PLANt
PLANt
PLANt is a digital climate change adaptation tool initiate by UN FAO to reduce risk at planting. UNFAO works with PlantVillage @ PENN STATE to use high-resolution data to perform real time analyses.
Combining weather and climate data provides useful and actionable information for decision-making.
The key to formulating adequate adaptation options for regions that operate close to critical thresholds is looking into short-term climate variability and how it relates to longer-term change. This project is about looking into short-term variabilities within Zambia, Malawi, and other countries for maize crops.
The objective of PLANt is to use high-resolution data to perform real-time analyses; combining weather and climate data provides useful and actionable information for decision making in:
- Choosing the most adapted crop varieties for a location
- Help to identify the best day for planting
- Plan planting activities and services
Zambia agro-ecological regions map updated using observed climate data
PlantVillage contribute to PLANt by:
- Developing a state frequency memory model (SFM) on maize performance under different climate and soil conditions to model maize growth at 250m*250m resolution across all of Zambia. Predictions on crop growth 10, 30, 60, 120 days into the future with the expectation that the further out the time, the less accurate the prediction.
- Detecting Patterns in multiple large datasets. The datasets (SoilGrids, WaPOR, CHIRPS) are publicly available and will be used to model the expected future growth of maize across Zambia. Data will be integrated from breeding trials as well as farmer trials. Data from FAMEWS app will also be used.
- Automation of advice based on models/maps to farmers and extension workers specific to their location.
The clustering of Zambia agro-ecological regions was based on three climate variables and eight soil variables. Those variables were decided after several iterations of discussion with Zambia and FAO partners according to 1) climatic and soil heterogeneity in Zambia; 2) critical yet limiting factors for maize growth in Zambia. Three climate variables are average total rainfall amount in rain seasons from 1981 – 2019, average length of rain seasons from 1981 – 2019, and average start date of rain seasons from 1981 – 2019. Daily precipitation from 1981 – 2019 at the spatial resolution of 5km was used to calculate rain season parameters, based on the rain season criteria suggested by Hachigonta et al., 2008. The onset of a rain season is the first dekad after November 1st with total rainfall of 25mm or above followed by two dekads with total rainfall of 20mm or above each. The first day of the first dekad meeting the onset criteria is the start date of the rain season. The cessation of a rain season is the three earliest and consecutive dekads after Feb 25 with daily rainfall of 2mm or less. The first day of the first dekad meeting the cessation criteria is the end date of the rain season. The rain season length is the number of days between the start and end date of the rain season; and the total rainfall amount during a rain season is the sum of daily rainfall between start and end date of a rain season. These three climate variables on rain season characteristics are critical to choose maize varieties which have proper maturation time and water demand for specific locations, and to determine suitable planting dates for chosen maize variety. Data source and spatial and temporal resolutions of rainfall data are shown in Table 1.
Eight soil variables are soil pH, available water capacity, sand content, organic carbon, nitrogen, exchangeable Al, extractable Fe and extractable Mn contents in top soils. These soil variables are either impacting soil moisture contents or a measurement of soil macro and micro nutrient contents, which are all key factors for estimating maize yields and suggesting farming practices such as fertilizer application. The data source and spatial resolution of soil variables are shown in Table 1.
Table 1. Data source and temporal and spatial resolutions of variables used in Zambia agro-ecological zone clustering.
|
Variable |
Note |
Data Source |
Temporal /Spatial Resolution |
URL |
|
Rainfall amount in rain season, 1981-2019, mm |
calculated from historical daily rainfall |
CHIRPS |
Daily/5km |
|
|
Rain season length, 1981-2019, days |
||||
|
Rain season start time (Julia day of a year), 1981-2019 |
||||
|
pH |
pH measured in water at soil depth of 0-5cm |
ISRIC |
NA/250m
|
|
|
OC, g/kg |
organic carbon content at soil depth of 0-5cm |
|||
|
Nitrogen, mg/kg |
total nitrogen at soil depth of 0-20cm |
|||
|
Available water capacity, % |
available soil water capacity with FC = pF 2.0 at soil depth of 0-5cm |
|||
|
Sand, % |
sand content at soil depth of 0-5cm |
|||
|
Exchangeable Al, cmol/kg |
exchangeable aluminum at soil depth of 0-20cm |
|||
|
Extractable Fe, mg/kg |
extractable iron content at soil depth of 0-30cm |
|||
|
Extractable Mn, mg/kg |
extractable manganese content at soil depth of 0-30cm |
The algorithm used to generate agro-ecological zone clustering based on above environmental variables is CLARA (Clustering Large Applications, (Kaufman and Rousseeuw, 1990)) in R version 4.0.2. CLARA is an extension of k-medoids method with abilities to deal with large size of input data. In this case where the input data was a large raster stack of eleven environmental variables, CLARA was able to reduce computation time to less than 1 minute and save memory space. The core of CLARA, k-medoids method, is an unsupervised clustering algorithm which minimizes the distance between points in a cluster and the center of this cluster. Before running the algorithm, we first resampled all environmental raster layers to the spatial resolution of the precipitation data (5km) and stacked all raster layers together to create a raster stack. This raster stack was then input to CLARA. The number of clusters, a user-defined input parameter to CLARA, was set to be three in this case, considering that 1) the previous Zambia agro-ecological zone map has three zones which can provide a base reference for our new zoning map; and 2) there are in general three categories of maize varieties accessible to farmers.
The resulted clustering map and the average values of eleven environmental variables in each zone were shown in Figure 1 and Table 2, respectively. The resulted clustering map agrees with general knowledge of Zambia climate and soil and previous Zambia agro-ecological zone map. The precipitation decreases from north to south with Zone 3 having greatest rainfall while Zone 1 having the least. Northern areas with greater precipitation also have lower pH and lower sand contents due to leaching, compared to drier areas in the south. Northern areas are also more enriched with soil nutrients (e.g., organic carbon, nitrogen, and Al) except for Fe and Mn.
Figure 1. The final agro-ecological zoning clustering map in Zambia at the spatial resolution of 5km, based on eleven climate and soil variables in Table 2.
Table 2. Average values of eleven environmental variables used in clustering in each zone
|
Zone ID |
1 |
2 |
3 |
|
Rainfall amount in rain season, 1981-2019, mm |
785.31 |
968.82 |
1209.95 |
|
Rain season length, 1981-2019, days |
159.54 |
160.33 |
169.36 |
|
Rain season start time, 1981-2019, Julia day of a year |
316 |
314 |
307 |
|
pH |
5.71 |
5.96 |
5.41 |
|
OC, g/kg |
10.34 |
14.04 |
16.49 |
|
Nitrogen, mg/kg |
0.79 |
0.98 |
0.94 |
|
Available water content, % |
11.05 |
13.52 |
13.52 |
|
Sand, % |
71.27 |
63.27 |
51.95 |
|
Exchangeable Al, cmol/kg |
0.4 |
0.53 |
1.21 |
|
Extractable Fe, mg/kg |
83.07 |
95.13 |
68.3 |
|
Extractable Mn, mg/kg |
140 |
116.09 |
114.49 |