the way the data collection is still very manual at the village. One of the research agencies has already started the pilot using the NDVI satellite to estimate the harvest but is still in the research stage.

Does anyone have a more religious way of projecting a harvest? That was a real try, not a theoretical one.

288 7 0

satu minggu yang lalu

the way that i use it and quite accurately: green fruit sampling at some point of a representational orchard at the age of 24-26 weeks after interest. counting density per tree, times the conversion factor of the previous season experience. not scientific pure but from a three-year experience, the result is no more than 10-12%.

it's not usually a model, but there's been an unexpected incident: pest attacks, heavy rain in the harvest, or farmers in the same area that harvest early in the plan.

satu minggu yang lalu

i'm from the business side and this is a real problem. suppliers often give overoptimistic estimate numbers at the start and revision at the end. sometimes i suspect that the original number was given so i could lock the contract first.

i've tried asking for monthly progress reports but it's still not religious because farmers don't have good visibility. it's not about intent, but it is that no tools are proxy.

satu minggu yang lalu

i'm curious about the NDVI they call in the post. That's actually how it works. Can we use our own little farmer or should we go through a research agency? And the accuracy is how good is a small-scale farm that's scattered through the mountains?

satu minggu yang lalu

NDVI is a Normalized Divergence Vegetation Index, essentially subtracting health vegetation from the color of plant reflecting via satellite or drone. The ESA sentinels are free and the resolution is 10 meters, accessible via Google Earth Engine is also free. The paid there is a Planet Labs resolution higher up to 3 meters.

For coffee farms, research in Ethiopia and Colombia has proven correlation between NDVI before the harvest season with the yield estimate strong enough, a average error of about 12-18%. Small farmers can't walk themselves, need any technical facilitators or operatives that have this capacity.

satu minggu yang lalu

my farmer's acquaintance on Flores used the simplest way: every year he took photos of garden conditions at the same point in the same month and compared them. It's not scientific, but he can pick up trends whether this season is denser than last year. He said it was pretty help setting up expectations for the buyer. Low tech but more importantly consistent.

satu minggu yang lalu

if i may ask who made this thread, what model or method is it to project? and the contract with the buy-in is a volume adjustment clause if you miss significant or get a direct penalty?

satu minggu yang lalu

use a direct observation combination and rainfall data from the BMKG. contract with buyer usually has a 15% tolerance, more than that goes into the renegotiation zone. So far, no hard penalty, That's why I found a more accurate way.

the gap between the data of harvest projections versus field reality is that the code makes a hole

Write a Reply

Report Thread

Create New Post

Unsaved Writing

Member Info

Sign In to Clorofile Forum